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
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-5-291557595410.1186/1471-2296-5-29Research ArticleWalkability and self-rated health in primary care patients Rohrer James [email protected] JR [email protected] Anne [email protected] Department of Family and Community Medicine, Texas Tech University Health Sciences Center, Health Services Research, 1400 Wallace Blvd. Amarillo, Texas, 79106, USA2 Womens' Health and Research Institute, 1400 Wallace Blvd, Amarillo Texas 79106, (PH) 806-354-5786, (FX) 806-356-59083 Department of Internal Medicine, Texas Tech University Health Sciences Center, Amarillo, Texas USA4 Texas Tech University Health Sciences Center, Amarillo, Texas USA2004 2 12 2004 5 29 29 30 6 2004 2 12 2004 Copyright © 2004 Rohrer et al; licensee BioMed Central Ltd.2004Rohrer 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 objective of this study was to investigate the relationship between perceived walkability and overall self-rated health among patients who use community-based clinics.
Methods
A cross-sectional survey was distributed to a convenience sample in three community clinics. Forms were completed by 793 clinic patients. Multiple logistic regression analysis was to control for the effects of demographic variables and lifestyles.
Results
Perceiving the availability of places to walk was related to better self-rated health. The most important places were work (OR = 3.2), community center (OR = 3.12), park (OR = 2.45) and day care (OR = 2.05). Respondents who said they had zero (OR = .27) or one (OR = .49) place to walk were significantly less healthy than persons who said they had five or more places to walk.
Conclusion
Persons who perceived that they had no place to walk were significantly less healthy than persons who thought they had at least one place to walk (OR = .39). Support for walkable neighborhoods and education of patients about options for walking may be in the best interests of community medicine patients.
==== Body
Background
Much recent interest and research has been directed at the relationship between one's health and where one lives [1]. A number of measures of health (including all-cause mortality [2,3], low infant birth weight [4], unintentional childhood injuries [5], hospitalization for asthma [6], and the risk for certain arboviral diseases [7,8] have been associated with neighborhood effects that are independent of demographic health markers such as age, gender, race, and poverty. The neighborhood environment appears to exert health effects independent of or in addition to the health behaviors of the neighborhoods inhabitants [9].
Neighborhoods where walking is convenient might encourage their inhabitants to exercise. Indeed, having convenient places to walk in the neighborhood has been related to the proportion of persons in the neighborhood who met current activity recommendations [10] and with a decreased prevalence of overweight [11]. Neighborhoods with convenient places to walk are characterized by good "walkability." Walking is now recommended for the prevention and treatment of many common diseases, such as hypertension, diabetes, coronary heart disease, osteoporosis, colon cancer, and obesity [12]. Therefore, one would postulate that neighborhoods characterized by good walkability would be inhabited by healthier residents.
The hypothesis that perceived walkability is directly related to an individual's overall self-rated health has not previously been investigated. The purpose of this project was to test that hypothesis in patients attending community-based clinics. Self-rated health has been shown to accurately predict overall health as measured by other more traditional measures of health [13].
Methods
A cross-sectional survey was used to test the hypothesis that patients with good self-rated health perceived the neighborhoods in which they live as having good walkability, and that this effect was independent of demographic characteristics and lifestyle habits that would predict good health. The sample was drawn from three clinics that primarily serve low-income populations. Clients, parents (if the patient was a child) and accompanying visitors were asked to complete surveys and drop them into a box. Participation was voluntary. A total of 1471 surveys were distributed with an overall return of 825 (56.1%) surveys of which 793 (54%) met eligibility parameters. Pregnant women and persons under age 18 were excluded. Assuming 80% power, p < .05, 20% poor health among the persons inhabiting neighborhoods with good perceived walkability and 30% poor health among persons inhabiting neighborhoods with poor perceived walkability, 626 cases were needed to test the hypothesis. Completed forms were received from a total of 793 persons.
Return rates varied by clinic. Clinic 1 is a university-based family medicine clinic providing a full range of primary care services to cross-generational clients. It is staffed by family medicine physicians and residents. Census was approximately 85 clients daily, of which, less than 5 % were non-English speaking. Clinic personnel distributed 500 survey forms over an eight week period with an 80.8% return rate.
Clinic 2 serves women and children, providing obstetrical, well care (including immunizations), and acute care services to a targeted high-risk, low socioeconomic sub-population. It is staffed by pediatric and OB-GYN physicians and residents. Approximately 30% of the clinic clients do not speak English. A total of 471 surveys were distributed over a period of 18 weeks with a return of 37.4%. Both the large number of obstetrical patients (ineligible for survey) and percentage of non-English speaking clients contributed to the low return rate.
Clinic 3 provides primary care services to a population of indigent adults meeting residential and income screening requirements. It is staffed by internal medicine physicians and residents. A total of 500 surveys were distributed in this clinic over a ten week period with a return of 42.6%.
The dependant variable for the study was self-rated health. Subjects were asked whether in general they would say their health was excellent, very good, good, fair or poor. Excellent, very good, and good responses were combined to form a category called 'good health' while fair and poor comprised 'poor health'.
Leyden's scale of walkability was modified for a US population for this study[11] by dropping the terms newsagent, chemist, and crèche. The question reads: "A lot of people are very dependent on a car these days to get where they want to go. If you or another family member wanted to which of the following could you walk to without too much trouble. Circle all you could walk to without too much trouble." Possible answers were a local corner shop, a church, a park, a local school, a community center or recreation center, a day care center, a drug store, a bar or pub, the place that I work, or "none of the above. It is really hard to go anywhere without a car."
Both demographic characteristics and lifestyle variables were used to adjust the associations between perceived walkability and self-rated health. Lifestyle variables were: numbers of fruits and vegetables eaten per day (zero, one, two, three, four, five or more), smoking status (not a smoker, smokes one-20 cigarettes per day, smokes more than 20 per day), days of physical activity per week that involve at least 20 minutes of exercise (zero, one, two, three, four, five or more), and obese (yes vs no). Obese was defined as body mass index (BMI) >30. BMI was computed from self-reported height and weight.
Demographic variables were race/ethnicity (Hispanic, non-Hispanic white, non-Hispanic black, non-Hispanic Asian, non-Hispanic other), gender, age category, marital status (married vs. not), and highest level of education achieved.
Chi square tests were performed to test for any unadjusted associations between self-rated health and each categorical independent variable. Multivariate logistic regression modeling was employed to determine if associations between perceived walkability and self-rated health remained significant after adjustment for demographic and lifestyle variables. Separate logistic regression models were run for each variable. Statistical analysis was performed using EpiInfo 3.2.2.
Results
The question about self-rated overall health was answered by 793 respondents. Of these, 67 percent were classified as healthy because they said their health was excellent, very good, or good. The remainder was classified as having poor health because they said their health was fair or poor.
The typical respondent was non-Hispanic white, female, had at least a high school education. The sample was evenly spread across age groups. About half were married and about half were not married (Table 1).
Table 1 Association between good self-rated health and demographic variables (chi-square)
Variable Overall Percent Percent Healthy Pct Not Healthy P
Age 0.0717
18–25 17.0 77.6 22.4
26–35 21.6 66.5 33.5
36–45 17.0 64.2 35.8
46–55 19.1 64.0 36.0
56–65 12.3 69.1 30.9
Over 65 12.8 60.4 39.6
Race/ethnicity 0.0303
Hispanic 15.5 58.5 41.5
NH* asian .5 75.0 25.0
NH* black 7.8 54.8 45.2
NH* white 74.9 69.9 30.1
NH* other 1.3 70.0 30.0
Gender 0.4852
Male 18.7 64.2 35.8
Female 81.3 67.6 32.4
Education 0.0011
Less than high school 9.0 54.9 45.2
High school 37.8 62.0 38.0
Some College 37.7 69.6 30.4
College degree 15.5 79.7 20.3
Married <0.0001
Yes 49.8 74.7 25.3
No 50.2 59.3 40.7
*NH = Non-Hispanic
Unadjusted relationships between demographic variables and self-rated health are shown in Table 1. Statistical significance was determined using chi-square tests. As expected, the percent healthy declined with age. Non-Hispanic white respondents were more likely to report good self-rated health than Hispanics or non-Hispanic blacks. The percent healthy increased with level of education and married persons were more likely to be healthy. No significant difference was seen between male and female subjects in self-rated health.
The most common number of days of exercise per week was zero (34.9 percent). Most people said they ate at least one fruit or vegetable each day, though only less than half had three or more. The percent obese was 43.6. Over 70 percent were non-smokers (see Table 2).
Table 2 Association between good self-rated health and lifestylevariables (chi-square)
Variable Overall Percent Percent Healthy Pct Not Healthy p
Days of exercise per week 0.0104
Zero 34.9 62.9 37.1
One 8.0 52.4 47.6
Two 16.6 67.9 32.1
Three 18.8 72.3 27.7
Four 5.8 80.4 19.6
Five or more 16.0 71.4 28.6
Fruit and vegetables 0.1522
Zero 15.2 62.5 37.5
One 19.6 70.8 29.2
Two 27.6 68.7 31.3
Three 20.1 66.5 33.5
Four 9.9 56.4 43.6
Five or more 7.6 75.0 25.0
Obese <0.0001
Yes 43.6 58.1 41.9
No 56.4 73.8 26.2
Smoking Status 0.1496
Non-smoker 72.0 68.9 31.1
One – 20 per day 26.1 61.8 38.2
Over 20 per day 1.9 60.0 40.0
Self-rated health was related to lifestyle variables (Table 2). Persons who exercised more times per week were more likely to report good self-rated health. Obese persons were less likely to be healthy. Smoking and consumption of fruit and vegetables were not significantly related to self-rated health, though more smokers reported poor health.
The association between perceiving particular places to walk and self-rated overall health is shown in Table 3. Respondents typically said they could not walk to the places included in the walkability scale. The highest percents were corner store (43.1), park (44.0), and school (40.9). When the association between particular places to walk and self-rated health was adjusted for the demographic and lifestyle variables, eight were significant: workplace (OR = 3.2, p = .0011), church (OR = 1.76, p = .0031), community center (OR = 3.12, P = .0019), corner store (OR = 1.71, p = .0032), day care (OR = 2.05, p = .0216), drug store (OR = 1.88, p = .0055), park (OR = 2.45, p < 0.0001), and school (OR = 1.86, p = .0008).
Table 3 Association between Good Self-Rated Health and Places to Walk (N = 793)
Variable Overall Percent Percent Healthy Pct Not Healthy p Ajusted Odds Ratio (Confidence Interval)* p
67.0 33.0
Pub 0.5043
Yes 10.1 72.5 27.5 1.38 (.77–2.48) 0.2750
No 83.9 66.2 33.8 reference
Workplace 0.0146
Yes 8.7 82.6 17.4 3.20 (1.5955–6.4271) 0.0011
No 85.1 65.3 34.7 reference
Church 0.0010
Yes 32.2 75.7 24.3 1.76 (1.21–2.57) 0.0031
No 61.8 62.2 37.8 reference
Community Center 0.0030
Yes 9.6 84.2 15.8 3.12 (1.52–6.38) 0.0019
No 84.4 64.9 35.1 reference
Corner Shop 0.0002
Yes 43.1 74.6 25.4 1.71 (1.20–2.44) 0.0032
No 50.8 60.3 39.7 reference
Day Care 0.0176
Yes 11.3 80.0 20.0 2.05 (1.11–3.78) 0.0216
No 82.6 65.0 35.0 reference
Drug Store 0.0037
Yes 21.7 77.3 22.7 1.88 (1.20–2.93) 0.0055
No 72.3 63.7 36.3 reference
Park <0.0001
Yes 44.0 76.8 23.2 2.45 (1.69–3.55) <0.0001
No 49.9 58.1 41.9 reference
School <0.0001
Yes 40.9 76.2 23.8 1.86 (1.29–2.67) 0.0008
No 53.1 59.6 40.4 reference
*adjusted for marital status, age, gender, obesity, smoking status, days of exercise per week, educational level, race/ethnicity using multiple logistic regression (N = 775)
Table 4 shows the association between the number of places to walk and self-rated health. Nearly 30 percent of respondents reported that they had no places to which they might walk. Using five or more places to walk as the reference category and adjusting for demographic variables and lifestyle, persons with no places to walk had lower odds of being in good health (OR = .27, p < 0.0001). Persons who had only one place to walk also were not as healthy (OR = .49, p = .0257). The other levels of walkability were not significantly different from five or more places.
Table 4 Association between Good Self-Rated Health and Number of Places to Walk
Variable Overall Percent Percent Healthy Pct Not Healthy p Ajusted Odds Ratio (Confidence Interval)* p
67.0 33.0
Walk category <0.0001
None 27.9 51.1 48.9 .27 (.16–.47) <0.0001
One 14.6 65.5 34.5 .49 (.26–.92) 0.0257
Two 11.6 75.0 25.0 .76 (.37–1.54) 0.4414
Three 12.5 72.7 27.3 .66 (.34–1.30) 0.2281
Four 10.2 72.8 27.2 .72 (.35–1.45) 0.3505
Five + 17.2 80.1 19.9 reference
Missing 6.1 68.8 31.3 .59 (.26–1.30) 0.1898
No place to walk <0.0001
Yes 27.9 51.1 48.9 .39 (.27–.57) <0.0001
No 72.1 73.1 26.9 reference
*adjusted for marital status, age, gender, obesity, smoking status, days of exercise per week, educational level, race/ethnicity using multiple logistic regression (N = 775)
Also shown in Table 4 is the result of a multiple logistic regression analysis in which perceived walkability is scored simply as none versus more than none. In this model, the adjusted odds ratio for perceived walkability was .39 (p < 0.0001)
Discussion
Studies of physical activity in public health may be classified according to the intensity of physical activity on which they focus (the "meet recommended guidelines" group versus the "any activity will do" group). Meeting recommended physical activity guidelines is related to better self-rated health [14]. Certainly, more fitness is better, but arguments can be made against setting high standards for physical activity in the population. Foremost among these is the observation that effective interventions to promote high levels of fitness in the general population may not be available. Harlan reported that a brief intervention in primary care was not effective [15], as did Yeazel [16]. In contrast, Long's study supported primary care based promotion of physical activity [17] and Eakin also offered support [18]. Overall, most studies of the elderly do not show that exercise reduces disability [19].
If promotion of intensive physical activity is problematic, then more moderate activity might be a more reasonable goal. Simple walking, for example, reduces the cost of medical care for the elderly [20] and interventions have been developed to increase walking [21]. Besides, sedentary adults may not be able to accurately recall the intensity of physical activity, casting doubt on the accuracy with which it can be measured [22].
In addition to valuing leisure-time walking, researchers and clinicians should take into account the benefits of work related activity and housework [23]. So called "lifestyle" interventions are more cost-effective than supervised center-based exercise [24]. In short, encouraging moderate physical activity is important. Successfully doing so in the most sedentary and unfit portion of the population (the bottom 20 or 25 percent) would generate large gains in population health [25].
One approach to encouraging moderate physical activity is to increase walkability. The convenience of places to exercise is widely recognized to be important; adults generally support local policies that increase the availability of places to exercise [26-28]. Walkability in Georgia was investigated by Powell et al, using the state-wide Behavioral Risk Factor Surveillance System (BRFSS) [10]. Over 90 percent of Georgians reported that they knew of a place where they felt safe walking. The most common place was the respondent's neighborhood (32%). About half said that they could get to their walking place in less than ten minutes. A direct relationship was found between convenience of the walking place and the proportion of respondents meeting current activity recommendations. Better health should result from more walking, but this relationship was not tested in the Georgia study.
Our purpose was to investigate the effect of perceived walkability on the health of patients attending community based clinics. The results should not be interpreted as being relevant to all persons in the community or even to all persons who live in neighborhoods they perceive to be unwalkable, but only to patients attending clinics that serve a disproportionate share of disadvantaged persons. This is important to public health because the success of strategies designed to improve in overall community health will have to focus on the disadvantaged people who bear the greatest burden of poor health. We believe that more studies of disadvantaged clinic populations such as this one are needed in the public health literature.
Conclusions
The study reported here of low income primary care patients confirms the relationship between perceptions of convenient walking locations and self-rated health. Health status is measured by just one question, a practice which has become increasingly common in the public health literature[29,30]. However, since the study employs a cross-sectional design the relationships between perceived walkability and good health might be due to an omitted third variable, such as a tendency to look for excuses for inactivity among persons of poor health. Or a tendency for negative persons to report no places to walk may have influenced the results of this study. In addition, since a large proportion of our respondents were female, the results might be less generalizable to male populations.
Another limitation of the study is the subjective nature of the walkability measure. Perceptions of walkability may not be accurate and thus objective assessment of walkability might have led to a different conclusion. However, since our purpose was to investigate the relationship between perceptions of walkability and self-rated health, the results are valid for this sample. If community medicine patients are incorrect in their perceptions of neighborhood walkability, then public health education campaigns and personal health education in clinics can inform them about options for walking in their neighborhoods and how they might overcome any perceived barriers.
Our results also may have been influenced by the context from which subjects were drawn. Amarillo is located in the Panhandle of Texas. Most of the year the climate is mild, but during the summer season temperatures can average over 90 degrees. In some neighborhoods, crime rates are above the national average and walking after dark might be worrisome. Furthermore, cities in West Texas are designed for automobile traffic, to the detriment of pedestrians and bicyclists. Furthermore, the vast open spaces in the region make it impossible to reach many locations by foot. Accordingly the, culture of the area has not evolved to be supportive of walking. Therefore, our results may not be generalizable to communities where walking has historical been a visible aspect of the culture.
Despite these limitations, the results of this analysis are intriguing and warrant further investigation using a prospective study design. The evidence presented suggests that support for walkable neighborhoods and health education about options for walking in their neighborhoods may be in the best interests of community medicine patients.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JER planned the project, analyzed the data and wrote the first draft of the manuscript. JRP contributed comments and revisions to the manuscript. AD organized data collection and wrote a portion of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This manuscript was partially supported by funding from the City of Amarillo Department of Public Health. Conclusions are the authors' own and do not reflect official policy of the City of Amarillo.
==== Refs
Kawachi I Berkman LF eds Neighborhoods and Health 2003 Oxford: Oxford University Press
Kawachi I Berkman LF Kawachi I, Berkman LF Chapter 1. Introduction Neighborhoods and Health 2003 Oxford: Oxford University Press 1 19
Kawachi I Kennedy BP Income inequality and health: pathways and mechanisms Health Serv Res 1999 34 215 227 10199670
Collins J Schulte NF Drolet A Differential effect of ecologic risk factors on low birthweight component of African-American, Mexican-American, and non-Latino white infants in Chicago J Natl Med Assoc 1998 90 223 232 9581442
Shenessa ED Stubbendick A Beown MJ Social disparities in housing and related pediatric injury: a multilevel study Am J Public Health 2004 94 633 639 15054017
Wright RJ Fischer EB Kawachi I, Berkman LF Chapter 11. Putting asthma into context: community influences on risk, behavior, and intervention Neighborhoods and Health 2003 Oxford: Oxford University Press 233 264
Ruiz MO Tedesco C McTighe TJ Austin C Kitron U Environmental and social determinants of human risk during a West Nile Virus outbreak in the greater Chicago area, 2002 International J Health Geographics 2004 3 8 10.1186/1476-072X-3-8
Luby JP Miller G Gardner P Pigford CA Henderson BE Eddins D The epidemiology of St. Louis encephalitis in Houston, Texas, 1964 Am J Epidemiol 1967 86 584 597 6081383
MacIntyre S Ellanay A Kawachi I, Berkman LF Chapter 2. Neighborhoods and health Neighborhoods and Health 2003 Oxford: Oxford University Press 20 44
Powell KE Martin LM Chowdhury PP Places to Walk: Convenience and Regular Physical Activity Am J Public Health 2003 93 1519 1521 12948973
Leyden KM Social capital and the built environment: the importance of walkable neighborhoods Am J Public Health 2003 93 1546 51 12948978
U.S. Dept of Health and Human Services Physical Activity and Health A report of the Surgeon General 1996 Washington DC
Rohrer JE Arif AA Pierce JR JrBlackburn CB Unsafe neighborhoods, social group activity, and self-rated health J Public Health Manag Pract 2004 10 124 129 14967979
Brown DW Balluz LS Health GW Moriarty DG Ford ES Giles WH Mokdad AH Associations between recommended levels of physical activity and health-related quality of life. Findings from the 2001 Behavioral Risk Factor Surveillance System survey Prev Med 2003 37 520 528 14572437 10.1016/S0091-7435(03)00179-8
Harland J White M Drinkwater C Chinn D Farr L Howel D The Newcastle exercise project: a randomised controlled trial of methods to promote physical activity in primary care BMJ 1999 319 828 832 10496829
Yeazel MW Bunner SH Kofron PM Weiss PJ Put prevention into practice (PPIP): evaluating PPIP in two family practice residency sites Fam Med 2002 34 17 22 11838522
Long BJ Calfas KJ Wooten W Sallis JF Patrick K Goldstein M Marcus BH Schwenk TL Chenoweth J Carter R Torres T Palinkas LA Heath G A Multisite Field Test of the Acceptability of Physical Activity Counseling in Primary Care: Project PACE Am J Prev Med 1996 12 73 81 8777071
Eakin EG Glasgow RE Riley KM Review of primary care-based physical activity intervention studies effectiveness and implications for practice and future research J Fam Pract 2000 49 158 68 10718694
Keysor JJ Jette AM Have we oversold the benefit of late-life exercise? J Gerontol A Biol Sci Med Sci 2001 56 M412 23 11445600
Perkins AJ Clark DO Assessing the association of walking with health services use and costs among socioeconomically disadvantaged older adults Prev Med 2001 32 492 501 11394953 10.1006/pmed.2001.0832
Humpel N Marshall AL Iverson D Leslie E Owen N Trial of telephone delivered interventions to influence walking Prev Med 2004 39 635 641 15313106 10.1016/j.ypmed.2004.02.032
Duncan GE Sydeman SJ Perri MG Limacher MC Martin AD Can sedentary adults accurately recall the intensity of their physical activity? Prev Med 2001 33 18 26 11482992 10.1006/pmed.2001.0847
Brownson RC Eyler AA King AC Brown DR Shyu YL Sallis JF Patterns and correlates of physical activity among US women 40 years and older Am J Public Health 2000 90 264 70 10667189
Sevick MA Dunn AL Morrow MS Marcus BH Chen GJ Blair SN Cost-effectiveness of lifestyle and structured exercise interventions in sedentary adults: results of project ACTIVE Am J Prev Med 2000 19 1 8 10865157 10.1016/S0749-3797(00)00154-9
Blair SN Connelly JC How much physical activity should we do? The case for moderate amounts and intensities of physical activity Res Q Exerc Sport 1996 67 193 205 8835998
Brownson RC Baker EA Housemann RA Brennan LK Bacak SJ Environmental and policy determinants of physical activity in the United States Am J Public Health 2001 91 1995 2003 11726382
Rütten A Abel T Kannas L von Lengerke T Lüschen G Rodríguez-Diaz JA Vinck J van der Zee J Self reported physical activity, public health, and perceived environment: results from a comparative European study J Epidemiol Community Health 2001 55 139 146 11154254 10.1136/jech.55.2.139
Wang G Macera CA Scudder-Soucie B Schmid T Pratt M Buchner D Heath G Cost Analysis of the Built Environment: The Case of Bike and Pedestrian Trials in Lincoln, Neb Am J Public Health 2004 94 549 553 15054000
Rohrer J Borders T Healthy Skepticism Prev Med 2004 39 1234 1237 15539061 10.1016/j.ypmed.2004.04.038
Rohrer JE Young R Self-esteem, stress and self-rated health in family planning clinic patients BMC Family Practice 2004 5 11 15176984 10.1186/1471-2296-5-11
| 15575954 | PMC539238 | CC BY | 2021-01-04 16:28:59 | no | BMC Fam Pract. 2004 Dec 2; 5:29 | utf-8 | BMC Fam Pract | 2,004 | 10.1186/1471-2296-5-29 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-4-541556939310.1186/1471-2334-4-54Research ArticleRegional risks and seasonality in travel-associated campylobacteriosis Ekdahl Karl [email protected] Yvonne [email protected] Department of Epidemiology, Swedish Institute for Infectious Disease Control (SMI), SE-17182 Solna, Sweden2 Department of Medical Epidemiology and Biostatistics, Karolinska Institute, SE-17177 Stockholm, Sweden2004 29 11 2004 4 54 54 13 9 2004 29 11 2004 Copyright © 2004 Ekdahl and Andersson; 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.
Backgound
The epidemiology of travel-associated campylobacteriosis is still largely unclear, and various known risk factors could only explain limited proportions of the recorded cases.
Methods
Using data from 28,704 notifications of travel-associated campylobacteriosis in Sweden 1997 to 2003 and travel patterns of 16,255 Swedish residents with overnight travel abroad in the same years, we analysed risks for travel-associated campylobacteriosis in 19 regions of the world, and looked into the seasonality of the disease in each of these regions.
Results
The highest risk was seen in returning travellers from the Indian subcontinent (1,253/100,000 travellers), and the lowest in travellers from the other Nordic countries (3/100,000 travellers). In Africa, large differences in risk between regions were noted, with 502 /100,000 in travellers from East Africa, compared to 76/100,00 from West Africa and 50/100,000 from Central Africa. A distinct seasonal pattern was seen in all temperate regions with peaks in the summer, while no or less distinct seasonality was seen in tropical regions. In travellers to the tropics, the highest risk was seen in children below the age of six.
Conclusions
Data on infections in returning travellers together with good denominator data could provide comparable data on travel risks in various regions of the world.
==== Body
Background
Campylobacter infection is a zoonotic disease, observed in most parts of the world. The disease is caused by Campylobacter jejuni, or less commonly Campylobacter coli. It is estimated to cause 5–14% of diarrhoea, worldwide [1]. The incubation period is usually 2 to 5 days (range 1 to 10 days), and persons not treated with antibiotics may excrete the organisms for as long as 7 weeks [2]. Also in the Western world Campylobacter infection has emerged to be most important bacterial cause of gastrointestinal infection. Animals (variety of fowl, swine, cattle, sheep, dogs, cats, and rodents) are the major reservoir for the bacteria. Campylobacter does not easily grow in food, but the critical infective dose is low [3].
Unlike salmonellosis with well-known routes of transmission, the epidemiology of campylobacteriosis is still largely unclear, and various known risk factors could only explain limited proportions of the recorded cases [4]. Known risk factors for the disease include ingestion of undercooked meat, contaminated food and water or raw milk, direct contact with pets, farm animals and small children, and swimming in lakes, but also travel abroad [3,5-7]. Direct person-to-person transmission between adults appears to be uncommon. In temperate regions, campylobacteriosis has a distinct seasonal pattern, with the peak incidence in the summer months [4,6,8,9], but seasonal data on campylobacteriosis from tropical regions are scarce.
Approximately 80 million persons from industrialized countries travel every year to places in Africa, Asia, Pacific Islands, Latin America and remote areas of Eastern Europe, and between, 25 and 50 % of travellers to these areas experience travellers' diarrhoea [10-12]. About 80 % of all episodes of traveller's diarrhoea have a bacterial cause, and Campylobacter infection is a leading cause together with infections due to enterotoxigenic Escherichia coli (ETEC), salmonellosis, and shigellosis [11,13].
In this study we have used returning travellers to Sweden as sentinels to estimate the comparative risks for travel-associated campylobacteriosis in 19 regions of the world, and looked into the seasonality of the disease in each of these regions.
Methods
Notification data on campylobacteriosis
Campylobacteriosis is a notifiable disease according to the Swedish communicable disease act. Cases are notified in parallel to the Swedish Institute for Infectious Disease Control (SMI) by the clinician having seen the patient (clinical notification) and the laboratory having diagnosed the pathogen (laboratory notification). At the SMI the notifications from the two sources are merged into case records, using a unique personal identification number issued to all Swedes, and used in all health care contacts. The clinical notifications contain epidemiological information of relevance, including country of infection. For this study we retrieved notification information (age, sex, country of infection and month of infection) from the national surveillance database [14] on all cases of campylobacteriosis notified in the period 1997–2003, with country of infection outside Sweden. All information in the database is derived from the notifications, and the data (including "country of infection") are thus based on the best judgment of the notifying clinician based on the patient history and knowledge of the characteristics of the pathogen in question. Since we focused on travel-associated infections, refugees and newly arrived immigrants (with incomplete personal identification number) were sorted out before analysis.
Denominator data on travel patterns
Data on travel patterns were obtained from a commercial database, the Swedish Travel and Tourist Database (TDB) [15]. This database contains data from monthly telephone interviews with 2,000 randomly selected Swedish residents, with travel-related questions. Out of the total database, containing data from almost 170,000 interviews, we used 16,255 records of persons with recent overnight travel outside Sweden. Each record included information on principal geographical country/region of travel, age, sex, and month of travel, but no data on any illness. Data from the TDB are often given as regions rather than countries, to account for low numbers of respondents outside the most popular travel destinations.
Statistical methods
The age, sex and geographical distribution of the interviewees in the TDB, were standardised against the total population of Sweden to give an extrapolation of the actual number of travellers to each country during the seven years. We then estimated risks per 100,000 travellers (divided on the exposures sex, age and region of travel) using notifications on Campylobacter infection (cases) as numerator and extrapolated total numbers of travellers from the TDB as denominator. The actual numbers of interviewed persons (controls) were used for the calculations of 95% confidence intervals (95% CI) for the estimates, using the formula:
eIn risk ± 1.96*√ (1/cases+1/controls)
To adjust for possible confounding and test for interaction, we also calculated odds ratios (OR) with corresponding 95% CI for the same exposures with a logistic regression model.
In an initial crude analysis, odds ratios (ORs) for all exposures (age, sex, and travel destination) on the outcome campylobacteriosis were analysed, with the lowest incidence in each category used as reference. Confounding was then assessed using Mantel-Haenszel stratification. ORs for exposures with significant association with the outcome were included in a logistic regression analysis if they were shown to contribute significantly to the model in a Wald test. The presence of significant interaction was tested with tests for homogeneity. For each region we analysed seasonality separately (OR for disease per month, adjusted for age, sex and number of cases/travellers). All analyses were done using the Stata 6.0 software (Stata Corporation, College Station, Tx, USA).
Ethical considerations
Notification data is regulated by the Swedish Communicable disease act, and contain full personal identification. The TDB contains aggregated data only. The Medical Ethics Committee of the Karolinska Institute, Stockholm, Sweden, approved the study.
Results
Of 53,223 persons notified with campylobacteriosis in the period 1997 to 2003, 28,704 (54%) were travel-associated, corresponding to 42.3 cases per 100,000 travellers (Table 1). The total number of infections from single countries both reflected the risk of disease in the various countries, but to a large extent also the travel pattern of Swedes. The five most commonly stated countries of infection were Thailand (n = 6,129), Spain (n = 5,646), Turkey (n = 1,812), Morocco (n = 1,501), and India (n = 1,086).
The 16,255 respondents with overnight travel outside Sweden in 1997–2003 from the TDB database corresponded to almost 68 million travel episodes; 78% leisure trips and 22% business trips (Table 2). Travel to several countries within one region was quite common, but overnight stay in more than one region was rare (less than 0.1% of travellers).
Comparing the number of cases with the projected number of travellers, we estimated the risk for Campylobacter infection in each of the 19 regions under study. The highest unadjusted risks were seen in the Indian Subcontinent (1,253 per 100 000 travellers; 95 % CI 878–1,787), East Africa (502 per 100 000; 95 % CI 314–804), East Asia (386 per 100 000; CI 353–422), North Africa (362 per 100 000; 95 % CI 313–418) and Arab countries/Iran (197 per 100 000; 95 % CI 144–268). Adjusting for age, sex, and month in the logistic regression model did not change the rank between the regions (Tables 1 and 2, Figure 1).
In the crude risk estimate women were at significant higher risk for campylobacteriosis than men; 44.0 cases per 100,000 (95 % CI 42.8–45.2) versus 40.8 cases per 100,000 (95 % CI 39.7–41.9). After adjusting for destination, age, and month in the multivariate logistic regression model, the risks were reversed with a significantly higher OR in males (1.17; 95 % CI 1.11–1.23). However, travel destination was an effect modifier on the association between sex and campylobacteriosis, and this higher risk for males were only significant for travellers returning from a European country (OR 1.21; 95% CI 1.15–1.27) (Table 3).
The highest adjusted age risks were seen in young/middle-aged adults 19–45 years old (OR 2.52; 95 % CI 2.27–2.80) and in small children 0–6 years old (OR 2.34; 95 % CI 1.99–2.76). Also the association between age and campylobacteriosis was modified by travel destination, and in travellers from tropical destinations, especially from Africa and Asia/Oceania the highest risk was seen in the youngest children (Table 3).
There was a marked seasonality in the temperate regions with peak risks mainly in the summer; Nordic countries (peak in June and nadir in March, OR 11.8; 95% CI 5.9–23.4), Western Europe (peak in June and nadir in December, OR 2.4; 95% CI 1.8–3.3), Eastern Europe (peak in June and nadir in November, OR 2.4; 95% CI 1.8–3.3), North America (peak in June and nadir in March, OR 5.8; 95% CI 1.5–23.4), Southern Europe (peak in September and nadir in January, OR 3.9; 95% CI 3.0–5.0), Northern Africa (peak in September and nadir in May, OR 4.3; 95% CI 1.7–11.0), Arab countries and Iran (peak in April and nadir in August, OR 10.1; 95% CI 1.7–26.4), and Australia/New Zealand (peak in November and Nadir in July, OR 33.1; 95% CI 2.8–394). In the Eastern Mediterranean the peak risk was seen in the spring (peak in March and nadir in January, OR 5.1; 95% CI 2.4–10.8), and in Russia and former USSR in late fall (peak in November and nadir in May, OR 6.7; 95% CI 1.3–59.2). In the tropical regions the seasonality was considerably less distinct. In East Asia the risk peak was in December with nadir in May (OR 4.5; 95% CI 2.8–7.2) and in the Caribbean in February with nadir in September (OR 7.8; 95% CI 2.2–27.7). In the Indian Subcontinent, Sub-Saharan Africa, and Central/South America no distinct, significant seasonal peaks could be identified.
Discussion
Methodological issues
In this study we report the risks for travel-associated campylobacteriosis and seasonality of the risks in various parts of the world, based on more than 28,000 notified cases. The large number of cases gives more precise risk estimates for this disease than in previous studies, although the estimates are given for quite large regions in parts of the world with few Swedish travellers. The denominator data from the TDB has previously been used in studies on dengue fever [16] and rickettsiosis [17]. We have also tested the reliability of the TDB by comparing the TDB data with in-flight passenger data obtained from some countries with such requirements. For destinations with many travellers, the two sources of information were highly compatible, e.g. less than 5 % difference for travel to Thailand.
Notification data only reflects a small (but unknown) proportion of all travel-related Campylobacter infections. One should therefore be cautious in drawing conclusions from the magnitude of the figures, and rather focus on the relative risks between the various regions, as estimated by the odds ratios. Since the data are all from the same source, the risk figures from the various regions are directly comparable. However, there may be a tendency of investigating travellers from the tropics more vigorous than travellers from e.g. the other Nordic countries, thus underestimating the risks in nearby countries. However, such selection bias could likely not explain the huge differences between say West Africa and East Africa.
Since we have no comparable data on the length of stay among the cases in our study, we were not able to include length of stay in our logistic regression model. However, the TDB clearly shows a longer median stay among travellers in far-away destinations. For instance the median stay in Spain was 6 nights, while in Thailand it was 14 nights. On the other hand, only cases detected after the return to Sweden are included in the analysis. The disease data therefore mainly reflect infections contracted during the last week of stay at the travel destination. Differences in length of travel are therefore to some extent evened out.
It has previously been suggested that the risk of travellers' diarrhoea is higher during the first two weeks in highly endemic areas [10,18]. The calculated risks in this study may therefore be underestimated in travel destinations with more prolonged stay. However, persons staying long periods abroad are also less likely to be telephone interviewed in Sweden, balancing the missed cases.
Regional risks
The differences in risk between various regions were considerable, not only between industrialized and developing countries, but also between different developing countries. The Indian Subcontinent, East Africa, East Asia, and North Africa stood out as special high-risk areas. In a recent Finnish study, the risk of travel associated Campylobacter jejuni infection was 10 per 100,000 travellers returning from Spain and Portugal, and 50, 60, and 80 per 100,000 returning travellers from China, Thailand and India, respectively [19]. The lower risks, and lesser differences between the countries may be explained by a much smaller number of cases (n = 205) to base the risk estimates on. East Africa and India have also previously been identified as high-risk areas for travel-associated diarrhoea of various aetiology [20,21], but the very large differences in the risk of campylobacteriosis between East Africa, and West, Central and Southern Africa have to our knowledge not previously been described.
Age, sex and season (month of travel/infection) were identified as possible confounders for the association between travel destination and risk for campylobacteriosis and were thus included in the logistic regression model. All three variables contributed significantly to the model, but the overall effect of these confounders did not alter the rank order between the regions in the logistic regression model compared to the crude analysis.
Age and gender
The highest risks were seen in young adults and small children, and especially in the tropics the highest risks were seen in the youngest. This is consistent with previous findings that in developed countries the disease most of hits children below the age of 5 years and young adults, while in developing countries it is most often seen in children below the age of 2 years, with an annual incidence of 40–60 % [1]. The data are also consistent with the results from other studies on traveller's diarrhoeas [11]. De Las Casas has suggested that the high risk in young adults is due to a more adventurous lifestyle when it comes to eating habits, and the elevated risk in the youngest is due to increased faecal/oral contamination and decreased immunity [22], explanations that seem plausible. An alternative explanation put forward is that young people with a greater appetite ingest more bacteria, and thereby increasing their risk of infection.
In travellers returning from Europe, male gender was an independent risk factor for Campylobacter infection. This pattern is also seen in domestically acquired Swedish campylobacteriosis cases, where 56% of the notified cases in 2003 were males, and in the US where campylobacteriosis is more common in males of all age groups [8,9]. The higher risk in males in the US has been attributed to sex-specific differences in food-handling practices and consumption practices as well as a higher susceptibility to gastro-intestinal infections in males [9]. For travel destinations to tropical or subtropical destinations, the risk was not influenced by gender, consistent with other studies on travel-associated diarrhoea [23].
Seasonality
In each of the 19 regions in the study, we looked closely at the seasonality of the disease. As has previously been shown [3,6-9], there was a striking seasonal pattern in all temperate regions, with distinct peaks in the summer. Previously these summer peaks have been partly attributed to returning travellers [4], but obviously this could not explain the same peaks in our study. The magnitude of the summer peaks was also in the same order in domestic Swedish cases, as in the returning travellers from other temperate countries.
In the tropical regions, seasonal peaks of campylobacteriosis have not previously been recognized [8]. Also in this study the seasonal pattern was much less distinct in tropical than in temperate regions, and only in East Asia (peak incidence in December) and in the Caribbean (peak incidence in February) could a seasonal pattern be discerned. In a study on US medical students in Mexico, the peak incidence of Campylobacter infection was seen between November and April [18]. With only 15 cases and 8 TDB respondents, our study did not have the power to detect any seasonality in Central America.
Conclusions
Data on infections in returning travellers together with good denominator data could provide comparable data on travel risks in various regions of the world. This study has revealed large and unexplained regional incidence differences, e.g. between East and Central Africa. The very distinct seasonal pattern seen in all temperate regions could not be discerned in the tropics.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KE raised the original idea of the study, did the statistical analyses, and prepared the first draft of the manuscript. YA contributed with in depth knowledge of campylobacteriosis and revised the draft manuscript. Both authors have read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Map showing Campylobacter risk per 100 000 returning travellers to Sweden from different regions of the world. In regions with a distinct seasonality, the month with the highest risk (OR) is given.
Table 1 Notified cases of Campylobacter infection in Sweden 1997–2002, per category of infection.
Disease and category of infection 1997 1998 1999 2000 2001 2002 2003 1997–2003
Sweden 1 856 2 616 2 208 2 462 2 839 2 479 2 688 17 148
Travel associated 3 013 3 769 4 564 4 932 4 730 3 914 3 782 28 704
Immigrants/refugees 309 166 167 150 134 111 133 1 170
Unknown 1 703 846 730 870 875 633 544 6 201
Total 6 881 7 397 7 669 8 414 8 578 7 137 7 147 53 223
Table 2 Estimated number of Swedish travellers, respondents in the Travel and Tourist database – TDB (controls) and notified Swedish cases with travel associated campylobacteriosis 1997–2003, with an unadjusted risk estimate (per 100,000) and multivariate odds ratios from a logistic regression model adjusted for the risk factors age, sex month of travel and travel destination.
Age/Sex/Regiona Estimated no, of travellers Controls (TDB) Notified, cases Risk per 100,000 95% CI Multivariable odds ratio 95% CI
Total 67,870,000 16,255 28,704 42.3 41.5–43.1 - -
0 to 6 years 3,300,000 524 1,234 37.4 33.8–41.4 2.34 1.99–2.76
7 to 18 years 8,150,000 1,599 1,957 24.0 22.5–25.7 1.33 1.17–1.51
19 to 45 years 30,520,000 6,708 16,207 53.1 51.6–54.6 2.52 2.27–2.80
46 to 65 years 21,850,000 5,990 8,200 37.5 36.3–38.8 1.50 1.35–1.66
65+ years 4,050,000 1,434 1,106 27.3 25.2–29.5 Reference
Men 36,020,000 8,145 14,694 40.8 39.7–41.9 1.17 1.11–1.23
Women 31,850,000 8,110 14,007 44.0 42.8–45.2 Reference
Nordic countries 22,730,000 5,350 606 2.67 2.45–2.90 Reference
Western Europe 14,800,000 3,584 2,238 15.1 14.3–15.9 5.58 5.05–6.17
Southern Europe 12,070,000 2,931 6,730 55.8 53.4–58.2 22.2 20.1–24.4
Eastern Europe (incl. Baltic Republics) 3,320,000 818 1,414 42.6 39.1–46.4 14.9 13.2–16.8
Eastern Mediterranean 7,740,000 1,817 3,260 42.1 39.8–44.6 14.5 13.0–16.0
Russia and former USSR 260,000 59 96 36.9 26.7–51.1 14.3 10.2–20.1
Arab countries and Iran 220,000 44 433 197 144–268 92.0 66–128
Indian Subcontinent 120,000 31 1,503 1,253 878–1787 532 369–769
East Asia 2,050,000 517 7,910 386 353–422 173 152–197
Australia, New Zealand and the Pacific 450,000 116 75 16.7 12.5–22.3 6.64 48.9–9.02
North Africa 770,000 196 2,788 362 313–418 164 138–195
West Africa 80,000 22 61 76.3 47–124 28.8 17.4–47.6
East Africa 90,000 18 452 502 314–804 243 151–394
Central Africa 30,000 8 15 50.0 21–118 17.4 7.2–41.8
Southern Africa 170,000 42 150 88.2 63–124 37.2 26.0–53.2
North America 2,170,000 503 77 3.5 2.8–4.5 1.39 1.08–1.80
Central America 170,000 43 148 87.1 62–122 35.0 24.6–49.9
Caribbean 380,000 95 299 78.7 62.5–99.1 33.0 25.7–42.4
South America 250,000 61 449 180 138–235 76.0 57–101
aNordic countries = Denmark, Finland, Iceland, Norway; Western Europe = Austria, Belgium, France, Germany, Ireland, Luxembourg, The Netherlands, Switzerland, United Kingdom; Southern Europe = Italy, Malta, Monaco, Portugal, Spain; Eastern Europe = Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia; Eastern Mediterranean = Albania, Cyprus, Former Yugoslavia, Greece, Israel, Turkey; Russia and former USSR = Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Russia, Tajikistan, Turkmenistan, Ukraine, Uzbekistan; Arab countries and Iran = Bahrain, Iraq, Iran, Jordan, Kuwait, Lebanon, Oman, Qatar, Saudi Arabia, Syria, United Arab Emirates, Yemen; Indian Subcontinent = Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, Sri Lanka; East Asia = Brunei, Burma, Cambodia, China, Hong Kong, Indonesia, Japan, Laos, Malaysia, Mongolia, North Korea, Philippines, South Korea, Singapore, Taiwan, Thailand, Tibet, Viet Nam; Australia, New Zeeland, and the Pacific = American Samoa, Australia, Cook Islands, Fiji, French Polynesia, Guam, Kiribati, Marshall Islands, Micronesia, Nauru, New Caledonia, New Zealand, Niue, Palau, Papua New Guinea, Samoa, Tokelau, Tonga, Tuvalu, Vanuatu, Wallis and Futuna; North Africa = Algeria, Egypt, Libya, Morocco, Tunisia; West Africa = Benin, Burkina Faso, Cape Verde, Ghana, Guinea, Guinea-Bissau, Ivory Coast, Liberia, Mali, Mauritania, Senegal, Sierra Leone, The Gambia, Togo; East Africa = Burundi, Djibouti, Eritrea, Ethiopia, Kenya, Rwanda, Seychelles, Somalia, Sudan, Tanzania, Uganda; Central Africa = Cameron, Central African Republic, Chad, Congo Brazzaville, Equatorial Guinea, Gabon, Niger, Nigeria, Republic of Congo, São Tomé et Principe; Southern Africa = Angola; Botswana, Lesotho, Madagascar, Malawi, Mauritius, Mozambique, Namibia, South Africa, Zambia, Zimbabwe; North America = Canada, USA; Central America = Belize, Costa Rica, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama; Caribbean = Antigua and Barbuda, Bahamas, Barbados, Bermuda, Cayman Islands, Cuba, Dominica, Dominican Republic, Grenada, Guadeloupe, Jamaica, Haiti, Martinique, Netherlands Antilles, Puerto Rico, S:t Christopher and Nevis, S:t Lucia / S:t Vincent, Saint Kitts-Nevis, The Grenadines, Trinidad and Tobago, Virgin islands; South America = Bolivia, Brazil, Colombia, Ecuador, French Guiana, Guyana, Honduras, Paraguay, Peru, Suriname, Uruguay, Venezuela.
Table 3 Multivariable odds ratios (per continent) for the risk of being notified with travel-associated campylobacteriosis from a logistic regression model adjusted for the risk factors age, sex month of travel and travel destination. For North America there were too few cases for any meaningful results.
All regions Europe Asia + Oceania Africa Latin America + Carribean
Age/sex OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Men 1.17 1.11–1.23 1.21 1.15–1.28 1.10 0.93–1.29 1.06 0.83–1.37 0.85 0.62–1.17
Women Reference Reference Reference Reference Reference
0 to 6 years 2.34 1.99–2.76 2.00 1.67–2.39 9.39 4.66–18.90 7.67 2.76–21.29 2.89 0.89–9.41
7 to 18 years 1.33 1.17–1.51 1.28 1.11–1.48 1.82 1.16–2.85 1.64 0.91–2.98 1.25 0.47–3.31
19 to 45 years 2.52 2.27–2.80 2.41 2.27–2.80 4.04 2.81–5.84 3.86 2.44–6.10 2.15 1.01–4.56
46 to 65 years 1.50 1.35–1.66 1.45 1.29–1.62 1.99 1.38–2.89 1.74 1.11–2.74 1.41 0.65–3.04
65+ years Reference Reference Reference Reference Reference
==== Refs
Coker AO Isokpehi RD Thomas BN Amisu KO Obi CL Human campylobacteriosis in developing countries Emerg Infect Dis 2002 8 237 44 11927019
Heymann D Campylobacter enteritis Control of communicable diseases manual 2004 18 Washington DC: American Public Health Association
Blaser MJ Mandell GL, Bennett JE, Dolin Campylobacter and related species Principles and practice of infectious diseases 1995 4 New York: Churchill Livingstone
Nylen G Dunstan F Palmer SR Andersson Y Bager F Cowden J Feierl G Galloway Y Kapperud G Megraud F Molbak K Petersen LR Ruutu P The seasonal distribution of campylobacter infection in nine European countries and New Zealand Epidemiol Infect 2002 128 383 90 12113481 10.1017/S0950268802006830
Kapperud G Skjerve E Bean NH Ostroff SM Lassen J Risk factors for sporadic Campylobacter infections: resultsof a case-control study in southeastern Norway J Clin Microbiol 1992 30 3117 21 1452694
Schonberg-Norio D Takkinen J Hanninen ML Katila ML Kaukoranta SS Mattila L Rautelin H Swimming and Campylobacter infection Emerg Infect Dis 2004 10 1474 7 15496253
Studahl A Andersson Y Risk factors for indigenous Campylobacter infection: a Swedish case-control study Epidemiol Infect 2000 125 269 75 11117949 10.1017/S0950268899004562
Allos BM Campylobacter jejuni infections: update on emerging issues and trends Clin Infect Dis 2001 32 1201 6 11283810 10.1086/319760
Samuel MC Vugia DJ Shallow S Marcus R Segler S McGivern T Kassenborg H Reilly K Kennedy M Angulo F Tauxe RV Emerging Infections Program FoodNet Working Group Epidemiology of sporadic Campylobacter infection in the United States and declining trend in incidence, FoodNet 1996–1999 Clin Infect Dis 2004 38 S165 S174 15095186 10.1086/381583
Steffen R de Bernardis C Banos A Travel epidemiology-a global perspective Int J Antimicrob Agents 2003 21 89 95 12615369 10.1016/S0924-8579(02)00293-5
Peltola H Gorbach SL DuPont HL, Steffen R Travelers' diarrhea epidemiology and clinical aspects Textbook of Travel Medicine and Health 1997 Hamilton, Ontario: BC Decker Inc
Ericsson CD Travellers' diarrhoea Int J Antimicrob Agents 2003 21 116 24 12615374 10.1016/S0924-8579(02)00282-0
Okhuysen PC Ericsson CD Travelers' diarrhea. Prevention and treatment Med Clin North Am 1992 76 1357 73 1405823
Swedish Institute for Infectious Disease Control (SMI): Updated statistics covering the notifiable diseases in Sweden
Swedish Travel and Tourist Data Base, TDB: Resurs AB, Sweden
Lindback H Lindback J Tegnell A Janzon R Vene S Ekdahl K Dengue fever in travelers to the tropics, 1998 and 1999 Emerg Infect Dis 2003 9 438 42 12702223
Rahman A Tegnell A Vene S Giesecke J Rickettsioses in Swedish travellers, 1997–2001 Scand J Infect Dis 2003 35 247 50 12839153 10.1080/00365540310005422
Ericsson C DuPont HL Travelers' diarrhea: approaches to prevention and treatment Clin Infect Dis 1993 16 616 24 8507751
Hakanen A Jousimies-Somer H Siitonen A Huovinen P Kotilainen P Fluoroquinolone resistance in Campylobacter jejuni isolates in travelers returning to Finland: association of ciprofloxacin resistance to travel destination Emerg Infect Dis 2003 9 267 70 12604004
von Sonnenburg F Tornieporth N Waiyaki P Lowe B Peruski LF JrDuPont HL Mathewson JJ Steffen R Risk and aetiology of diarrhoea at various tourist destinations Lancet 2000 356 133 4 10963251 10.1016/S0140-6736(00)02451-X
Jiang ZD Lowe B Verenkar MP Ashley D Steffen R Tornieporth N von Sonnenburg F Waiyaki P DuPont HL Prevalence of enteric pathogens among international travelers with diarrhea acquired in Kenya (Mombasa), India (Goa), or Jamaica (Montego Bay) J Infect Dis 2002 185 497 502 11865402 10.1086/338834
De Las Casas C Adachi J Dupont H Review article: travellers' diarrhoea Aliment Pharmacol Ther 1999 13 1373 8 10571591 10.1046/j.1365-2036.1999.00638.x
Kollaritsch H Travelers' diarrhea among Austrian tourists in warm climates countries. I Epidemiology Eur J Epidemiol 1989 5 74 81 2785058 10.1007/BF00145049
| 15569393 | PMC539239 | CC BY | 2021-01-04 16:03:30 | no | BMC Infect Dis. 2004 Nov 29; 4:54 | utf-8 | BMC Infect Dis | 2,004 | 10.1186/1471-2334-4-54 | oa_comm |
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-2-431558506610.1186/1477-7819-2-43Case ReportPancreatic metastasis from gastric carcinoma: a case report Wente Moritz N [email protected] Frank [email protected]öhlich Boris E [email protected] Peter [email protected]üchler Markus W [email protected] Helmut [email protected] Department of General Surgery, University of Heidelberg, Heidelberg, Germany2 Institute of Pathology, University of Heidelberg, Heidelberg, Germany2004 7 12 2004 2 43 43 18 9 2004 7 12 2004 Copyright © 2004 Wente et al; licensee BioMed Central Ltd.2004Wente 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 pancreas is a rare but occasionally favored target for metastasis. Metastatic lesions in the pancreas have been described for various primary cancers, such as carcinomas of the lung, the breast, renal cell carcinoma and sarcomas.
Case presentation
We report the case of a 60-year old female with a mass in the pancreatic head four years after partial gastrectomy for gastric adenocarcinoma. The patient underwent a pancreatoduodenectomy. Pathological examination revealed metastases of the primary gastric carcinoma within the pancreatic head and in regional lymph nodes.
Conclusions
Pancreatic tumors in patients with a history of non-pancreatic malignancy should always be considered to be a putative metastatic lesion at an unusual site. If the pancreas can be identified as the only site of spread, radical resection may prolong survival.
==== Body
Background
The pancreas is an uncommon location for solitary metastasis from other primary cancers [1]. Despite this, in large autopsy series the prevalence of pancreatic metastasis has been described to be as high as 6% to 11% [2]. Whereas renal cell carcinoma appears to be the most common primary tumor to cause secondary pancreatic tumors, a variety of other cancers may spread to the pancreas, such as colon cancer, non-small cell lung cancer, and sarcomas [3]. This article presents the case of a pancreatic metastasis presenting as first site of gastric cancer recurrence four years after primary diagnosis.
Case presentation
A 60-year-old woman presented with elevated blood levels of the tumor markers CEA (17.3 μg/L, normal <2.5 μg/L) and CA 19-9 (121 U/ml, normal <37 U/ml). Four years before being referred to our institution, the patient had undergone gastric resection (Billroth II gastrectomy) for an adenocarcinoma of the stomach. The tumor was located at the lesser curvature of the gastric antrum, measuring 3 cm in the largest diameter. Pathologic examination revealed a gastric carcinoma of low differentiation, which infiltrated the gastric wall into the subserosal layer without penetrating the serosa. Microscopically, the carcinoma was mainly composed of tubular formations of mitotically active, atypical epithelial cells (Figure 1A). The tumor also displayed areas of marked desmoplastic stromal reaction, as well as areas of rather glandular differentiation. The latter two were mainly observed in paragastric lymph node metastases (Figure 1B). The carcinoma at stage pT2 pN1 (6/15) M0 G2 was completely resected (R0 resection). No recurrence was detected during the regular follow-ups.
Figure 1 Histomorphologic appearance of the primary gastric carcinoma (A) and a paragastric lymph node metastasis (B). Photomicrograph shows that the primary tumor is mainly composed of solid and tubular formations, whereas, a marked desmoplastic stromal reaction is seen in the lymph node metastasis. (hematoxylin and eosin × 40).
Four years after gastric resection, however, ultrasound examination, computed tomography and magnetic resonance imaging revealed an inhomogeneous mass of the pancreatic head, measuring 4 cm in largest diameter (Figure 2). Radiographically, no other masses were detected. For differential diagnosis, a primary carcinoma of the pancreas and a metastasis of the gastric carcinoma were considered. Following explorative laparotomy, the pancreatic mass was resected performing a partial pancreatoduodenectomy with resection of the distal bile duct (Whipple's procedure). Furthermore, the former gastroenterostomy was resected and revised. On pathologic examination, the tumor of the pancreatic head grossly presented as white to yellowish, firm mass. Microscopically, the tumor consisted of solid and glandular formations of atypical epithelial cells with distinct nuclear pleomorphism and presented marked desmoplastic stromal reaction, as well as areas of necrosis (Figure 3). The duodenal wall and the peripancreatic tissue were infiltrated by the tumor. Lymph node metastases were detected in two peripancreatic lymph nodes. The histomorphological appearance of the pancreatic tumor was in good accordance to some areas of the primary gastric tumor, and especially the growth pattern found in the paragastric lymph node metastases coped well with the microscopic picture found in the pancreatic tumor. Immunohistochemical analyses revealed identical expression patterns in the gastric carcinoma and the pancreatic mass, both displaying positive reactions with antibodies towards cytokeratins 8, 18 and 19, as well as carcinoembryonic antigen (CEA), whereas no reactions were seen with antibodies towards cytokeratins 7 and 20. Because of these findings and due to the lack of pancreatic cancer progenitor lesions, pancreatic intraepithelial neoplasias (PanINs), within the non-neoplastic pancreatic tissue of the Whipple's resection specimen, the pancreatic tumor and the two regional lymph node metastases were considered to be metastases of the primary gastric carcinoma.
Figure 2 Computed tomography of the abdomen four years after Billroth II resection for gastric cancer, revealing an inhomogenous mass in the pancreatic head, 4 cm in diameter. (Picture courtesy the Division of Radiology, German Cancer Research Center, provided by PD Dr. med. S. Delorme).
Figure 3 Histomorphologic appearance of the delayed pancreatic metastasis (hematoxylin and eosin × 100). As in the lymph node metastasis, a marked desmoplastic stromal reaction is seen in the pancreatic metastasis.
The patient was discharged from the hospital without any perioperative morbidity on the ninth postoperative day. The postoperative blood levels of the tumor markers declined to normal values (CEA 2.6 μg/l, CA 19-9 24 U/ml). Due to the complete surgical resection and the lack of risk factors for recurrence, the patient received no further adjuvant therapy. Under regular follow-up for one year with determination of the tumor markers and computed tomography, the patient revealed no signs or symptoms of local or systemic recurrence.
Discussion
Death from recurrence of gastric adenocarcinoma occurs in 70–75% of patients during the first two years after surgical intervention, however, reports of recurrences more than 10 years after primary diagnosis have been reported as well [4]. The most frequent sites of tumor recurrences include local, regional and peripheral lymph nodes, as well as the liver, the lungs, and the peritoneum [5]. Furthermore, solitary metastasis in other organs, such as the thyroid gland or the spleen have been described [6,7]. In contrast to direct infiltration into the pancreas, metastases of gastric cancer into the pancreas are considered to be extremely rare and to our knowledge only four cases have been reported in the English literature [8-10].
Adenocarcinomas of the pancreas and of other primary sites frequently display a large histomorphological and immunohistochemical overlap. Thus the differential diagnosis of primary pancreatic cancer versus solitary metastases of other adenocarcinomas may be very difficult – if not impossible – using common pathological and immunohistochemical techniques. According to Robbins et al [3], solitary pancreatic masses can be classified as secondary tumors to the pancreas in only 2% of the cases, and they are frequently misdiagnosed as primary pancreatic cancers. As a consequence from this, the subtle diagnostic work-up for isolated masses in the pancreas needs to inherit a meticulous elaboration of the medical history of the patients, in particular focused on previous non-pancreatic malignancy.
Pancreatic resections can nowadays be performed with low morbidity and mortality rates, in particular in high-volume centers [11,12]. Results of surgical extirpation of isolated metastases to the pancreas from various primary tumors provide improvement with regard to long-term survival [1,2]. Therefore, a resection of isolated metastases in the pancreas should be considered as a treatment option in patients with the history of non-pancreatic malignancy [13].
Competing Interests
The authors declare that they have no competing interests.
Authors' Contributions
MNW collated the information, searched the literature and wrote the manuscript.
FB and PS contributed to the pathological aspects of the manuscript and helped in preparing the manuscript.
BEF assisted in literature search and writing of the manuscript.
MWB and HF managed the patient and helped in preparing the manuscript and edited the final version with PS.
All authors read and approved the final version of the manuscript.
Acknowledgements
Written consent was obtained from the patient for publication of this case.
==== Refs
Hiotis SP Klimstra DS Conlon KC Brennan MF Results after pancreatic resection for metastatic lesions Ann Surg Oncol 2002 9 675 679 12167582 10.1245/aso.2002.9.7.675
Z'graggen K Fernandez-del Castillo C Rattner DW Sigala H Warshaw AL Metastases to the pancreas and their surgical extirpation Arch Surg 1998 133 413 417 9565122 10.1001/archsurg.133.4.413
Robbins EG Franceschi D Barkin JS Solitary metastatic tumors to the pancreas: a case report and review of the literature Am J Gastroenterol 1996 91 2414 2417 8931428
Shiraishi N Inomata M Osawa N Yasuda K Adachi Y Kitano S Early and late recurrence after gastrectomy for gastric carcinoma. Univariate and multivariate analyses Cancer 2000 89 255 261 10918153 10.1002/1097-0142(20000715)89:2<255::AID-CNCR8>3.0.CO;2-N
Ushijima T Sasako M Focus on gastric cancer Cancer Cell 2004 5 121 125 14998488 10.1016/S1535-6108(04)00033-9
Ok E Sozuer E Thyroid metastasis from gastric carcinoma: report of a case Surg Today 2000 30 1005 1007 11110395 10.1007/s005950070021
Yamanouchi K Ikematsu Y Waki S Kida H Nishiwaki Y Gotoh K Ozawa T Uchimura M Solitary splenic metastasis from gastric cancer: report of a case Surg Today 2002 32 1081 1084 12541027 10.1007/s005950200218
Brannigan AE Kerin MJ O'Keane JC McEntee GP Isolated resectable pancreatic metastasis 10 years post gastrectomy Ir J Med Sci 2000 169 227 11272885
Nakai T Shimomura T Nakai H A case of isolated pancreatic metastasis of gastric cancer presenting problematic discrimination from gastropancreatic double cancer Hepatogastroenterology 2004 51 1571 1574 15362804
Roland CF van Heerden JA Nonpancreatic primary tumors with metastasis to the pancreas Surg Gynecol Obstet 1989 168 345 347 2928909
Birkmeyer JD Siewers AE Finlayson EV Stukel TA Lucas FL Batista I Welch HG Wennberg DE Hospital volume and surgical mortality in the United States N Engl J Med 2002 346 1128 1137 11948273 10.1056/NEJMsa012337
Büchler MW Wagner M Schmied BM Uhl W Friess H Z'graggen K Changes in morbidity after pancreatic resection: toward the end of completion pancreatectomy Arch Surg 2003 138 1310 1314 14662530 10.1001/archsurg.138.12.1310
Sperti C Pasquali C Liessi G Pinciroli L Decet G Pedrazzoli S Pancreatic resection for metastatic tumors to the pancreas J Surg Oncol 2003 83 161 166 12827684 10.1002/jso.10262
| 15585066 | PMC539240 | CC BY | 2021-01-04 16:38:57 | no | World J Surg Oncol. 2004 Dec 7; 2:43 | utf-8 | World J Surg Oncol | 2,004 | 10.1186/1477-7819-2-43 | oa_comm |
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-3-451557596310.1186/1475-925X-3-45Book ReviewReview of "System approach to engineering design" by P.H. Sydenham Mondry Adrian [email protected] Medical and Clinical Informatics Group, Bioinformatics Institute, Singapore2004 3 12 2004 3 45 45 Sydenham PH: Systems approach to engineering design. Boston, USA: Artech House Publishers; 2004 349 pages, ISBN 1580534791, Hard cover29 11 2004 3 12 2004 Copyright © 2004 Mondry; licensee BioMed Central Ltd.2004Mondry; 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.
==== Body
Based on a semester-long seminar on the topic, this book aims to fill a gap in the current engineering curricula by taking a wide-angle view at the process of engineering design rather than focusing on a more narrow and in-depth approach. As part of Artech House Publishers' technology management and professional development library, this is an excellent introduction to the topic and may serve for later reference.
The author, Peter. H. Sydenham is the inaugural professor and head of the University of South Australia's School of Electronic Engineering, co-founder of the Australian Centre for Test and Evaluation, and a director of Global Systems Engineering Consulting Pty Ltd. This brief biographical sketch explains his qualification to write a book of such practical value and dimension, and why the book is pedagogically sound.
The book, according to the preface, aims at those readers who are engineers "who have, or aspire to team leadership or want to take on increased team interfacing responsibilities". As such it builds on and expands what the author perceives to be too little taught in the regular university courses, and covers what is lacking: "breadth of the knowledge now needed to be an effective engineering designer."
Twelve chapters offer a step-by-step introduction to the topic. The first chapter, "Systems Thinking and Systems Engineering", gives an overview and the "philosophical" background, while the last looks at "Change and Future Trends". The remaining ten chapters cover the practical aspects of the whole process of systems engineering design with a more hands-on approach, including basics of supporting knowledge such as staff management from planning over recruitment through training to promotion or termination of contract (chapter 3) or information technology support. As such, chapters 2–4 cover the more general aspects, while chapters 5–9 cover the design process from idea to evaluation. Chapter 10 intersperses legal issues, while chapter 11 covers the prototype build, i.e. the step from abstract design to material product.
This is an introductory textbook, and as such, it does a very good job of the task at hand. The language is clear, the case examples are well chosen, and the message is conveyed without fail. Some issues are discussed somewhat at length; for example, the basic IT issues are explained in a very detailed way, but one might assume that the basics would be self-evident for today's user: the target audience will have grown up in the age of the personal computer and the internet and as such, will neither question the computer's usefulness nor be overly naïve with regards to the many IT bugs one has to deal with.
The illustrations are about as clear as can be-some, in fact, are not clear at all but as they are meant to demonstrate the complexity of the issue at hand, the message gets conveyed the way it should be.
How does this general book relate to the specialist field of biomedical engineering? To quote from p. 12: "Researchers in the life sciences were driven by a need to better understand how nature works and controls itself. Out of this pioneering work emerged general systems theory cybernetics, self-organizing systems, automation, automaton systems, organizational science, operations research, systems science, and more-topics with which engineers are not usually that familiar". This lack of familiarity becomes very evident when young engineers join research teams with a focus on applications for the life sciences, such as biomechanical engineering, but also within the broad field of bioinformatics, and they usually take more time to acquire and apply that holistic view of their work than life scientists need to acquire and implement detailed and circumscript engineering knowledge to accomplish their tasks. As such, this book can be recommended to engineers working in biomedicine even at the outset of their careers as it may draw their attention to the importance of this view of things in the new field they enter. Competing recent titles include works by Blanchard [1], Hitchins [2] and Sheridan [3]. A more general outlook on the importance of systems thinking in many areas of life, and far beyond engineering design, is Gharajedaghi's [4] modern classic.
One point that a next edition seriously needs to address is the text editing. Using a word processor does not guarantee a perfect text: this one has a lot of extra words, and as many words missing. As faulty as the text editing is the punctuation check. These errors occur as often as once per page, and force the reader to repeat the study of entire paragraphs several times because, more often than not, it is the missing preposition or punctuation that poses a serious threat to understanding. The publishing house would do well in employing an old-fashioned human text editor to spare the reader such nuisance.
In summary, this is a basic textbook to project design management for aspiring team leaders, not only in engineering but for any scientific and some business projects as well. Well written, this is a recommended introduction to the matter and may serve as a reference book and reminder even to the experienced team leader.
Adrian Mondry is at Bioinformatics Institute, Singapore [email protected].
==== Refs
Blanchard BS System Engineering Management 2003 Indianapolis, Wiley Interscience
Hitchins DK Advanced Systems Thinking, Engineering, and Management 2003 Boston, Artech House Publishers
Sheridan T Humans and Automation: System Design and Research Issues 2002 Indianapolis, Wiley Interscience
Gharajedaghi J Systems Thinking: Managing Chaos and Complexity: A Platform for Designing Business 1999 Oxford, Butterworth- Heinemann
| 0 | PMC539241 | CC BY | 2021-01-04 16:37:32 | no | Biomed Eng Online. 2004 Dec 3; 3:45 | utf-8 | Biomed Eng Online | 2,004 | 10.1186/1475-925X-3-45 | oa_comm |
==== Front
Lipids Health DisLipids in Health and Disease1476-511XBioMed Central London 1476-511X-3-261557162910.1186/1476-511X-3-26ResearchThe apolipoprotein E polymorphism and the cholesterol-raising effect of coffee Strandhagen Elisabeth [email protected] Henrik [email protected] Nibia [email protected]ér Mona [email protected] Lars [email protected] Kaj [email protected] Dag S [email protected] the Cardiovascular Institute, Department of Medicine, Sahlgrenska University Hospital/Östra, Göteborg, Sweden2 Department of Clinical Chemistry and Transfusion Medicine, Sahlgrenska University Hospital/Sahlgrenska, Göteborg, Sweden3 Institute of Clinical Neuroscience, Department of Experimental Neuroscience, Sahlgrenska University Hospital/Mölndal, Mölndal, Sweden2004 30 11 2004 3 26 26 4 11 2004 30 11 2004 Copyright © 2004 Strandhagen et al; licensee BioMed Central Ltd.2004Strandhagen 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 response of serum cholesterol to diet may be affected by the apolipoprotein E (APOE) ε2/ε3/ε4 polymorphism, which also is a significant predictor of variation in the risk of coronary heart disease (CHD) and CHD death. Here, we test the hypothesis that the APOE polymorphism may modulate the cholesterol-raising effect of coffee.
Objective
We determined the effect of a coffee abstention period and a daily intake of 600 mL coffee on serum cholesterol and triglycerides with respect to the APOE polymorphism.
Design
121 healthy, non-smoking men (22%) and women (78%) aged 29–65 years, took part in a study with four intervention periods: 1 and 3) a coffee free period of three weeks, 2 and 4) 600 mL coffee/day for four weeks.
Results
APOE ε2 positive individuals had significantly lower total cholesterol concentration at baseline (4.68 mmol/L and 5.28 mmol/L, respectively, p = 0.01), but the cholesterol-raising effect of coffee was not influenced significantly by APOE allele carrier status.
Conclusions
The APOE ε 2 allele is associated with lower serum cholesterol concentration. However, the APOE polymorphism does not seem to influence the cholesterol-raising effect of coffee.
controlled studyAPOE polymorphismserum lipidsfiltered coffee
==== Body
Introduction
Apolipoprotein E (apoE) is a structural component of triglyceride-rich lipoproteins, chylomicrons, very-low-density lipoproteins (VLDL), and high-density-lipoproteins (HDL) [1]. Variation in the APOE gene sequence results in the 3 common alleles (ε2, ε3 and ε4), which can produce 6 different genotypes (ε2/ε2, ε2/ε3, ε2/ε4, ε3/ε3, ε3/ε4 and ε4/ε4). The ε2, ε3 and ε4 alleles encode three distinct forms of apoE (E2, E3 and E4) and have approximate frequencies of 8%, 77%, and 15%, respectively, in white populations [2]. ApoE3 seems to be the normal isoform in all known functions, while apoE4 and apoE2 can each be dysfunctional [3,4]. In most populations, individuals with the APOE ε2 allele display lower levels of plasma cholesterol compared with individuals carrying the APOE ε3 allele, whereas individuals with the APOE ε4 allele show higher levels of plasma cholesterol, especially LDL-cholesterol [1,2,5]. Subjects with APOE ε3/ε4 and ε4/ε4 genotypes absorb cholesterol effectively and have higher non-fasting serum triglyceride values than ε4 negative individuals [6,7]. The allelic variation in the APOE gene is shown to be a significant predictor of variation in the risk of coronary heart disease (CHD) and CHD death [2-4,8-10], but the results from an extensive prospective study showed no associations [11]. Both the MONICA Project [12] and the Scandinavian Simvastatin Survival Study [13] suggest an increased risk of CHD for individuals carrying the APOE ε4 allele. The APOE ε4 allele is also considered a strong risk factor for Alzheimer's disease [14-16].
The serum cholesterol-raising effect of coffee is due to the diterpenes kahweol and cafestol [17]. Earlier studies have shown a cholesterol-raising effect mainly of unfiltered coffee, because a major part of the diterpenes is retained by a paper filter [18-20]. A recent trial, however, indicates that filtered coffee has a more pronounced serum cholesterol-raising effect than previously anticipated [21]. This finding was further corroborated in a randomized intervention study, where we demonstrated a considerable cholesterol-raising effect of filtered coffee [22]. In the study two coffee abstention periods were associated with a significant decline in serum cholesterol of 0.22 and 0.36 mmol/L, respectively, while 600 mL filtered coffee a day during two different periods increased serum cholesterol by 0.25 and 0.15 mmol/L, respectively. Here, we test the hypothesis that these effects might be modulated by the APOE ε2/ε3/e4 polymorphism.
Subjects and methods
Trial design
The study was organised as a prospective, controlled study with four consecutive trial periods. The first and third periods were 3 weeks of total coffee abstention. The second and fourth period consisted of 4 weeks with the subjects consuming 600 mL filter brewed coffee/day.
The main outcome or effect variable was total serum cholesterol and the effect was assessed as the difference between the measurements at the beginning and the end of the coffee free periods (coffee abstention) and the difference between measurements at the beginning and at the end of the four weeks of coffee consumption (Figure 1). Trial duration of 3–4 weeks has previously been shown to be sufficient to get an effect of coffee on serum cholesterol [21,23].
Figure 1 Study design
Subjects and procedure
Participants were recruited by advertising in Gothenburg's major newspaper. Inclusion criteria were age-range 30–65 years, free from clinically recognized chronic diseases, such as cardiovascular diseases, cancer, renal disorders, liver disease and diabetes mellitus. The participants were required to be free from anti-epileptic or cholesterol lowering drugs, and had been using coffee on a regular basis for at least five years and were currently non-smokers (at least for the last six months).
During the coffee drinking periods the participants were instructed to drink about 600 mL filter brewed coffee/day (4 cups), according to standardised measures. The coffee was provided to guarantee that they were all exposed to the same brand and quality of filter brewed coffee. All participants also got the same kind of standardised coffee filter and measuring spoon.
The coffee filters used were of the brand Euro-Shopper, made by Indupa N.V., Zaventem, Belgium. Divergence from the 4 cups was reported. The participants were allowed to drink tea and other caffeine containing beverages during the coffee-free periods.
Effect variables
Non-fasting blood samples were drawn at inclusion and at three, seven, ten and fourteen weeks after start of the study. Prior to analysis, prepared serum was stored at -70°C.
The blood samples were analysed for blood lipids (total cholesterol, HDL cholesterol, triglycerides, lipoprotein(a) (Lp(a)) and urate in serum. Serum cholesterol and triglycerides were determined by an enzymatic procedure on a Hitachi 917 analyzer. HDL cholesterol was determined enzymatically after precipitation of VLDL, LDL and chylomicrones by α-cyklodextrinsulphate and dextransulphate. Determination of Lp(a) was done by the method Tint Elize Lp(a) of Biopool International. Serum urate was analysed by Hitachi 917 autoanalyser. Body Mass Index (BMI; kg/m2) was recorded once during the study. Blood pressure was recorded by manual device and EKG and heart rate were recorded at all five visits.
The dietary habits were assessed by dietary frequency questionnaires at the beginning of the study. A follow-up survey with special emphasis on changes in food habits during the four different periods was undertaken. The dietary questionnaire was based upon a Norwegian version, which has been used in a number of previous studies [24].
Genotype Analysis
APOE genotypes were determined by solid-phase minisequencing as previously described by Blennow et al [25].
Statistical methods
All analyses were performed using the SAS©software version 8. Signed rank test was used to test differences in the groups. Wilcoxon rank sum test was used to test differences at baseline and differences between the groups. The mean was used as location measure and measures of variation were described in terms of standard deviation. P-values < 0.05 were considered statistically significant.
Results
A total of 156 persons responded to the advertisement and of these 124 fulfilled the criteria and were able to take part. Three persons decided to withdraw during the study, leaving a total of 121 participants. One person was not able to take part during the first two periods and five persons were not able to take part in the last two periods, which resulted in 120 participants in the first trial period and 116 in the subsequent trial period.
Genotype frequencies
The APOE allele frequencies were 6.1% for the ε2 allele, 75.6% for the ε3 allele and 18.2% for the ε4 allele. This distribution agrees well with those reported in other populations in northern Europe [2,3]. Genotype and allele frequencies for the APOE polymorphism are given in Table 1.
Table 1 APOE genotype and allele frequency, n = 121
n %
ε2/ε2 2 1.7
ε2/ε3 9 7.4
ε2/ε4 2 1.7
ε3/ε3 69 57.0
ε3/ε4 36 29.8
ε4/ε4 3 2.5
ε2 15 6.2
ε3 183 75.6
ε4 44 18.2
Serum cholesterol concentrations according to genotype and coffee exposure
Individual APOE genotypes (six subgroups, Table 1) did not influence baseline values or coffee-induced changes in serum cholesterol, serum HDL cholesterol, serum triglycerides or serum Lp(a), possibly due to a small sample size (data not shown). ε4-positive individuals had similar serum cholesterol levels and coffee-induced changes in cholesterol concentration as ε4-negative individuals (data not shown). However, when grouping ε2-positive individuals it was revealed that these displayed significantly lower cholesterol at baseline (Table 2). There was a significant difference in cholesterol decrease between week 0 and 3 for both groups. There was no difference between the two groups regarding the cholesterol decrease in the first coffee abstention period but there was a significant difference in cholesterol decrease in the second coffee abstention period, where ε2-negative individuals displayed a larger decrease in cholesterol (Table 2).
Table 2 Serum cholesterol concentration (mmol/L) at baseline and after two 3-week periods of coffee abstention (week 0 – 3 and week 7 – 10) for APOE ε2-positive (n = 13) and APOE ε2-negative (n = 108) individuals
APOE ε2-positive APOE ε2-negative p
n = 13 n = 107/103 a
First trial period
week 0 4.68 (0.80) 5.28 (0.93) 0.01b
week 3 4.49 (0.71) 5.05 (0.90)
diff week 0–3 -0.18 (0.24) -0.23 (0.55) 0.30 c
p (diff 0–3) 0.02 d <0.0001d
Second trial period
week 7 4.52 (0.71) 5.34 (0.93)
week 10 4.34 (0.64) 4.95 (0.89)
week 7–10 -0.18 (0.41) -0.39 (0.55) 0.08 c
p (diff 7–10) 0.13 <0.0001d
a 107 participants in the first trial period and 103 participants in the second trial period
b Significant difference between APOE ε2-positive and APOE ε2-negative at baseline, Wilcoxon rank sum test
c No significant difference between APOE ε2-positive and APOE ε2-negative in differences between week 0–3 or week 7–10, Wilcoxon rank sum test
d Significant difference between week 0–3 for the APOE ε2-positive group and between week 0–3 and week 7–10 for the APOE ε2-negative group, Signed rank test
Coffee consumption resulted in a significant cholesterol increase in the ε2-negative group in both trial periods (w 3–7 and w 10–14), but not in the ε2-positive group (Table 3). There were no differences between the ε2-positive and the ε2-negative group according to baseline characteristics as sex, age, body mass index (BMI) and coffee consumption prior to the study (Table 4).
Table 3 Serum cholesterol concentration (mmol/L) after two 4-week periods of coffee consumption (week 3 – 7 and week 10 – 14) for APOE ε2-positive (n = 13) and APOE ε2-negative (n = 108) individuals
APOE ε2-positive APOE ε2-negative p
n = 13 n = 107/103 a
First trial period
week 3 4.49 (0.71) 5.05 (0.90)
week 7 4.52 (0.71) 5.34 (0.93)
diff week 3–7 0.03 (0.57) 0.29 (0.57) 0.09 b
p (diff 3–7) 0.70 <0.0001 c
Second trial period
week 10 4.34 (0.64) 4.95 (0.89)
week 14 4.54 (0.84) 5.09 (0.85)
diff week 10–14 0.20 (0.47) 0.14 (0.59) 0.37 b
p (diff 10–14) 0.15 0.009 c
a 107 participants in the first trial period and 103 participants in the second trial period
b No significant difference between 2-positive and 2-negative in differences between week 3–7 or week 10–14, Wilcoxon rank sum test
c Significant difference between week 0–3 and week 7–10 for the APOE ε2-negative group, Signed rank test
Table 4 Baseline characteristics for APOE ε2-positive (n = 13) and APOE ε2-negative (n = 108) individuals
APOE ε2-positive APOE ε2-negative p
n = 13 n = 108
Sex (% women) 77 79 ns a
Age (years) 46.6 48.7 0.44 b
BMI (kg/m2) 25.7 25.8 0.87 b
Coffee consumption (cups/day) 4.3 3.7 0.12 b
a No significant difference between APOE ε2-positive and APOE ε2-negative at baseline, Chi square test
b No significant difference between APOE ε2-positive and APOE ε2-negative at baseline, Wilcoxon rank sum test
Dietary monitoring and compliance
The dietary survey did not reveal any important changes during the four intervention periods [22]. Coffee consumption or non-compliance was reported by six persons during the first coffee abstention period (mean 1.8 cups/period), whereas four persons reported coffee consumption in the second coffee abstention period (mean 0.7 cups/period).
Discussion
Subjects with different APOE genotypes differ in the absorption efficiency of cholesterol from the intestine, in the synthesis rates of cholesterol and bile acids, and in the production of LDL apoprotein B [3,26]. This suggests that the response of serum cholesterol to diet may be affected by the APOE e2/e3/e4 polymorphism [27,28]. One previous study examined the effect of purified cafestol on serum lipids in relation to the APOE polymorphism [26] and found that responses of LDL-cholesterol to cafestol were slightly smaller among carriers of the APOE ε4 allele. Here, we investigate for the first time the possible influence of the APOE polymorphism on the cholesterol-raising effect of filtered coffee.
APOE ε4-positive individuals did not differ significantly from ε4-negative individuals with regard to baseline cholesterol concentration or coffee-induced changes in cholesterol concentration. However, we confirm that ε2-positive individuals display significantly lower cholesterol levels at baseline than ε2-negative individuals. A tendency was seen that ε2-positive individuals might be partly protected from the cholesterol increasing effect of coffee consumption. This was, however, only seen in the first trial period and will require further investigations. In conclusion, the APOE ε2/ε3/ε4 polymorphism is not a strong modulator of the cholesterol-increasing effect of coffee. Other genes should be discussed and further investigation is needed to see if there is a genetic factor in the cholesterol-raising effect of coffee.
Acknowledgements
This work was supported by grants from the Swedish Medical Research Council (project #12103), the Sahlgrenska University Hospital and the Swedish Council for Working Life and Social Research.
==== Refs
Ou T Yamakawa-Kobayashi K Arinami T Amemiya H Fujiwara H Kawata K Saito M Kikuchi S Noguchi Y Sugishita Y Hamaguchi H Methylenetetrahydrofolate reductase and apolipoprotein E polymorphisms are independent risk factors for coronary heart disease in Japanese: a case-control study Atherosclerosis 1998 137 23 28 9568733 10.1016/S0021-9150(97)00244-X
Davignon J Gregg RE Sing CF Apolipoprotein E polymorphism and atherosclerosis Arteriosclerosis 1988 8 1 21 3277611
Eichner JE Dunn ST Perveen G Thompson DM Stewart KE Stroehla BC Apolipoprotein E polymorphism and cardiovascular disease: a HuGE review Am J Epidemiol 2002 155 487 495 11882522 10.1093/aje/155.6.487
Mahley RW Rall SC APOLIPOPROTEIN E: Far More Than a Lipid Transport Protein Annu Rev Genomics Hum Genet 2000 1 507 537 11701639 10.1146/annurev.genom.1.1.507
Song Y Stampfer MJ Liu S Meta-analysis: apolipoprotein E genotypes and risk for coronary heart disease Ann Intern Med 2004 141 137 47 15262670
Tammi A Rönnemaa T Rask-Nissilä L Miettinen TA Gylling H Valsta L Viikari J Välimäki I Simell O Apolipoprotein E Phenotype Regulates Cholesterol Absorption in Healthy 13-Month-Old Children-The STRIP Study Pediatr Res 2001 50 688 691 11726725
Tammi A Rönnemaa T Viikari J Jokinen E Lapinleimu H Ehnholm C Simell O Apolipoprotein E4 phenotype increases non-fasting serum triglyceride concentration in infants – the STRIP study Atherosclerosis 2000 152 135 141 10996348 10.1016/S0021-9150(99)00436-0
Stengård JH Pekkanen J Ehnholm C Nissinen A Sing CF Genotypes with the apolipoprotein ε4 allele are predictors of coronary heart disease mortality in a longitudinal study of elderly Finnish men Hum Genet 1996 97 677 684 8655152 10.1007/s004390050115
Ilveskoski E Perola M Lehtimäki T Laippala P Savolainen V Pajarinen J Penttilä A Lalu KH Männikkö A Liesto KK Koivula T Karhunen PJ Age-dependent association of apolipoprotein E genotype with coronary and aortic atherosclerosis in middle-aged men. An autopsy study Circulation 1999 100 608 613 10441097
Corbo RM Scacchi R Vilardo T Ruggeri M Polymorphisms in the Apolipoprotein E gene regulatory region in relation to coronary heart disease and their effect on plasma Apolipoprotein E Clin Chem Lab Med 2001 39 2 6 11256794 10.1515/CCLM.2001.002
Liu S Ma J Ridker PM Breslow JL Stampfer MJ A prospective study of the association between APOE genotype and the risk of myocardial infarction among apparently healthy men Atherosclerosis 2003 166 323 9 12535745 10.1016/S0021-9150(02)00335-0
Stengård JH Weiss KM Sing CF An ecological study of association between coronary heart disease mortality rates in men and the relative frequencies of common allelic variations in the gene coding for apolipoprotein E Hum Genet 1998 103 234 41 9760210 10.1007/s004390050811
Gerdes LU Gerdes C Kervinen K Savolainen M Klausen IC Hansen PS Kesaniemi YA Faergeman O The apolipoprotein allele determines prognosis and the effect on prognosis of simvastatin in survivors of myocardial infarction. A substudy of the Scandinavian Simvastatin Survival Study Circulation 2000 101 1366 71 10736278
Corder EH Saunders AM Strittmatter WJ Schmechel DE Gaskell PC Small GW Roses AD Haines JL Pericak-Vance MA Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer's disease in late onset families Science 1993 261 921 923 8346443
Poirier J Davignon J Bouthillier D Kogan S Bertrand P Gauthier S Apolipoprotein E polymorphism and Alzheimer's disease Lancet 1993 342 697 699 8103819 10.1016/0140-6736(93)91705-Q
Strittmatter WJ Saunders AM Schmechel D Pericak-Vance M Enghild J Salvesen GS Roses AD Apolipoprotein E: high-avidity binding to β-amyloid and increased frequency of type 4 allele in late-onset familial Alzheimer's disease Proc Natl Acad Sci USA 1993 90 1977 1981 8446617
Urgert R Katan B The cholesterol-raising factor from coffee beans Annu Rev Nutr 1997 17 305 324 9240930 10.1146/annurev.nutr.17.1.305
Bak A Coffee and cardiovascular risk; an epidemiological study PhD thesis 1990 Erasmus Univerisity Rotterdam
Ahola I Jauhiainen M Aro A The hypercholesterolaemic factor of boiled coffee is retained by a paper filter J Inter Med 1991 230 293 297
Van Dusseldorp M Katan MB Van Vliet T Demacker PNM Stalenhoef A Cholesterol-raising factor from boiled coffee does not pass a paper filter Arterioscler Thromb 1991 11 586 593 2029499
Christensen B Mosdøl A Retterstøl L Landaas S Thelle DS Abstention from filtered coffee reduces the levels of homocysteine and cholesterol – a randomized, controlled trial Am J Clin Nutr 2001 74 302 307 11522552
Strandhagen E Thelle DS Filtered coffee raises serum cholesterol: results from a controlled study Eur J Clin Nutr 2003 57 1164 1168 12947437 10.1038/sj.ejcn.1601668
Grubben MJ Boers GH Blom HJ Broekhuizen R de Jong R van Rijt L de Ruijter E Swinkels DW Nagengast FM Katan MB Unfiltered coffee increases plasma homocysteine concentrations in healthy volunteers: a randomized trial Am J Clin Nutr 2000 71 480 484 10648261
Nes M Frost Andersen L Solvoll K Sandstad B Hustvedt BE Løvø A Drevon C Accuracy of a quantitative food frequency questionnaire applied in elderly Norwegian women Eur J Clin Nutr 1992 46 809 21 1425535
Blennow K Ricksten A Prince JA Brookes AJ Emahazion T Wasslavik C Bogdanovic N Andreasen N Batsman S Marcusson J Nagga K Wallin A Regland B Olofsson H Hesse C Davidsson P Minthon L Jansson A Palmqvist L Rymo L No association between the alpha2-macroglobulin (A2M) deletion and Alzheimer's disease, and no change in A2M mRNA, protein, or protein expression J Neural Transm 2000 107 1065 79 11041282 10.1007/s007020070052
Weggemans RM Zock PL Ordovas JM Pedro-Botet J Katan MB Apoprotein E genotype and the response of serum cholesterol to dietary fat, cholesterol and cafestol Atherosclerosis 2001 154 547 555 11257255 10.1016/S0021-9150(00)00720-6
Masson LF McNeill G Avenell A Genetic variation and the lipid response to dietary intervention: a systematic review Am J Clin Nutr 2003 77; 1098 1111 12716659
Ordovas JM Gene-diet interaction and plasma lipid responses to dietary intervention Biochem Soc Trans 2002 30 68 73 12023827 10.1042/0300-5127:0300068
| 15571629 | PMC539242 | CC BY | 2021-01-04 16:39:18 | no | Lipids Health Dis. 2004 Nov 30; 3:26 | utf-8 | Lipids Health Dis | 2,004 | 10.1186/1476-511X-3-26 | oa_comm |
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-2-681556662710.1186/1477-7525-2-68ResearchRelationship between three palliative care outcome scales Higginson Irene J [email protected] Nora [email protected] Department of Palliative Care and Policy, King's College London, Weston Education Centre, Cutcombe Road, London SE5 9RJ, UK2004 29 11 2004 2 68 68 6 10 2004 29 11 2004 Copyright © 2004 Higginson and Donaldson; licensee BioMed Central Ltd.2004Higginson and Donaldson; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Various scales have been used to assess palliative outcomes. But measurement can still be problematic and core components of measures have not been identified. This study aimed to determine the relationships between, and factorial structure of, three widely used scales among advanced cancer patients.
Methods
Patients were recruited who received home or hospital palliative care services in the south of England. Hope, quality of life and palliative outcomes were assessed by patients in face to face interviews, using three previously established scales – a generic measure (EQoL), a palliative care specific measure (POS) and a measure of hope (Herth Hope Index). Analysis comprised: exploratory factor analysis of each individual scale, and all scales combined, and confirmatory factor analysis for model building and validation.
Results
Of 171 patients identified, 140 (81%) consented and completed first interviews; mean age was 71 years, 54% were women, 132 had cancer. In exploratory analysis of individual means, three out of the five factors in the EQoL explained 75% of its variability, four out of the 10 factors in POS explained 63% of its variability, and in the Hope Index, nine out of the 12 items explained 69% of its variability. When exploring the relative factorial structure of all three scales, five factors explained 56% of total combined variability. Confirmatory analysis reduced this to a model with four factors – self-sufficiency, positivity, symptoms and spiritual. Removal of the spiritual factor left a model with an improved goodness of fit and a measure with 11 items.
Conclusion
We identified three factors which are important outcomes and would be simple to measure in clinical practice and research.
palliative carequality of lifeassessmenthopesymptomshospiceday care
==== Body
Background
Measurement of the effect of illness and its treatment on patients is now an accepted part of clinical trial design [1]. Such measurement is also proposed as an aid to improve clinical practice and decision making [2,3]. However, as the illness becomes more advanced the value of many well validated quality of life instruments has been challenged [4-9]. There are three main difficulties. First, many quality of life scales focus on the assessment of physical functioning, which deteriorates as illness progresses [4,8]. This can render the measure insensitive to, or mask, other changes. Second, most quality of life scales have been validated among patients in early stage illness, such as cancer or whilst undergoing chemotherapy or curative treatment [8,9]. Sometimes their validation was founded on an assumption that patients in terminal disease had a poorer quality of life than those at diagnosis [10]. This assumption has been consistently challenged [8]. Concerns among patients with more advanced illness are often different to earlier stages, as patients reframe their priorities in the face of impending death [8]. Existential, relationships, information, the provision of care, and use of remaining time become more important [9]. Third, collecting information from patients at late stages of disease is practically difficult; questionnaires need to be kept short, be easy to use, and be few in number. Even then there are often difficulties of missing data and loss to follow-up [8,9].
In response to these difficulties, different measures have been developed and tested among patients receiving palliative and hospice care in different countries and contexts [8,11]. These include scales that assess, to different degrees, symptoms, existential aspects or spirituality, the impact of therapy, hope, information, social and family concerns [8,9,12]. Some are completed directly by patients, some by family members or other proxies, and some by a combination of these. However, there is little information on how different measures compare, particularly in relation to more traditional measures. Clinicians and researchers need such information to determine which core factors should be measured, especially when it is not possible to collect a battery of measures. This study therefore sought to determine the relationships between three such scales and their factorial structures to recommend short, self-contained scales for future use among people with advanced cancer.
Methods
Design
Secondary analysis of a prospective observational study.
Patients and setting
Patients living in Chichester in the South of England receiving home or hospital palliative care support, from community, hospice or hospital palliative care team staff, were approached to take part in the study. Local research ethics committee approval was obtained. The local hospice was planning to develop a day care unit and patients were recruited during this period. A historical group was recruited before the day care unit was established. Consecutive consenting patients were recruited for both series. Patients were eligible if they were in the care of the hospice home care team, or neighbouring home care teams, that had access to the day care unit. Patients were excluded if they were judged by staff to be too ill for interview, if staff felt it would distress them, or they lived outside the catchment area of the hospice day unit. Two concurrent groups were recruited after day care was established – patients who did (Group AD) and did not (Group AN) choose to receive day care.
Data collection
Data was collected using trained interviewers. Interviews took place in the patients' preferred location, usually their own home. Interviewers explained the background to the study and used a structured schedule to collect clinical, demographic and use of service data. They then administered three scales. All were short, taking less than 10 minutes on average to complete, and were acceptable to advanced cancer patients. Scales were administered in the order they are listed below.
1. EQoL EQ-5D. This generic questionnaire defines health in five dimensions: mobility, self-care, usual activities, pain or discomfort, and anxiety or depression. Each dimension is divided into three categories – whether the respondent has no problem, a moderate problem, or an extreme problem. A sixth item scores the person's overall health on a visual analogue (0 – 100) scale. The questionnaire has been validated and applied in a wide range of patient groups [13-16].
2. Palliative care Outcome Scale (POS). This 10 item scale (plus an open question) was specifically developed and validated for palliative care and covers physical symptoms, patient and family or caregiver anxiety/fears and well being. The effect of the items on the daily life of the patient over the past three days is scored on a five-point Likert scale ranging from 'none' (0) to 'overwhelmingly' (4). For example: "over the past 3 days, have you been feeling anxious or worried about your illness or treatment? (0) not at all – (4) overwhelmingly" [17,18]. In the POS the term 'family' describes the caregiver or significant other, such as a partner, spouse or other closest individual.
3. Herth Hope Index (Hope). This 12 item instrument assesses hope in adults in clinical settings, and is designed to assess change. For example: "I have a positive outlook toward life? strongly disagree (1) to strongly agree (4) -". Patients are asked how much they agree with the statement right now [19].
Full details of the scales are shown in the Appendix 1 (see additional file 1). Patients were interviewed immediately after referral to the study. Follow-up interviews occurred but these data are not considered here.
Analysis
Data were analysed separately for the historical and concurrent (post day care) groups. The relative factorial structure of the three scales was explored in two steps. First, we performed a preliminary exploratory factor analysis (EFA) on each individual scale and on all the items of the three scales combined, using Principal Component Analysis on the historical sample. Second, we performed further exploration and final validation using confirmatory factor analysis (CFA) on the combined historical and concurrent samples. The EQS software [20] was used to compare several models to the covariance matrix of the 28 variables. Although this was an observational follow-up study, for the purpose of this paper we always used the baseline measures, when complete data for all patients was available.
Results
171 patients were identified and asked to take part in the study, 82 in the historical group, and 89 in the concurrent group (40 were AD). Of these, 140 (81%) were successfully approached, agreed to take part in the study, and completed the first interview. As shown in Table 1, 66 were from the historical group and 74 were from the concurrent group (of whom 28 were AD). Failure to interview was due to: refused 12, felt too unwell 11, died 8. Complete data in all three scales were obtained in 137 patients. As Table 1 shows the AD and AN were similar, and so were subsequently merged to form the concurrent group. The concurrent and historical samples were very similar in terms of characteristics like age, ethnicity, willingness to take part in the study, diagnosis, as well as their relationship to the carer and whether they resided with family, spouse or alone and housing. In spite of the age similarity the proportion of retired people was slightly larger in the concurrent sample. Differences between the two samples were only observed for place of death and for gender. Although not statistically significant, the concurrent sample tended to have more patients dying at home while the historical sample tended to have more patients dying in hospice. The proportion of women was larger in the historical sample (60% vs 40% P = 0.02). The distribution of cancers was similar to those in the general population.
Table 1 Patient socio-demographics (completed 1st interview) for historical group and concurrent group who did (AD) and did not (AN) receive day care
Demographics Historical Group (n = 66) Group AD (n = 28) Group AN (n = 46)
Age in years
Mean (SD) 69.2 (12.4) 74.0 (10.1) 70.8 (11.9)
Median/range 71.0/34–94 77.0/50–94 72.0/39–90
Gender
Women 40 (61%) 12 (43%) 23 (50%)
Men 26 (39%) 16 (57%) 23 (50%)
Ethnicity
White UK 66 (100%) 28 (100%) 46 (100%)
Employment Status
Working (F/T or P/T) 6 (9%) 2 (7%) 2 (4.5%)
Not working (unable) 14 (22%) 2 (7%) 9 (20.5%)
Retired 45 (69%) 23 (85%) 33 (75%)
Carer
Spouse 43 (65%) 20 (71%) 30 (70%)
Other carer 12 (18%) 5 (18%) 9 (21%)
No carer 11 (17%) 3 (11%) 4 (9%)
Carer Contact
Lives with spouse 43 (65%) 20 (71%) 30 (70%)
Lives with family 2 (3%) 0 2 (5%)
Lives alone 21 (32%) 8 (29%) 11(25%)
Carer employment
Working (F/T or P/T) 18 (29%) 6 (21%) 8 (19%)
Not working (unable) 5 (8%) 2 (7%) 4 (9%)
Retired 29 (46%) 17 (61%) 27 (63%)
No carer 11 (17%) 3 (11%) 4 (9%)
Housing
Own/private 28 (42%) 11 (39%) 17 (37%)
Own/council 7 (11%) 7 (25%) 7 (15%)
Own/rented 28 (42%) 9 (32%) 20 (44%)
Other (N/home) 3 (5%) 1 (4%) 2 (4%)
Primary diagnosis
Lung cancer 11 (17%) 4 (14%) 11 (26%)
Gastrointestinal 11 (17%) 8 (29%) 9 (21%)
Breast 9 (14%) 4 (14%) 4 (10%)
GU/Prostate 11 (17%) 6 (21%) 8 (19%)
Gynae 7 (11%) 0 3 (7%)
Other cancer 10 (15%) 4 (14%) 7 (17%)
Non-cancer 6 (9%) 2 (7%) 0
Place of death (n = 46) (n = 11) (n = 22)
Home 8 (17%) 4 (36%) 6 (27%)
Hospital 7 (15%) 1 (9%) 3 (14%)
Hospice 31 (67%) 6 (55%) 13 (59%)
Individual Scales
Summaries of the distribution of scores on the three instruments assessing hope, quality of life and palliative outcomes for the combined sample as well as results of the exploratory factor analysis, are shown in Table 4 (see additional file 2). On principal component analysis (unrotated), three factors in EQoL explained 75% of the total variability brought up by the six items in this scale. The first factor, explaining 40%, comprised general health: Health Status and the three self-sufficiency items. Anxiety-Depression defined the second factor, which explained 20% of the variability, and Pain-Discomfort formed the third factor, which explained 15% of the variability. For POS, the exploratory factor analysis grouped the 10 POS items in four factors explaining 67% of its variability. The first factor, which alone explained 27% of the variability, consisted of the two items measuring positivism (life-worthwhile and feel-good) and in addition, worry-anxiety. The second factor, which alone explained 16% of the variability, was mainly determined by information, followed by time-wasted and practical-matters. The third factor, which explained 12% of the variability, was solely determined by the item family-anxious. The fourth factor, also explaining 12% of the variability, was determined by pain and symptom-control. In the individual exploration we found that the 12 items of Hope grouped into four factors that explained approximately 69% of the variability present in the scale. The first factor was items 1, 8, 10, 11, and 12 representing positivity (39%), the second factor had items 2 and 4, measures of goals (12%), the third was items 3 and 6 (10%) and the fourth was items 7, 9 and 5 (9%). These last two factors represented a measure of support loneliness.
Three scales combined
The exploratoty factor analysis of EQoL, POS and Hope on the historical sample alone gave rather consistent results for different extraction methods. Table 4 (see additional file 2) shows the results of the unrotated principal component analysis. Five factors explained 54% of the total variability present in the three combined scales. The first principal factor, explaining 25% of the total variability of the combined scales, was determined by the three items of positivity contained in POS (share-feelings, feel-good and life worthwhile), together with all the Hope items and the anxiety items in both, POS and EQoL. The EQoL items "General Health" and items of "self-sufficiency", which constituted the most important factor of the EQoL scale, loaded together into the second factor, explaining 10% of the total variability of the combined scale. The third most important principal component, explaining 8% of the variability, comprised a general measure of patient anxiety (measured by both EQoL and POS), and family anxiety (measured by POS). The fourth principal component explained 6% of the variability and was defined by pain (measured by both EQoL and POS). The POS items "information" and "time-wasted" loaded together into the fifth factor, explaining only 6% of the total variability. In addition, the POS item "symptom control" did not load into any of these five factors and appeared to be acting independently.
One of the extractions explored was principal axis factoring with a varimax rotation. This provided a better definition of the structure, with items loading more exclusively onto one of the factors. The first factor that we had obtained with the unrotated matrix essentially separated into two. The first axis, explaining 29% of the variability, was defined by the POS item "life worthwhile" loading with those items of Hope that reflected positivity and direction: positive outlook, goals, inner strength, loving, sense of direction, days have potential and life has value. The second axis, explained 11% of the variability and contained the anxiety items of EQoL and POS, the "feel good" and "share feeling" items of POS and the items of Hope that reflected pessimism or anxiety: "alone", "scared of future" and "past memories". The third factor was the EQoL general health and self-sufficiency and explained 7% of the variability. The fourth factor was solely defined by the pain items in EQoL and POS and explained 6% of the variability. The rest of the items played only a minor role. The POS items: practical matters, information and time wasted loading in a minor factor while the POS item "family-anxious" and the Hope items "tunnel" and "faith" disappeared altogether from the rotated matrix. Therefore, the model derived from this data is one in which the following items are omitted: POS2 and POS4 from the POS scale and Hope4 Hope5 from the Hope scale, leaving the rest of the items to define four major latent factors in the following manner: Spiritual, positivity, symptoms and self-sufficiency.
Confirmatory Factor Analysis (CFA)
Several models were explored and the most relevant are presented in Table 2 with the various measures of fit given by EQS. Model 1 was a three-factor model allowing each scale (EQoL, POS and Hope) to individually determine each factor. The goodness of fit measures suggest that the model does not provide a good fit for the data, although most of the residuals (observed-predicted covariances) were found to be relatively small and their frequency distribution is symmetric and centred around zero [21]. This model confirmed that the POS and Hope factors were indeed very highly correlated [Estimated correlation = 0.81; 95% CI (0.71–0.91)]. Consequently, a second model was fitted to the data in which only two factors were postulated, the first was the EQoL items as in the previous model and the second factor having as its indicators both the POS variables and the HOPE variables. The fit was very similar to the fit of Model 1 but it appeared that the two factor model needs to be considered as a serious alternative to model 1. In addition, the results of these two models suggest that some of the POS variables (family anxious, information given, time wasted and practical matters) are not needed for defining the second factor. As a result of this, we explored a range of models, allowing for the strong correlation between POS and Hope and giving special attention to those items that were unimportant in either the exploratory or preliminary confirmatory factor analysis. Three of the POS items, which confirmed a latent construct that we called "practical", proved to be superfluous in the overall construct. These items were: information given, time wasted and practical matters. The POS item family-anxious did not particularly disrupt the identifiability of the model but its presence reduced the goodness of the fit. Similarly, four Hope items were discarded – alone, light at the end of the tunnel, faith and scared of future – to give a total of seven items discarded. We arrived to two models, exhibited in Table 2: Model 3, fitting the four factors listed in Table 3, and Model 4, fitting only the first three factors, leaving out the spiritual factors construed by the Hope scale. Table 2 includes the goodness of fit statistics for these models.
Table 2 Goodness of fit summaries for the four models derived by Confirmatory Factor Analysis (CFA)
Model 1 Model 2 Model 3 Model 4
Independence 1076 1076 717 279
Chi-square (378 df) (378 df) (171) (55 df)
Average standardised residuals 0.11 0.09 0.09 0.09
Average off-diagonal st. residuals 0.12 0.10 0.10 0.11
Chi-squared fit 534.7 520 213 67.7
(df)s (347 df) (347 df) (150 df) (43)
P-value 0.00001 0.00001 0.001 0.01
Free parameters 59 57 40 23
Akaike's information criterion (AIC) -193 -173 -87 -18
Bozodgan's version of AIC (C-AIC) (-1437) (-1424) (-627) (-173)
Comparative Fit Index (CFI) 0.73 0.75 0.89 0.90
Normed Fit Index (NFI) 0.50 0.52 0.70 0.76
Non-normed Fit Index (NNFI) 0.71 0.73 0.87 0.86
Model 1 comprised the basic 3 factors: EQoL, POS and HOPE. Model 2 was 2 factors: EQoL, and POS and HOPE combined. Model 3 was 4 factors: items relating to self-sufficiency, positivity, symptoms and spiritual. Model 4 was 3 factors, items relating to self-sufficiency, positivity and symptoms.
AIC and CAIC measure degree of fit. The smaller, the better the fit. The larger are NFI, NNI and CFI, the better the fit, with an upper value of 1.
Table 3 The factorial structure of the proposed model (MLE Estimators of regression coefficients (Standard Error)
Scale Item SYMPTOMS SELF SUFFICIENCY POSITIVITY SPIRITUAL
EQoL1 Mobility 0.32 (0.06)
EQoL2 Self-care 0.49 (0.08)
EQoL3 Usual activities 0.39 (0.08)
EQoL4 Pain-Discomfort 0.38 (0.08)
EQoL5 Anxty-Depression 0.34 (0.06)
EQoL6 Health Status -6.9 (2.34)
POS1 Pain Control 0.93 (0.17)
POS2 Symptom Control 0.16 (0.11)
POS3 Anxious/Worried 0.52 (0.11)
POS4 Family anxious 0.26 (0.15)
POS5 Information
POS6 Share feelings 0.69 (0.15)
POS7 Life Worthwhile 0.72 (0.10)
POS8 Feel Good 0.97 (0.12)
POS9 Time Wasted
POS10 Practical matters
HOP1 Positive outlook 0.42 (0.07)
HOP2 Goals 0.47 (0.08)
HOP3 Alone
HOP4 Tunnel
HOP5 Faith
HOP6 Scared of future
HOP7 Happy memories 0.27 (0.07)
HOP8 Inner strength 0.65 (0.08)
HOP9 Loving 0.32 (0.07)
HOP10 Sense of direction 0.77 (0.08)
HOP11 Days Potential 0.67 (0.07)
HOP12 Life has value 0.55 (0.08)
Significant coefficients are highlighted.
In all the models presented, the matrix was positively definite, the test of independence was significant and the frequency distribution of the standardised residuals was symmetrical around 0. Models 3 and 4, not only omit the superfluous items but also separate the factors on clinical considerations. Both provide a huge improvement over the first two models. Model 3 allowed for a high correlation between the positivity and spiritual factors. More remarkably, the results show that Model 4, which disposes completely of the spiritual factor defined by the remaining Hope items, is an enormous improvement on the other models. The chi-square statistic was greatly reduced and almost reached the threshold indicating that no lack of fit was detected.
Discussion
An important next step in quality of life measurement is the translation of measurement into clinical practice to improve patient care [2]. One important barrier among patients with advanced illness is ensuring that relevant items are captured from a sufficiently small range of instruments relevant to this stage of illness. The three measures used in this study all have relevance in advanced illness. The EQoL deals with general aspects of health related quality of life, generating within the scale 243 possible health states. It has been used to provide indexed preferences for health states [22], and health state valuations in national and cross cultural studies [23]. Standardised measures, such as the Medical Outcome Study (MOS) short form 12 (SF-12) map to this scale [24]. Among our patients with advanced illness, primarily cancer, we found variability within the EQoL, although patients tended to score at the poorer end of the scale. Health status and the self-sufficiency items of mobility, self-care and usual activities explained 40% of the variation of this scale in our patient population. We included the self-sufficiency aspect in our model of summary factors, but it is debatable whether items such as mobility, self-care and usual activities are relevant outcomes in palliative care. Functional status and those items within quality of life measures that reflect functional status are highly correlated to survival [25], thus the scores will inevitably deteriorate towards death. However, to provide consistency with other scales used in general health care and cancer treatment, measurement may be valuable [24].
A factor which we entitled 'positivity' appeared to be highly important among people with advanced illness. Spirituality/positivity has also been related to communication outcomes [26]. In the exploratory factor analysis it explained 24% of the total variability of the combined scales. Its importance was maintained in the confirmatory factor analysis. In model 3 positivity could be seen as separate from spirituality, but if a smaller model is required, spirituality can be assessed through positivity, because it is strongly correlated. Items that reflect this domain of positivity are found in a number of measures of palliative care [9,12,18]. However, our study is the first to quantify the extent to which this positive domain is relevant in patients with advanced illness. Our data suggests it can be measured in a variety of ways, through questions related to sharing feelings, feeling good, anxiety, as well as questions directly tapping hope.
When attempting to develop a reduced scale we identified two models, one with four factors (19 items) and another one with three factors which provided a good fit (11 items). All the Herth Hope items are excluded from the latter model, which captured self-sufficiency, symptoms and positivity. Positivity appeared very close to spirituality, as measured by the Hope index. Further work is needed to determine the relationship of these questions with specific spirituality scales [27-29].
Symptom control was absent in the structures obtained by EFA. We suspected that this was because this item was not a structured question designed for any specific symptom, but elicited in an open way what symptoms had troubled the patient. In the CFA this item loaded with the General Health status factor. Measures which specifically address symptoms are available and have been used in palliative care populations [29,30]. Work with the POS has now developed to separate symptom modules and these are in the process of further testing and validation [31].
Special attention was given to the three items forming the practical factor in POS. The information item in POS (POS-5) was constant in the concurrent sample and only a few patients in the historical sample recorded non zero for this item. The time wasted item of POS was also essentially constant. It may be that the grading for these items needs to be reviewed to ensure that they give greater sensitivity to change. In our analysis this could have contributed to the poor fit shown when attempting to fit a general POS factor containing these items. It may also be because that none of the three items is an indicator of QOL; they are rather items of the provision of health care. In this study all of the patients were in receipt of a wide range of services, including specialist care teams and their practical needs were likely to have been met. Research among patients in different circumstances has shown deficiencies in these aspects of care [32,33].
The POS item family-anxious was intriguing. It did not disrupt the validity of the model but if kept and loaded in the positivity factor, it reduced the goodness of fit. This item also showed erratic behaviour in the exploratory factor analysis. Family anxiety may be related to a large number of factors, some of which are determined by the circumstances of the patient and some of which are determined by other events. Family needs often increase as patients deteriorate and are difficult to resolve. Further work is required directly targeting the needs of families [34].
Our study was limited by the comparison of these scales among patients in one setting: we do not know if similar results would be found if patients were not in receipt of specialist palliative care available in the United Kingdom. However, our findings are consistent with other work assessing quality of life measurement in palliative care and in advanced cancer [11]. Correlation between the scales may also have occurred because individuals were aware of the answers they had given for the different scales. It would be impossible to avoid this process in the completion of the questionnaires. We did not vary the order in which the questionnaires were administered. However, we believed that the questions appeared to be sufficiently different for patients not to be influenced by their prior responses. Future research should analyse this.
Our reduced model suggests that clinicians may sensibly target quality of life measurement in advanced cancer towards three main components, positivity, self-sufficiency, and symptoms. This might be achieved by the 12 items in model 4 of our factor analysis. Such a scale would be short and simple to use in both clinical practice and research, improving the measurement of outcomes in this population.
Supplementary Material
Additional File 1
Appendix 1: Full details of the three palliative care outcome scales
Click here for file
Additional File 2
Table 4: Summary Statistics and Principal Component Analysis (unrotated) of the three scales (POS-EQoL-Hope) in the historical sample.
Click here for file
Acknowledgements
We thank the staff, patients and families at Chichester hospice, in particular Dr Brendan Amesbury, Medical Director, for their participation in this study, Danielle Goodwin and members of the Project Advisory Group for their participation in study design and data collection and Professor Brian Everitt for helpful discussions. We thank the NHS Executive, South East, for funding this project.
==== Refs
Bottomley A Efficace F Thomas R Vancoorden V Ahmedzai SH Health-related quality of life in non-small-cell lung cancer: methodologic issues in randomized controlled trials J Clin Oncol 2003 21 2982 2992 12885819 10.1200/JCO.2003.01.203
Levine MN Ganz PA Beyond the development of quality-of-life instruments: where do we go from here? J Clin Oncol 2002 20 2215 2216 11980989 10.1200/JCO.20.2.405
Higginson IJ Carr AJ Measuring quality of life: Using quality of life measures in the clinical setting BMJ 2001 322 1297 1300 11375237 10.1136/bmj.322.7297.1297
McMillan SC Weitzner MA Methodologic issues in collecting data from debilitated patients with cancer near the end of life Oncol Nurs Forum 2003 30 123 129 12515990
Rustoen T Fossa SD Skarstein J Moum T The impact of demographic and disease-specific variables on pain in cancer patients J Pain Symptom Manage 2003 26 696 704 12906954 10.1016/S0885-3924(03)00239-2
Hahn EA Cella D Health outcomes assessment in vulnerable populations: measurement challenges and recommendations Arch Phys Med Rehabil 2003 84 S35 S42 12692770 10.1053/apmr.2003.50245
Di Maio M Perrone F Quality of life in elderly patients with cancer Health and Quality of Life Outcomes 2003 1 44 14525617 10.1186/1477-7525-1-44
Hearn J Higginson IJ Outcome measures in palliative care for advanced cancer patients: a review J Public Health Med 1997 19 193 199 9243435
Cohen SR Mount BM Tomas JJN Mount LF Existential well-being is an important determinant of quality of life: Evidence from the McGill Quality of Life Questionnaire Cancer 1996 77 576 586 8630968 10.1002/(SICI)1097-0142(19960201)77:3<576::AID-CNCR22>3.0.CO;2-0
Spitzer WO Dobson AL Hall J Chesterman E Levi J Shepherd R Battista RN Catchlove BR Measuring the quality of life of cancer patients: a concise QL- Index for use by physicans J Chron Dis 1980 34 585 597 7309824 10.1016/0021-9681(81)90058-8
Kaasa S Loge JH Quality of life in palliative care: principles and practice Palliat Med 2003 17 11 20 12597461 10.1191/0269216303pm662ra
Steinhauser KE Bosworth HB Clipp EC McNeilly M Christakis NA Parker J Tulsky JA Initial assessment of a new instrument to measure quality of life at the end of life Journal of Palliative Medicine 2002 5 829 841 12685529 10.1089/10966210260499014
Brazier J Jones N Kind P Testing the validity of the EuroQoL and comparing it with the SF-36 health survey questionnaire Qual Life Res 1993 2 169 180 8401453 10.1007/BF00435221
Kind P Dolan P Gudex C Williams A Variations in population health status: results from a United Kingdom national questionnaire survey BMJ 1998 316 736 741 9529408
Van Agt H Essink-Bot ML Krabbe P Bonsel G Test re-test reliability of health state valuations collected with the EuroQoL questionnaire Soc Sci Med 1994 39 1537 1544 7817218 10.1016/0277-9536(94)90005-1
Hurst NP Jobanputra P Hunter M Lambert M Lochead A Brown H Validity of EuroQoL - a generic health status instrument - in patients with rheumatoid arthritis British Journal of Rheumatology 1994 33 656 662
Aspinal F Hughes R Higginson IJ Chidgey J Drescher U Thompson M A user's guide to the Palliative Outcome Scale 2002 London , Palliative Care and Policy Publications
Hearn J Higginson IJ Development and validation of a core outcome measure for palliative care: the palliative care outcome scale Quality in Health Care 1999 8 219 227 10847883
Herth K Hope in the family caregiver of terminally ill people J Adv Nurs 1993 18 538 548 8496501 10.1046/j.1365-2648.1993.18040538.x
Bentler PM EQS: Structural Equations Manual 1989 Los Angeles, BMDP Statistical Software
Dunn G Everitt B Pickles A Modelling Covariances and Latent Variables using EQS 1993 Chapman and Hall
Dolan P Modelling valuations for EuroQoL health states Medical Care 1997 35 1095 1108 9366889 10.1097/00005650-199711000-00002
Johnson JA Coons SJ Ergo A Szava-Kovats G Valuation of EuroQoL (EQ-5D) health states in an adult US sample Pharmacoeconomics 1998 13 421 433 10178666
Franks P Lubetkin EI Gold MR Tancredi DJ Jia H Mapping the SF-12 to the EuroQoL EQ-5D index in a national US sample Medical Decision Making 2004 24 247 254 15185716 10.1177/0272989X04265477
Khayat D Quality of life: It's never too late J Clin Oncol 2002 20 2915 2917 12089218
Higginson IJ Costantini M Communication in end-of-life cancer care: a comparison of team assessments in three European countries J Clin Oncol 2002 20 3674 3682 12202669 10.1200/JCO.2002.11.008
King M Speck P Thomas A The Royal Free Interview for religious and spiritual beliefs: development and standardization Psychol Med 1995 25 1125 1134 8637943
McClain CS Rosenfeld B Breitbart W Effect of spiritual well-being on end-of-life despair in terminally-ill cancer patients Lancet 2003 361 1603 1607 12747880 10.1016/S0140-6736(03)13310-7
Portenoy RK Thaler HT Kornblith AB Lepore JM Friedlander-Klar H Kiyasu E Sobel K Coyle N Kemeny H Norton L al The Memorial Symptom Assessment Scale: an instrument for the evaluation of symptom prevalence, characteristics and distress European Journal of Cancer 1994 30A 1326 1336 7999421 10.1016/0959-8049(94)90182-1
Modi S Panageas KS Duck ET Bach A Weinstock N Dougherty J Cramer L Hudis C Norton L Seidman A Prospective explanatory analysis of the association between tumour response, quality of life, and expenditures among patients receiving paclitaxel monotherapy for refractory metastatic breast cancer J Clin Oncol 2002 20 3665 3673 12202668 10.1200/JCO.2002.08.057
Sherry KL McAuley G Symptom prevalence and the use of systematic symptom assessment Palliat Med 2004 18 75 76 14982215
Higginson IJ Wade AM McCarthy M Effectiveness of two palliative support teams J Public Health Med 1992 14 50 56 1376133
Gysels M Higginson IJ Improving supportive and palliative care for adults with cancer: Research Evidence 2004 London, National Institute of Clinical Excellence
Harding R Higginson IJ What is the best way to help caregivers in cancer and palliative care? A systematic review of interventions and their effectiveness Palliat Med 2003 17 63 74 12597468 10.1191/0269216303pm667oa
| 15566627 | PMC539243 | CC BY | 2021-01-04 16:38:11 | no | Health Qual Life Outcomes. 2004 Nov 29; 2:68 | utf-8 | Health Qual Life Outcomes | 2,004 | 10.1186/1477-7525-2-68 | oa_comm |
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-2-781556373810.1186/1477-7827-2-78ResearchThe role of transforming growth factor-beta (TGF-beta) during ovarian follicular development in sheep Juengel Jennifer L [email protected] Adrian H [email protected] Karen L [email protected] Stan [email protected] Laurel D [email protected] Lisa J [email protected] Kenneth P [email protected] AgResearch, Wallaceville Animal Research Centre, Upper Hutt, New Zealand2004 25 11 2004 2 78 78 22 6 2004 25 11 2004 Copyright © 2004 Juengel et al; licensee BioMed Central Ltd.2004Juengel 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
Recently, several members of the transforming growth factor-beta (TGF-beta) superfamily have been shown to be essential for regulating the growth and differentiation of ovarian follicles and thus fertility.
Methods
Ovaries of neonatal and adult sheep were examined for expression of the TGF-betas 1–3 and their receptors (RI and RII) by in situ hybridization using ovine cDNAs. The effects of TGF-beta 1 and 2 on proliferation and differentiation of ovine granulosa cells in vitro were also studied.
Results
The expression patterns of TGF-beta 1 and 2 were similar in that both mRNAs were first observed in thecal cells of type 3 (small pre-antral) follicles. Expression of both mRNAs continued to be observed in the theca of larger follicles and was also present in cells within the stroma and associated with the vascular system of the ovary. There was no evidence for expression in granulosa cells or oocytes. Expression of TGF-beta 3 mRNA was limited to cells associated with the vascular system within the ovary. TGFbetaRI mRNA was observed in oocytes from the type 1 (primordial) to type 5 (antral) stages of follicular growth and granulosa and thecal cells expressed this mRNA at the type 3 (small pre-antral) and subsequent stages of development. The TGFbetaRI signal was also observed in the ovarian stroma and vascular cells. In ovarian follicles, mRNA encoding TGFbetaRII was restricted to thecal cells of type 3 (small pre-antral) and larger follicles. In addition, expression was also observed in some cells of the surface epithelium and in some stromal cells. In granulosa cells cultured for 6 days, both TGF-beta 1 and 2 decreased, in a dose dependent manner, both the amount of DNA and concentration of progesterone.
Conclusion
In summary, mRNA encoding both TGF-beta 1 and 2 were synthesized by ovarian theca, stroma and cells of the vascular system whereas TGF-beta 3 mRNA was synthesized by vascular cells. Luteinizing granulosa cells also responded to both TGF-beta 1 and beta 2 in vitro. These findings in sheep are consistent with TGF-beta potentially being an important autocrine regulator of thecal cell function and possibly a paracrine regulator of ovarian cell function at various development stages.
==== Body
Background
Members of the transforming growth factor-beta (TGF-β) superfamily are important intraovarian growth factors [1-6]. Three key members of the TGF-β subfamily, namely TGF-β1, TGF-β2 and TGF-β3, have been shown to be produced by ovarian cells [7-13]. However, the cellular distribution of these proteins varies between species. Likewise, the effects of TGF-βs on granulosa cell function also vary between species. In rodents, TGF-βs are potent stimulators of granulosa cell proliferation [14-16] whereas in other species, such as cattle and pigs, these growth factors have only a mild stimulatory or even inhibitory effect [17-20]. Likewise, TGF-βs stimulate progesterone synthesis from rodent granulosa cells [21-23] where inhibitory effects are observed in granulosa cells collected from sheep, cattle and pigs [17,24-26]. Exceptional ovulation rates and sterility have been observed in lines of sheep with mutations in two members of the TGF-β superfamily, namely growth differentiation factor 9 or bone morphogenetic protein 15 or one of their receptors, activin like kinase-6 [6,27]. However, little is known about the roles of other members of the TGF-β superfamily in sheep and thus the potential interactions of members of the TGF-β superfamily are unclear. The objectives of this study in sheep were to localize the ovarian cellular types expressing mRNA encoding TGF-β1, TGF-β2, TGF-β3, TGFβRI and TGFβRII and to determine the effects of TGF-β1 and TGF-β2 on granulosa cell proliferation/survival and progesterone production in vitro.
Methods
Generation of cDNAs encoding a portion of the coding region of genes of interest
Except where indicated, laboratory chemicals were obtained from BDH Chemicals New Zealand Ltd (Palmerston North, New Zealand), Invitrogen (Auckland, New Zealand) or Roche Diagnostics N.Z. Ltd. (Auckland, New Zealand).
Total cellular RNA was isolated from ovine ovary using TRIzol according to manufacturer's instructions. First strand cDNA was produced from total cellular RNA using a poly t primer. Complementary DNAs encoding a portion of the coding sequence of the genes were isolated using standard PCR techniques. For individual cDNAs generated, primer sequences and annealing temperature are given in table 1. Resulting PCR products were ligated into appropriate vectors and their nucleotide sequence determined by automated sequence analysis (Waikato DNA Sequencing Facility; The University of Waikato; Hamilton, New Zealand). These sequences were compared with known sequences to confirm identity using the GAP program of GCG (Wisconsin Package Version 10.2, Genetics Computer Group; Madison, Wisconsin). All sequences were >80% identical to those listed as reference sequences in table 1 indicating that the ovine homologue of the respective genes had been obtained.
Table 1 GenBank reference numbers used for primer design, primer sequences, annealing temperatures, and GenBank accession numbers for the resulting ovine sequence for the various genes amplified.
Gene Reference: (Genbank #) Forward Primer (5' to 3') Reverse Primer (5' to 3') Annealing temperature Genbank # (resulting sequence)
TGF-β1 NM_011577 ggaattcatgccgccctcggggctgcgg (EcoR I site and bases 867–888) ggtctagatcagctgcacttgcaggagcg (Xba I site and bases 2040–2020) 63°C ND
TGF-β2 M19154 ggaattcatgcactactgtgtgctgagc (EcoR I site and bases 468–488) ggtctagagctgcatttrcaagacttkac (Xba I site and bases 1794–1773) 64°C AY656797
TGF-β3 J03241 ggaattcgcaaagggctctggtggtcctgg (EcoR I site and bases 277–299) ggtctagaccagttctcctccaagttgcgg (Xba I site and bases 1206–1186) 62°C AY656798
TGFβRI U97485 cacagatgggctttgctttg (bases 180–199) ccttgggtaccaactatctc (bases 1007–988) 50°C AY656799
TGFβRII S69114 gtcctgtggacgcgcat (bases 80–97) aggagcacatgaagaaagtc (bases 449–430)* 50°C AY656800
TGFβRII (for PCR) various gccaacaacatcaaccac gggtcrtggtcccagca 53°C AY751461
TGFβRII (internal for PCR) AY751461 tcgccgaggtctacaagg atgccctggtggttgagc 55°C N/A
* Sequence is based on the corresponding ovine sequence obtained from an ovine est clone.
ND, the complete sequence of the clone was not determined, as the ovine TGF-β1 sequence is known. The clone was sequenced from both ends and resulting sequence compared to the known ovine TGF-β1 sequence to confirm identity of the isolated cDNA.
N/A as the sequence overlaps that of AY751461.
Collection of tissue samples
All experiments were performed in accordance with the 1999 Animal Welfare Act Regulations of New Zealand. All animals had ad lib access to pasture and water and lambs were kept with their mothers until just prior to tissue collection. Romney ewes and lambs were killed by administration of a barbiturate overdose (Pentobarbitone; 200 mg/kg, Southern Veterinary Supplies, Christchurch, New Zealand). Recovered ovaries were fixed in 4% (w/v) phosphate-buffered paraformaldehyde and embedded in paraffin wax.
In Situ Hybridization
Cellular localization of mRNAs was determined using the in situ hybridization protocol described previously with minor modifications [28]. Sense and anti-sense RNA probes were generated from cDNA encoding the gene of interest with T7, T3 or SP6 RNA polymerase using the Riboprobe combination system (Promega, Dade Behring Diagnostics Ltd., Auckland, New Zealand). For all in situ hybridizations, 4–6 μm tissue sections were incubated overnight at 55°C with 45,000 cpm/μl (approximately 48,000 dpm/μl) of 33P-labelled antisense RNA. Non-specific hybridization of RNA was removed by RNase A digestion followed by stringent washes (2 × SSC, 50% formamide, 65°C and 0.2 × SSC at 37°C). Following washing, sections were dehydrated, air dried and coated with autoradiographic emulsion (LM-1 emulsion; Amersham Pharmacia Biotech, New Zealand). Emulsion-coated slides were exposed at 4°C for 3 weeks, developed for 3 1/2 minutes in D19 developer (Eastman Kodak, Rochester, NY), development was stopped using a 1 minute incubation in 1% acetic acid and slides were fixed with a 10 minute incubation in Ilfofix II (Ilford Limited, Cheshire, England). Sections were stained with hematoxylin and then viewed and photographed using both light and dark field illumination on an Olympus BX-50 microscope (Olympus New Zealand Ltd., Lower Hutt, New Zealand). Non-specific hybridization was monitored by hybridizing at least two tissue sections from each age group (lamb and adult) with approximately equal concentrations of the sense RNA for each gene. Hybridization was considered to be specific when the intensity of silver grains, as measured by visual assessment, over a cellular type was greater than that observed in the area of the slide not containing tissue. For all genes, hybridization of the sense RNA over the tissue section was similar or lower in intensity to that observed on the areas of the slide not containing tissue of both the sense and antisense hybridized slides and thus was considered non-specific.
Follicular classification
Classification of follicles was based on the system outlined by Lundy et al. [29]. Briefly, type 1/1a follicles consist of an oocyte surrounded by a single layer of flattened or mixed flattened and cuboidal cells. Type 2 follicles contain 1 < 2 layers of cuboidal granulosa cells whereas type 3 follicles contain 2 < 4 layers of cuboidal granulosa cells. Type 4 follicles have >4 layers of granulosa cells and a well defined theca but have not yet formed an antrum. Type 5 follicles have multiple layers of granulosa cells, a well defined theca and a defined antrum. All follicles with signs of degeneration (i.e. pyknotic granulosa cells, lack of a distinct basement membrane or degenerate oocytes) were excluded from the study. Ovarian sections from a minimum of eight animals, including at least three lambs and three adults, were examined for each gene studied. In addition, each follicle class was observed in a minimum of three animals. No differences in expression patterns between lamb and adults ovaries were noted in this study.
Granulosa cell culture
Ovaries were collected from ewes following slaughter at the local abattoir, transported back to the laboratory at room temperature, washed in 3% bleach solution in PBS for 5 minutes, rinsed twice in PBS and stored in Leibovitz media containing 0.1% BSA, 100 U/ml penicillin and 100 μg/ml streptomycin. Follicles approximately 1–2 mm in diameter were dissected away from the ovaries and stored in Leibovitz media until collection of granulosa cells. The granulosa cells were collected by cutting follicles in half followed by manual scraping of cells from the follicular wall using a wire loop. Oocytes and follicular debris were removed from the cells using a micro-glass pipette. Remaining cells were collected by centrifugation at 300 g for 5 min at room temperature, washed once in 5 mls Leibovitz media, twice in 5 mls McCoys media (Sigma, Auckland, New Zealand) with 100 U/ml penicillin, 100 μg/ml streptomycin and 2 mM GlutaMAX-1 and resuspended using a syringe and needle. Cell viability was determined using trypan blue exclusion and 100,000 viable cells per well (250 μl total volume) were added in McCoys media containing 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM GlutaMAX-1, 5 ng/ml selenium (Sigma), 10 ng/ml insulin (Sigma), 5 μg/ml apo-transferrin (Sigma), 30 ng/ml androstenedione (Sigma), 3 ng/ml ovine FSH (purified in our laboratory; 1.4 X USDA-oFSH-19-SIAFP RP2), 1 ng/ml IGF-1 (Long-R3, GroPep, Adelaide, SA 5000, Australia) with varying doses (0–10 ng/ml) of purified human TGF-β1 and recombinant human TGF-β2 (R & D Systems, Minneapolis, MN). Cells were cultured at 37°C in a 5% CO2 incubator. Every 48 hours, 200 μl of media was removed from each well and replaced with 200 μl of warmed media that had been prepared at the start of the culture and stored at 4°C. Media samples from the last 48 hours of treatment were collected on day 6 of treatment and frozen at -20°C for later determination of progesterone concentrations by RIA. Unattached cells were removed by 2 washes with McCoys media at 37°C. Attached cells were lysed by incubating cells at 37°C in 100 μl distilled water for 1–2 hours followed by freezing at -70°C. All treatments were performed at least in triplicate with three independent pools of granulosa cells. Within an assay, individual values outside of 20% of the mean value for the treatment were discarded. Points in which at least 2 of the replicates were not within 20% of each other were regarded as missing data. This occurred for the 10 ng TGF-β1 measure of DNA in a single pool of granulosa cells.
Measurement of DNA
The amount of DNA present in each well was determined by comparing binding of Hoechst 33258 dye (Sigma, final concentration in well of 10 μg/ml) in samples to calf thymus DNA standard measured with a Wallac 1420 plate reader at 350 nm for excitation and 460 nm for emission. Sensitivity of the assay (+ two SD of control buffer value) was 33 ng per well and the intra- and inter-assay co-efficients of variation (CV), based on variability of the 100, 250, 1000 and 2500 ng standard curve points were 3.9% and 8.8%, respectively. No samples were below the sensitivity of the assay.
Measurement of Progesterone
Concentrations of progesterone in media were determined by RIA as described [30]. The sensitivity of the assay (90% maximum binding) was 13 pg/ml and the intra- and inter-assay CV, averaged for a standard pool sample at approximately 20%, 50% and 80% binding, was 8.3% and 19.7%, respectively. No samples were below the sensitivity of the assay.
Determination of expression of TGFβRII mRNA in cultured granulosa cells
Granulosa cells were collected as described above and either frozen immediately after collection or plated in 6 well culture dishes at a density of 1.0–1.5 × 106 viable cells per well in 2 mls of control (i.e. no TGF-β) culture media described above for 48 hours. At this time, unattached cells were removed by washing the wells twice with PBS. RNA was collected using TRIzol according to the manufacturer's instructions. First strand cDNA was produced from total cellular RNA using the SuperScript™ preamplification system for first strand cDNA synthesis. An initial PCR was performed with 4 week old ovary RNA to obtain the ovine sequence of a region of the TGFβRII gene which spans introns 4 and 5 in the human sequence (AY675319) and a second set of primers was designed based on the ovine sequence (see table 1). Expression of TGFβRII was determined by PCR using the Qiagen Taq PCR core Kit (Biolab Scientific Limited) and the internal ovine primers listed in table 1 with the following conditions: initial denaturing cycle of 3 minutes at 94°C followed by 40 cycles of denaturing at 94°C for 1 minute, annealing at 55°C for 1 minute and extension at 72°C for 2 minutes and a final extension at 72°C for 10 minutes. cDNA generated from a 4 week old lamb ovary was run as a positive control whereas replacement of cDNA with water was used as a negative control. Expression of TGFβRII was assessed by visualization of DNA bands of the correct size following gel electrophoresis. Identity of product was confirmed by sequencing.
Statistical analysis
Concentration of progesterone per μg DNA was calculated for individual wells before averaging for each treatment within each assay. Points in which at least 2 of the replicates were not within 30% of each other were regarded as missing data. Changes in the concentrations of progesterone in media and DNA content after culture were analysed with the general linear model procedure of SAS. Replicate was included in the model as baseline progesterone and DNA values varied among the granulosa cell pools. Differences between least square means were evaluated using least significant differences and were considered significant when p < 0.05. Data presented are least square means. The standard errors of least square means were 0.7 ng/well, 0.2 μg/well and 0.5 ng/μg for progesterone, DNA and p4 per DNA, respectively.
Results
In situ hybridization
TGF-β1
The mRNA for TGF-β1 was not observed in granulosa cells or oocytes of any follicles (Figure 1a,1b, table 2). However, TGF-β1 mRNA was observed in stromal and/or thecal cells of type 3 follicles, in the theca interna of type 4 and 5 follicles and also in the stroma and cells of the vascular system. Within the theca interna, the cells closest to the basement membrane usually had more intense signal than those further away (Figure 1a,1b).
Figure 1 Localization of expression of mRNA encoding TGF-βs in ovine ovaries. Panels a and b contain corresponding light field and dark field views of a type 5 follicle from an adult ewe hybridized to TGF-β1 antisense RNA. Silver grains indicating hybridization of the TGF-β1 antisense RNA are observed concentrated in thecal (t) cells close to the basement membrane with no specific hybridization observed in either the granulosa cells (gc) or oocyte (o). The inset in panel b contains a dark field view of the same area of the tissue hybridized to the TGF-β1 sense RNA. Note the lack of specific concentration of silver grains over any cellular type. Panels c and d contain several type 5 (5) follicles and a type 4 (4) follicle in a 4 week old lamb hybridized to TGF-β2 antisense RNA. Note the lack of hybridization in the oocytes and granulosa cells of the type 4 and 5 follicles and the concentration of silver grains in thecal cells around the follicles as well as the stromal cells between the follicles and scattered cells of the surface epithelium (se). Observe the equal distribution of silver grains over the thecal cells. The inset in panel d contains a dark field view of the same area of the tissue hybridized to the TGF-β2 sense RNA. Note the lack of specific concentration of silver grains over any cellular type. Panels e and f contain corresponding light field and dark field views of an ovarian section obtained from an adult ewe hybridized to TGF-β3 antisense RNA. There is a lack of hybridization in the section including the granulosa and thecal cells of the types 4 (4) and 5 follicle (5) as well as the oocyte of the type 4 follicle, stroma tissue and the surface epithelium (SE). Panels g and h contain light field and dark field views of a blood vessel (v) from an adult ewe hybridized to TGF-β3 antisense RNA. Observe the specific hybridization in the wall of the vessel (v ). Panel i contains a dark field view of the same area of the tissue hybridized to TGF-β3 sense RNA. Note the lack of specific concentration of silver grains over any cellular type. Scale bar equals approximately 100 μm for all panels.
Table 2 Summary of expression of mRNA encoding TGF-βs and receptors in ovine ovary.
Gene Follicular type Stroma Vascular System
1/1a 2 3 4 5
TGFβ1 - - t t t + +
TGFβ2 - - t t t + +
TGFβ3 - - - - - - +
TGFβRI o o o, gc, t o, gc, t o, gc, t + +
TGFβRII - - t t t + +
+, expression observed; -, expression not observed; o, oocyte; gc, granulosa; t, theca
TGF-β2
The pattern of expression of mRNA encoding TGF-β2 was similar to that observed for TGF-β1, with hybridization limited to the thecal cells of type 3 and larger follicles (Figure 1c,1d, table 2). However, hybridization within the thecal layer appeared evenly distributed in contrast to the signal for TGF-β1 (compare panels a, b and c, d in Figure 1). Expression of TGF-β2 mRNA was also observed in some surface epithelium and stromal cells as well as cells associated with the vascular system.
TGF-β3
Expression of TGF-β3 mRNA was exclusive to cells associated with the vascular system of the ovary. Expression was not observed in the granulosa, theca, or oocyte of any follicle examined (Figure 1e,1f,1g,1h,1i, table 2).
TGFβRI
Expression of TGFβRI mRNA was observed in oocytes of all types of follicles (Figure 2a,2b,2c,2d, table 2). Granulosa and thecal cells of type 3 and larger follicles also expressed TGFβRI mRNA (Figure 2c,2d). Signal was also observed in the surface epithelium, stromal cells (Figure 2a,2b,2c,2d and luteal tissue (data not shown).
Figure 2 Localization of expression of mRNA encoding TGF-β receptors in ovine ovaries. Panels a and b contain corresponding light field and dark field views of several small follicles from a 4 week old lamb following hybridization to the TGFβRI antisense RNA. Note specific hybridization in the oocytes of types 1/1a follicles (1) and type 2 follicles. Observe that some cells of the surface epithelium also express TGFβRI. Panels c and d contain corresponding light field and dark field views of a type 5 follicle from a 4 week old lamb following hybridization to the TGFβRI antisense RNA. Note the hybridization signal in the granulosa (gc), theca (t) and oocyte (o) of the type 5 follicle. Signal was also observed in many stromal cells. The inset in panel d contains a dark field view of the same area of the tissue hybridized to TGFβRI sense RNA. Observe the lack of specific concentration of silver grains over any cellular type. Panels e and f contain corresponding light field and dark field views of several small follicles from a 4 week old lamb following hybridization to the TGFβRII antisense RNA. Note the lack of specific hybridization in the type 1/1a and 2 follicles. Expression was observed in the theca of type 4 and 5 follicles however, note the lack of expression in the granulosa cells and oocytes of these follicles. Note also that some cells of the surface epithelium also express TGFβRII. The insert in panel f contains a dark field view of the same area of the tissue hybridized to TGFβRII sense RNA. Note the lack of specific concentration of silver grains over any cellular type. Panels g and h contain corresponding light field and dark field views of a type 5 follicle as well as several type 1/1a follicles from a 4 week old lamb ovary hybridized to the TGFβRII antisense RNA. Note that hybridization is limited to the theca (t) of the type 5 follicle and several stromal cells and is not observed in the granulosa cells (gc) or oocyte (o) of the type 5 follicle. In addition, signal is observed in the stroma around the type 1/1a follicles (1) but is not observed in the type 1/1a follicles. Scale bar equals approximately 100 μm for all panels.
TGFβRII
Expression of TGFβRII mRNA was not observed in types 1,1a or 2 follicles (Figure 2e,2f,2g,2h, table 2). Also, in larger follicles TGFβRII mRNA was not detected in granulosa cells or oocytes (Figure 2e,2f,2g,2h). In type 3 and larger follicles, expression of TGFβRII was localized to the theca interna (Figure 2e,2f,2g,2h, table 2). As was observed with TGF-β1, expression of TGFβRII within the theca was most intense in the cells adjacent to the basement membrane (Figure 2e,2f,2g,2h). Signal was also observed in some cells of the surface epithelium (Figure 2e,2f), and in stroma (Figure 2e,2f,2g,2h) and luteal tissue (data not shown).
Effects of TGF-βs on granulosa cell function in vitro and expression of TGFBRII in cultured cells
Both TGF-β1 and TGF-β2 inhibited progesterone synthesis of cultured granulosa cells, whether expressed as a function of number of cells placed in culture (Figure 3, top panel) or as a function of DNA content at the end of culture (Figure 3, bottom panel) with significant affects observed with as little as 0.1 ng/ml of either TGF-β. Treatment with either TGF-β also reduced DNA content at the termination of culture (Figure 3, middle panel). For both variables, no differences were observed between the effect of TGF-β1 and TGF-β2 at any dose of growth factor tested. In contrast to the lack of detectable expression of the TGFβRII mRNA observed in situ, freshly isolated or cultured granulosa cells expressed mRNA for the TGFβRII when assessed by RT-PCR (Figure 4).
Figure 3 Effects of TGFβs on granulosa cell function. Effects of TGF-β1 and TGF-β2 on secretion of progesterone during the last 48 hours of culture (top), content of DNA at the termination of culture (middle) and progesterone concentration per μg of DNA. Values are expressed as the LS mean from 3 separate experiments. The dose of either TGF-β1 or TGF-β2 is indicated along the bottom of the graphs. For each variable, asterisk(s) indicates values that are different from the control (0) value (* p < 0.05; ** p < 0.01, *** p < 0.001). Comparisons were also made between the values obtained for TGF-β1 and TGF-β2 at each dose; however, no significant differences were observed at any dose tested.
Figure 4 Expression of TGFβRII in cultured granulosa cells. Determination of expression of TGFβRII in granulosa cells immediately following collection and following 48 hours of culture. Lanes 1–3 contain PCR products (766 bases) following amplification with ovine TGFβRII primers from 3 separate pools of granulosa cells at the time of collection, lanes 4–6 contain PCR products (766 bases) following amplification with ovine TGFβRII primers from 3 separate pools of granulosa cells collected 48 hours after culture, lane 7 contains the negative control water blank whereas lane 8 contains the PCR product from the positive control 4 week old ovary sample. Migration of DNA molecular weight standards are indicated on the left hand side.
Discussion
In the ewe, expression of TGF-β1 and TGF-β2 mRNA in the follicle was limited to thecal cells during all stages of follicular growth examined. Furthermore, expression of TGF-β3 mRNA was not observed in any follicular cells. This is in contrast to the observed expression patterns for these proteins in other species where TGF-β1 and TGF-β2 have been localized to granulosa as well as thecal cells and sometimes also to the oocytes of many species [8,12,13,31-33]. Also in contrast to sheep, expression of TGF-β3 in cattle and cats was observed in the oocyte, theca and granulosa of follicles at various stages of development [12,13]. In pigs, the theca interna has been proposed to be the major source of TGF-β since granulosa cells express TGF-β1 mRNA without seeming to make the protein [11]. Similarly, expression of TGF-β2 mRNA has been observed in bovine oocytes, but no detectable TGF-β activity was observed [17], although other studies have demonstrated TGF-β protein in the oocytes using immunocytochemistry [12]. In addition, granulosa cells isolated from pigs and cattle produce little if any TGFβ bioactivity when cultured in vitro [9,11]. Thus, it seems likely that in some species, follicular TGF-β activity originates primarily from the thecal cells, with control of activity possibly occurring at several levels including gene transcription (this study), protein translation [11] or activation of the protein [17].
Similar to what has been observed in other species [32,34,35], expression of TGFβRI mRNA was observed in several different cell-types of the sheep ovary including the oocyte, granulosa cells, thecal cells, stroma, luteal cells and surface epithelium. While expression of TGFβRII mRNA was also observed in stroma, luteal cells and the surface epithelium, its expression within the follicle was limited to the theca. A similar pattern of expression for the TGFβRII mRNA was observed in mouse follicles, with expression most prominent in the theca and barely detectable in granulosa cells [8]. However, using immunocytochemistry, strong staining for TGFβRII has been observed in granulosa cells with no to little staining in oocytes and in the theca in other species [13,32,35-37]. The reasons for these observed differences in localization of the TGFβRII are uncertain but may be due to differences in techniques or species differences.
Expression of mRNA encoding all three TGF-β isoforms and the TGF-β type I and II receptors were observed in cells associated with blood vessels and both receptor types and TGF-β1 and 2 mRNAs were observed in the stroma surrounding follicles indicating a potential role for TGF-β in regulating certain functions in the ovarian stroma and vascular network. TGF-β is known to be important in regulating angiogenesis [38,39]. Moreover, in the ovary, both TGF-β1 and TGF-β3 mRNAs are upregulated during revascularization following autotransplantation of rat ovaries [40] further supporting a role for these factors in regulating vascular function.
The much more restricted pattern of expression of TGFβRII mRNA in sheep indicates that the TGFβRI may well be involved with other type II receptors in the signalling of other members of the transforming growth factor family. In agreement with this, TGFβRI has recently been shown to be involved in signalling of the oocyte-derived GDF-9 along with BMPRII [41,42]. In other species, GDF-9 has been shown to regulate granulosa cell mitosis and differentiation [6] and has been shown to be essential for normal follicular growth and development in both mice [43] and sheep [44,45]. Thus, expression of TGFβRI mRNA as well as BMPRII [46,47] in granulosa cells is probably mediating the effects of GDF-9. Localization of both of these receptors in granulosa cells from the type 3 (secondary) stage of development onwards is consistent with the presence of normal primary but not secondary follicles in both sheep [44] and mice [43] lacking biologically active GDF-9. Interestingly, in sheep, TGFβRI mRNA and BMPRII [46,47] are also both localized in oocytes from the type 1 (primordial) stage onwards suggesting that GDF-9 may also regulate oocyte function in this species.
The suppression of progesterone production and DNA content in granulosa cell cultures by TGF-β1 or TGF-β2 is similar to inhibitory to mild stimulatory effects observed in bovine, ovine and porcine granulosa cell cultures and contrary to the strong stimulatory effects observed in rodents [11,14-18,20-26]. The decreased DNA content observed following treatment accounts for some, but not all, of the decrease observed in progesterone concentration in the granulosa cell cultures. The suppression of progesterone synthesis indicates an anti-differentiative role for this growth factor as has been observed for other members of the TGF-β superfamily. The decreased content of DNA observed following culture could be related to a suppression of granulosa cell proliferation or survival. Since TGF-β can stimulate apoptotic pathways in concert with other factors [48,49], a role for TGF-β in regulating apoptosis of ovarian cells has been proposed.
No differences in the efficacy of TGF-β1 and TGF-β2 were observed in ovine granulosa cells. Similarly, TGF-β1 and TGF-β2 were equally efficacious in stimulating inhibin production in luteinized human granulosa cells [50] and in modulating gonadotrophin receptor expression in both rat and porcine granulosa cells [51]. Interestingly, while both TGF-β1 and TGF-β2 mRNA were synthesized by the theca interna, their spatial patterning within the theca was quite different. TGF-β1 mRNA was concentrated in the thecal cells closest to the basement membrane, similar to the localization observed for the TGFβRII mRNA. In contrast, TGF-β2 mRNA expression was observed throughout the thecal layer. The role, if any, of the apparent differential regulation of these two isoforms in subtypes of thecal cells is currently unknown.
Given the potent effects of both TGF-β1 and TGF-β2 on granulosa cell function in vitro, the lack of detectable expression of TGFβRII mRNA in these cells using in situ hybridization was very surprising. There are several potential explanations for these apparent conflicting results. It is possible that TGFβRII is expressed in ovine granulosa cells and the technique utilized simply failed to detect this message. The detection of mRNA encoding TGFβRII in isolated granulosa cells both before and after culture using RT-PCR would seem to support this assumption. However, it is possible that the isolation and culture of the granulosa cells potentially could be inducing expression of TGFβRII as most all cells in culture express TGFβRII [52]. In addition, strong expression of TGFβRII mRNA in luteal tissue is also consistent with up regulation of the TGFβRII in these cells as induction of progesterone synthesis by the ovine granulosa cells can be considered to indicate at least a partial luteinization of these cells. Finally, it is also possible that TGFβs are using another member of the type II receptor family to mediate their effects. The existence of a second type II receptor capable of mediating TGF-β effects is supported by the inability of cell lines expressing TGFβRII to bind to TGFβ2 but not TGFβ1 [53] and cell lines responsive to TGF-β without a detectible type II TGFβR [52].
Conclusions
Expression of mRNAs encoding TGF-β1 and TGF-β2 as well as both type I and II TGF-β receptors were observed in the theca of small growing follicles indicating that TGF-βs may be regulating thecal cell function in an autocrine manner. Expression of mRNA encoding TGF-β type I and II receptors is also observed in luteal cells, stroma, the vascular system and surface epithelium suggesting that TGF-βs may also regulate other cell types in the sheep ovary. Since granulosa cells showed no evidence of expressing any of the TGF-β ligands and expression of the TGF-β type II receptor was equivocal, it seems likely that any TGF-β effects in granulosa cells in vivo are due to paracrine or endocrine actions and possibly regulated through an alternative type II receptor.
Authors' contributions
AHB, LDQ and LJH cloned the ovine TGF-βs and receptors, completed sequencing projects and alignments, and performed the in situ hybridizations and PCRs. SL and KLR performed the granulosa cell bioassays including progesterone and DNA measurements. JLJ and KPM designed and co-ordinated the experiments, performed statistical analysis and drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
Supported by New Zealand Foundation for Research, Science and Technology, the Royal Society of New Zealand Marsden Fund and Ovita Limited, Dunedin, New Zealand. The authors would also like to thank Doug Jensen for the help with animal care, Lee-Ann Still, Peter Smith and Norma Hudson for technical assistance and Alan Barkus for help with preparation of the figures.
==== Refs
Knight PG Glister C Local roles of TGF-beta superfamily members in the control of ovarian follicle development Anim Reprod Sci 2003 78 165 183 12818643 10.1016/S0378-4320(03)00089-7
Chang H Brown CW Matzuk MM Genetic analysis of the mammalian transforming growth factor-beta superfamily Endocr Rev 2002 23 787 823 12466190 10.1210/er.2002-0003
Welt C Sidis Y Keutmann H Schneyer A Activins, inhibins, and follistatins: from endocrinology to signaling. A paradigm for the new millennium Exp Biol Med (Maywood) 2002 227 724 752 12324653
Durlinger AL Visser JA Themmen AP Regulation of ovarian function: the role of anti-Mullerian hormone Reproduction 2002 124 601 609 12416998 10.1530/rep.0.1240601
Lin SY Morrison JR Phillips DJ de Kretser DM Regulation of ovarian function by the TGF-beta superfamily and follistatin Reproduction 2003 126 133 148 12887271 10.1530/rep.0.1260133
Shimasaki S Moore RK Otsuka F Erickson GF The bone morphogenetic protein system in mammalian reproduction Endocr Rev 2004 25 72 101 14769828 10.1210/er.2003-0007
Chegini N Williams RS Immunocytochemical localization of transforming growth factors (TGFs) TGF-alpha and TGF-beta in human ovarian tissues J Clin Endocrinol Metab 1992 74 973 980 1569174 10.1210/jc.74.5.973
Schmid P Cox D van der Putten H McMaster GK Bilbe G Expression of TGF-beta s and TGF-beta type II receptor mRNAs in mouse folliculogenesis: stored maternal TGF-beta 2 message in oocytes Biochem Biophys Res Commun 1994 201 649 656 8002998 10.1006/bbrc.1994.1750
Mulheron GW Mulheron JG Danielpour D Schomberg DW Porcine granulosa cells do not express transforming growth factor-beta 2 (TGF-beta 2) messenger ribonucleic acid: molecular basis for their inability to produce TGF-beta activity comparable to that of rat granulosa cells Endocrinology 1992 131 2609 2614 1280203 10.1210/en.131.6.2609
Roy SK Hughes J Ontogeny of granulosa cells in the ovary: lineage-specific expression of transforming growth factor beta 2 and transforming growth factor beta 1 Biol Reprod 1994 51 821 830 7849185
May JV Stephenson LA Turzcynski CJ Fong HW Mau YH Davis JS Transforming growth factor beta expression in the porcine ovary: evidence that theca cells are the major secretory source during antral follicle development Biol Reprod 1996 54 485 496 8788203
Nilsson EE Doraiswamy V Skinner MK Transforming growth factor-beta isoform expression during bovine ovarian antral follicle development Mol Reprod Dev 2003 66 237 246 14502602 10.1002/mrd.10350
Bristol SK Woodruff TK Follicle-restricted compartmentalization of transforming growth factor beta superfamily ligands in the feline ovary Biol Reprod 2004 70 846 859 14656728 10.1095/biolreprod.103.021857
Dorrington J Chuma AV Bendell JJ Transforming growth factor beta and follicle-stimulating hormone promote rat granulosa cell proliferation Endocrinology 1988 123 353 359 3133196
Roy SK Epidermal growth factor and transforming growth factor-beta modulation of follicle-stimulating hormone-induced deoxyribonucleic acid synthesis in hamster preantral and early antral follicles Biol Reprod 1993 48 552 557 8452931
Saragueta PE Lanuza GM Baranao JL Autocrine role of transforming growth factor beta1 on rat granulosa cell proliferation Biol Reprod 2002 66 1862 1868 12021073
Gilchrist RB Morrissey MP Ritter LJ Armstrong DT Comparison of oocyte factors and transforming growth factor-beta in the regulation of DNA synthesis in bovine granulosa cells Mol Cell Endocrinol 2003 201 87 95 12706297 10.1016/S0303-7207(02)00429-X
Gangrade BK May JV The production of transforming growth factor-beta in the porcine ovary and its secretion in vitro Endocrinology 1990 127 2372 2380 2226322
May JV Frost JP Schomberg DW Differential effects of epidermal growth factor, somatomedin-C/insulin-like growth factor I, and transforming growth factor-beta on porcine granulosa cell deoxyribonucleic acid synthesis and cell proliferation Endocrinology 1988 123 168 179 3260173
Skinner MK Keski-Oja J Osteen KG Moses HL Ovarian thecal cells produce transforming growth factor-beta which can regulate granulosa cell growth Endocrinology 1987 121 786 792 3297652
Dodson WC Schomberg DW The effect of transforming growth factor-beta on follicle-stimulating hormone-induced differentiation of cultured rat granulosa cells Endocrinology 1987 120 512 516 3026778
Knecht M Feng P Catt K Bifunctional role of transforming growth factor-beta during granulosa cell development Endocrinology 1987 120 1243 1249 3030692
Hutchinson LA Findlay JK de Vos FL Robertson DM Effects of bovine inhibin, transforming growth factor-beta and bovine Activin-A on granulosa cell differentiation Biochem Biophys Res Commun 1987 146 1405 1412 3113435 10.1016/0006-291X(87)90806-0
Mondschein JS Canning SF Hammond JM Effects of transforming growth factor-beta on the production of immunoreactive insulin-like growth factor I and progesterone and on [3H]thymidine incorporation in porcine granulosa cell cultures Endocrinology 1988 123 1970 1976 3262054
Kubota T Kamada S Taguchi M Aso T Autocrine/paracrine function of transforming growth factor-beta 1 in porcine granulosa cells Hum Reprod 1994 9 2118 2122 7868683
Fabre S Pierre A Pisselet C Mulsant P Lecerf F Pohl J Monget P Monniaux D The Booroola mutation in sheep is associated with an alteration of the bone morphogenetic protein receptor-IB functionality J Endocrinol 2003 177 435 444 12773124 10.1677/joe.0.1770435
McNatty KP Juengel JL Wilson T Galloway SM Davis GH Hudson NL Moeller CL Cranfield M Reader KL Laitinen MP Groome NP Sawyer HR Ritvos O Oocyte-derived growth factors and ovulation rate in sheep Reprod Suppl 2003 61 339 351 14635946
Tisdall DJ Fidler AE Smith P Quirke LD Stent VC Heath DA McNatty KP Stem cell factor and c-kit gene expression and protein localization in the sheep ovary during fetal development J Reprod Fertil 1999 116 277 291 10615253
Lundy T Smith P O'Connell A Hudson NL McNatty KP Populations of granulosa cells in small follicles of the sheep ovary J Reprod Fertil 1999 115 251 262 10434930
Lun S Smith P Lundy T O'Connell A Hudson N McNatty KP Steroid contents of and steroidogenesis in vitro by the developing gonad and mesonephros around sexual differentiation in fetal sheep J Reprod Fertil 1998 114 131 139 9875165
Chegini N Flanders KC Presence of transforming growth factor-beta and their selective cellular localization in human ovarian tissue of various reproductive stages Endocrinology 1992 130 1707 1715 1537318 10.1210/en.130.3.1707
Juneja SC Chegini N Williams RS Ksander GA Ovarian intrabursal administration of transforming growth factor beta 1 inhibits follicle rupture in gonadotropin-primed mice Biol Reprod 1996 55 1444 1451 8949905
Teerds KJ Dorrington JH Immunohistochemical localization of transforming growth factor-beta 1 and -beta 2 during follicular development in the adult rat ovary Mol Cell Endocrinol 1992 84 R7 13 1639213 10.1016/0303-7207(92)90085-K
Qu J Godin PA Nisolle M Donnez J Expression of receptors for insulin-like growth factor-I and transforming growth factor-beta in human follicles Mol Hum Reprod 2000 6 137 145 10655455 10.1093/molehr/6.2.137
Roy SK Kole AR Ovarian transforming growth factor-beta (TGF-beta) receptors: in-vitro effects of follicle stimulating hormone, epidermal growth factor and TGF-beta on receptor expression in human preantral follicles Mol Hum Reprod 1998 4 207 214 9570266 10.1093/molehr/4.3.207
Roy SK Kole AR Transforming growth factor-beta receptor type II expression in the hamster ovary: cellular site(s), biochemical properties, and hormonal regulation Endocrinology 1995 136 4610 4620 7664681 10.1210/en.136.10.4610
Wehrenberg U Giebel J Rune GM Possible involvement of transforming growth factor-beta 1 and transforming growth factor-beta receptor type II during luteinization in the marmoset ovary Tissue Cell 1998 30 360 367 10091338
Distler JH Hirth A Kurowska-Stolarska M Gay RE Gay S Distler O Angiogenic and angiostatic factors in the molecular control of angiogenesis Q J Nucl Med 2003 47 149 161 12897707
Marchuk DA Srinivasan S Squire TL Zawistowski JS Vascular morphogenesis: tales of two syndromes Hum Mol Genet 2003 12 Spec No 1 R97 112 12668602 10.1093/hmg/ddg103
Dissen GA Lara HE Fahrenbach WH Costa ME Ojeda SR Immature rat ovaries become revascularized rapidly after autotransplantation and show a gonadotropin-dependent increase in angiogenic factor gene expression Endocrinology 1994 134 1146 1154 8119153 10.1210/en.134.3.1146
Mazerbourg S Klein C Roh J Kaivo-Oja N Mottershead DG Korchynskyi O Ritvos O Hsueh AJ Growth differentiation factor-9 signaling is mediated by the type I receptor, activin receptor-like kinase 5 Mol Endocrinol 2004 18 653 665 14684852 10.1210/me.2003-0393
Vitt UA Mazerbourg S Klein C Hsueh AJ Bone morphogenetic protein receptor type II is a receptor for growth differentiation factor-9 Biol Reprod 2002 67 473 480 12135884
Dong J Albertini DF Nishimori K Kumar TR Lu N Matzuk MM Growth differentiation factor-9 is required during early ovarian folliculogenesis [see comments] Nature 1996 383 531 535 8849725 10.1038/383531a0
Juengel JL Hudson NL Heath DA Smith P Reader KL Lawrence SB O'Connell AR Laitinen MP Cranfield M Groome NP Ritvos O McNatty KP Growth differentiation factor 9 and bone morphogenetic protein 15 are essential for ovarian follicular development in sheep Biol Reprod 2002 67 1777 1789 12444053 10.1095/biolreprod.102.007146
Hanrahan JP Gregan SM Mulsant P Mullen M Davis GH Powell R Galloway SM Mutations in the genes for oocyte-derived growth factors GDF9 and BMP15 are associated with both increased ovulation rate and sterility in Cambridge and Belclare sheep (Ovis aries) Biol Reprod 2004 70 900 909 14627550 10.1095/biolreprod.103.023093
Wilson T Wu XY Juengel JL Ross IK Lumsden JM Lord EA Dodds KG Walling GA McEwan JC O'Connell AR McNatty KP Montgomery GW Highly prolific Booroola sheep have a mutation in the intracellular kinase domain of bone morphogenetic protein IB receptor (ALK-6) that is expressed in both oocytes and granulosa cells Biol Reprod 2001 64 1225 1235 11259271
Souza CJ Campbell BK McNeilly AS Baird DT Effect of bone morphogenetic protein 2 (BMP2) on oestradiol and inhibin A production by sheep granulosa cells, and localization of BMP receptors in the ovary by immunohistochemistry Reproduction 2002 123 363 369 11882013 10.1530/rep.0.1230363
Foghi A Teerds KJ van der Donk H Moore NC Dorrington J Induction of apoptosis in thecal/interstitial cells: action of transforming growth factor (TGF) alpha plus TGF beta on bcl-2 and interleukin-1 beta-converting enzyme J Endocrinol 1998 157 489 494 9691982 10.1677/joe.0.1570489
Foghi A Teerds KJ van der Donk H Dorrington J Induction of apoptosis in rat thecal/interstitial cells by transforming growth factor alpha plus transforming growth factor beta in vitro J Endocrinol 1997 153 169 178 9135581 10.1677/joe.0.1530169
Eramaa M Ritvos O Transforming growth factor-beta 1 and -beta 2 induce inhibin and activin beta B-subunit messenger ribonucleic acid levels in cultured human granulosa-luteal cells Fertil Steril 1996 65 954 960 8612856
Gitay-Goren H Kim IC Miggans ST Schomberg DW Transforming growth factor beta modulates gonadotropin receptor expression in porcine and rat granulosa cells differently Biol Reprod 1993 48 1284 1289 8318582
Massague J The transforming growth factor-beta family Annu Rev Cell Biol 1990 6 597 641 2177343
Lawler S Candia AF Ebner R Shum L Lopez AR Moses HL Wright CV Derynck R The murine type II TGF-beta receptor has a coincident embryonic expression and binding preference for TGF-beta 1 Development 1994 120 165 175 8119124
| 15563738 | PMC539244 | CC BY | 2021-01-04 16:36:43 | no | Reprod Biol Endocrinol. 2004 Nov 25; 2:78 | utf-8 | Reprod Biol Endocrinol | 2,004 | 10.1186/1477-7827-2-78 | oa_comm |
==== Front
Int J Health GeogrInternational Journal of Health Geographics1476-072XBioMed Central London 1476-072X-3-281557419710.1186/1476-072X-3-28ReviewCurrent practices in spatial analysis of cancer data: data characteristics and data sources for geographic studies of cancer Boscoe Francis P [email protected] Mary H [email protected] Peggy [email protected] New York State Cancer Registry, New York State Department of Health, Albany, NY, USA2 Occupational and Environmental Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, NIH, DHHS, Bethesda, MD, USA3 California Department of Health Services, Environmental Health Investigations Branch, Oakland, CA, USA2004 1 12 2004 3 28 28 29 9 2004 1 12 2004 Copyright © 2004 Boscoe et al; licensee BioMed Central Ltd.2004Boscoe et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The use of spatially referenced data in cancer studies is gaining in prominence, fueled by the development and availability of spatial analytic tools and the broadening recognition of the linkages between geography and health. We provide an overview of some of the unique characteristics of spatial data, followed by an account of the major types and sources of data used in the spatial analysis of cancer, including data from cancer registries, population data, health surveys, environmental data, and remote sensing data. We cite numerous examples of recent studies that have used these data, with a focus on etiological research.
==== Body
Introduction
Understanding the spatial patterns of diseases in a population can provide insight as to their causes and controls. Indeed, this notion is at the very root of the field of epidemiology [1]. The recent explosion in data gathering, linkage and analysis capabilities fostered by computing technology, particularly geographic information systems (GIS), has greatly improved the ability to measure and assess these patterns. Large and complex georeferenced data sets are now readily available through Spatial Data Clearinghouses, facilitating analyses by researchers unaffiliated with the government agencies that have historically controlled data access. Meanwhile, increasingly sophisticated statistical tools have evolved to keep pace with the increased data availability and computing power.
The purpose of this article is to provide an overview of spatial data and its relevance to population-based cancer surveillance and research in the United States as of 2004. We begin by discussing a number of the distinctive characteristics of spatial data, which can sometimes hinder efforts to understand cancer etiology. We then proceed to describe the kinds of data sets that are available, accompanied by a survey of some applications using these data. Finally, we discuss several ongoing efforts to provide central repositories of geospatial data. Given the vast scope of cancer research taking place worldwide, our survey is necessarily partial, and we have chosen to emphasize etiology over other research themes with spatial dimensions, such as patterns of treatment or access to care [2].
Qualities of spatial data
Spatial data refer to data with locational attributes. Most commonly, locations are given in Cartesian coordinates referenced to the earth's surface. These coordinates may describe points, lines, areas or volumes. This need not be the only spatial framework; "relative spaces" may be defined in which distance is defined in terms of some other attribute, such as sociodemographic similarly or connectedness along transportation networks [3,4]. Spatial data have special qualities that require specialized statistical techniques and modeling approaches. A complete discussion of these special qualities is well beyond the scope of this article, but here we describe a number of the more compelling and recurring themes. For a focused discussion on the limitations on analysis that these data characteristics impose, see the companion piece to this article, "Current Practices in Spatial Analysis of Cancer Data: Flies in the Ointment, Or, The Limitations of Spatial Analysis" [5].
Individual humans represent the basic unit of spatial analysis in cancer research. Individuals are categorized as either having or not having a disease or attribute of a disease, and are assigned coordinates corresponding to the location of their place of residence, a technique known as geocoding. As with all measurements, geocoding involves some error. A growing body of literature is exploring the nature of this error and its potential to bias epidemiologic studies [6]. Among the topics that have been investigated are systematic problems with geographic reference files [7], the ramifications of different geocoding algorithms [8], positional accuracy [9] and how to handle non-residential addresses, such as rented post office boxes [10].
Assigning individuals to their place of residence also poses problems, although this is usually the only locational information that is available. Often the goal of a geographic analysis is to identify a common environmental exposure in a population, but exposures that are occupational or recreational may not necessarily reveal themselves in a residential analysis. Also, given the long latency period for many cancers and the mobility of the American population, the relevant exposure may be associated with a prior address. Difficult-to-measure behavioral risk factors such as smoking and diet often further confound attempts at geographic analysis.
Owing to confidentiality restrictions, researchers outside of central cancer registries typically do not have access to address-level data. In such instances, case data are aggregated by some functional or political unit such as census tract, county or ZIP code. Even when the case data are geocoded, population data must be aggregated, at least to the level of the census block, which is the smallest unit for which any population information is available. Knowing that there are four women with breast cancer living on the same street is not sufficient, by itself, to draw conclusions about whether the street displays an unusual incidence pattern; one must also know the number and ages of women without breast cancer on the same street. Short of conducting one's own thorough door-to-door census, this question cannot be answered, except by aggregating the street segments into blocks. When additional variables, such as measures of income or education, or also of interest, then still larger analytical units must be chosen.
The necessity for aggregating spatial data raises a whole set of analytic issues regarding the extent to which the act of aggregating introduces error and bias. It is theoretically possible to achieve dramatically different, even contradictory results, simply as a consequence of aggregating the data in a different fashion [11]. This is true not only for aggregations at different spatial scales, but also different aggregations at the same scale. Geographers have termed this the "modifiable areal unit problem" [12]. A special case of the modifiable areal unit problem is the ecological inference problem, which specifically refers to the lack of congruity between associations found in aggregated and individual-level data.
In practice, well-chosen scales and groupings can minimize the modifiable areal unit problem and allow reasonable consistency between aggregated and individual-level results [13]. There will always be exceptions to this, however, as evidenced by the many studies attempting to relate low-level indoor radon concentrations with lung cancer incidence. Individual-level studies have repeatedly found a positive correlation, while area-level studies have found a negative correlation at low radon levels [14-16]. Despite a general appreciation of how these discrepant results represent an example of aggregation bias, there is still active debate over what these results say about low-level radon risk [17,18].
Often analyses need to be performed on data that was collected at different spatial scales, such as a study using cancer cases aggregated by ZIP code and modeled air pollutant data at the census tract level. The resulting scale-translation problem is a recurring one that has inspired many independent solutions, and is known variously as areal interpolation, the polygon overlay problem, and the problem of inference with spatially misaligned data, among other terms [19]. The most naïve solution to this problem is to assume that each measured value is homogeneous within each spatial unit. Under this assumption, using our example, a ZIP code that is coincident with four census tracts would be broken into four polygons, each having the same cancer rate but different air pollutant values. More sophisticated cartographic overlay techniques have been developed that involve using covariate information to infer variation within spatial units. To date, these techniques have been primarily applied toward estimating population surfaces rather than cancer or other disease rate surfaces [20,21]. Hierarchical Bayesian and multi-level logit models have also shown promise [22-25].
Spatial autocorrelation is another distinctive quality of spatial data that requires the use of specialized analytic methods. Spatial autocorrelation is the tendency for nearby observations to have correlated attribute values. For most data sets involving the distribution of human populations and their characteristics, spatial autocorrelation is positive, meaning that neighboring individuals tend to have similar characteristics. Understanding the characteristics and qualities of spatial autocorrelation is essential to adequately model and interpret geographic patterns. For example, it is not appropriate to perform ordinary least squares regression on spatial data, because the presence of spatial autocorrelation means that the observations are not independent. Performing such a regression generally results in downwardly biased estimations of variance, which yields overstated levels of significance. In general, spatially autocorrelated data is less informative in a model than uncorrelated data. There is an ample literature on assessing and properly accounting for spatial autocorrelation in geographic analysis [26,27].
A final critically important characteristic of spatial data is spatial nonstationarity, or the tendency for relationships between and among variables to vary by geographic location [28]. First-order or strong stationarity refers to the degree to which measured values vary spatially, while second-order or weak stationarity refers to the degree to which the uncertainties in these measured values vary spatially. So-called global statistics ignore nonstationarity, suggesting that relationships across space are constant. The simple linear equation that has traditionally been used to express the relationship between rainfall and altitude is a well-known example. Local statistics, in contrast, take nonstationarity into account, at least first-order nonstationarity. Brunsdon et al. [29] used the technique of geographically weighted regression to demonstrate that both the slope and intercept of the rainfall-altitude equation vary considerably in space. The range and breadth of local statistics has seen rapid growth in recent years [27,30].
Local statistics are less adept at accounting for second-order nonstationarity. Indeed, many of these methods require the assumption of constant variance across space. Because of the uneven distribution of human populations, this assumption is seldom met for health data. Specifically, disease rates in areas with smaller numbers of cases are more variable than those in areas with larger numbers of cases, a property that has also been termed "variance instability" [31]. Variance instability is particularly pervasive on maps, since it is extremely difficult to design a map that is not visually biased toward either sparsely populated or densely populated areas [32,33]. A simple example is the tendency for rural counties to contain disproportionate numbers of unusually high or unusually low disease rates and thus visually dominate a choropleth map. The problem is compounded by the tendency of such counties to be large in size; for these reasons, maps of United States counties are often visually dominated by such states as Idaho, Nevada and Wyoming.
Efforts to include information about data uncertainty have shown promise, but have not seen widespread use [34]. One common way of addressing this problem is to produce smoothed maps, whereby the rate for a given area is influenced by the rates of neighboring areas. There are many algorithms available to accomplish this [35], ranging from conceptually straightforward spatial filters [36] to computationally-intensive Bayesian approaches [37,38]. Properly accounting for second-order spatial nonstationarity in maps and models remains an active research area.
Types and sources of data
In this section we described the primary types and sources of data most frequently used in the geographic analysis of cancer, along with examples of their application. These are summarized in Table 1.
Table 1 Sources of Cancer Registry Data
Dataset name Source Agency URL Geographic Resolution
SEER*Stat, Cancer Mortality Maps and Graphs, State Cancer Profiles National Cancer Institute County
Florida Cancer Data System University of Miami School of Medicine County
Cancer Incidence and Mortality Rates in Kentucky Kentucky Cancer Registry County
New York Cancer Incidence by ZIP code NYS Department of Health ZIP code
1. Cancer registries
A cancer registry is a data collection system that tracks cancer cases that have been diagnosed or treated in a specific institution or geographic area. Cancer registries typically collect information from medical records provided by hospitals, doctors, other care facilities, medical laboratories, and/or insurers. Data collected by cancer registries is stored under secure conditions so as to protect confidentiality.
Historically, observed geographic differences in cancer incidence have been of great interest in trying to understand more about factors which may influence risk of these diseases. Such differences have served as the basis for studies of migrant populations and acculturation differences in migrant groups. They have been possible because cancer is one of the few chronic diseases for which high quality population-based disease surveillance systems have been in place for many years in many countries of the world.
Cancer registry data has been widely applied toward the production of cancer atlases [39], studies analyzing the spatial distribution of particular cancer sites [40], and studies assessing spatial clustering [41]. Most recently, cancer studies have been undertaken which build on the combined resources of cancer registry data and increasingly available GIS tools. Because address at diagnosis is available for most registry cases it can be geocoded and integrated in a GIS with social and environmental attribute information available at a variety of geographic scales. Examples of such approaches include studies of childhood cancer which examine rate differences in areas of low versus intense agricultural pesticide use [42], heavy traffic patterns [43], or high air pollution [44]. Alternatively, cancer registry data can serve to identify population-based cases for studies using case-control or cohort designs, which can in turn be integrated into a GIS for area attribute data. Examples of this approach include case-control studies of childhood leukemia and traffic patterns [45-48]. and a studies of breast cancer incidence associated with residence in high pesticide use areas in a large case-control study [49,50]. and in a large cohort study [51].
For these types of studies, cancer registry data offer both a number of strengths and limitations. Primary strengths include the comprehensiveness of geographic coverage, detailed information on disease subgroups, and rich covariable information on demographic characteristics for each newly diagnosed case of cancer. Because registry data are abstracted from medical records and reflect information for a snapshot in time, primary limitations include the lack of historical information on various factors of potential interest including residential mobility and relevant personal behaviors. Cancer registries typically collect information on the residential address for individuals newly diagnosed with cancer at the time of that diagnosis. Since this is the locational information which serves as the basis for national and international statistics on area cancer rates, it is also useful for looking at area characteristics associated with rate differences, although inferences about etiologic associations are limited for these long latency diseases, and even more so for residentially mobile populations.
The Surveillance, Epidemiology and End Results (SEER) program of the National Cancer Institute (NCI) offers county-level incidence data for its member registries, which cover part or all of eight states, through its SEER*Stat software. Because it provides direct access to individual cancer records, users must first sign a data access agreement. County-level mortality data for the entire United States, collected and maintained by the National Center for Health Statistics (NCHS), is also accessible through SEER*Stat. These data include all causes of death, not just cancer deaths. Selected county-level cancer data may also be accessed through the NCI's Cancer Mortality Maps and Graphs and State Cancer Profiles web sites. The latter was launched in 2003 and contains a host of innovative statistical graphics. Many individual state registries also offer additional geographically referenced data. For example, the Florida Cancer Data System web site allows users to generate a variety of county- and facility-level tables and county-level maps on demand. The Kentucky Cancer Registry also offers a county-level mapping application. New York State offers a limited set of ZIP code level data for the four most common cancer types in the mid-1990s. Currently, county-level cancer incidence data is not available nationally.
2. Population data
The United States Census Bureau is the principal source of data on the entire population; most countries have comparable agencies. Since cancer rates are calculated by dividing the number of cases by the number of people at risk, census data is frequently referred to as "denominator data". Census data are readily available in electronic format through the Census Bureau web site, . Data are available in three basic formats. American FactFinder is a web-based application that allows users to drill down through geographic levels to find data tables of interest. It is most useful for data queries that are well-focused. Data may also be downloaded through an ftp server. This method obtains raw text files that require computer code to be written before the data can be easily accessed or manipulated. This method is most useful for users with large data needs who are in possession of some database programming skills. The third approach is to purchase DVDs from the Census Bureau's Customer Service center. The DVDs allow data output in many spreadsheet and database formats, facilitating the ability for users to process and analyze the data. There are also a large number of third-party vendors who offer similar services [52].
The four primary data files emanating from the 2000 census are named Summary File 1 through Summary File 4 (SF1–SF4). SF1 contains population counts by age, sex, race and ethnicity and basic housing characteristic information for the entire population, to the block level. SF2 contains similar information, detailed for ethnic subgroups, American Indian and Alaska Native tribes, and multiple-race individuals. These data are suppressed when the total number of individuals in a given geographic unit totals fewer than 100. SF3 contains detailed housing, demographic, and socioeconomic data to the census block group or census tract level, based on a long form that was sent to one in six households. Census block groups have an optimal population size of 1,500 and census tracts have an optimal population size of 4,000, though in practice populations vary widely. SF4 contains the same information as SF3 for detailed race and ethnic groups, with the same suppression rule as SF2. In addition to these four primary data files, the Census Bureau also provides digital cartographic boundary files for political entities in the country, as well as approximations of postal code boundaries known as ZIP code tabulation areas (ZCTAs).
The Census Bureau also conducts the American Community Survey (ACS), an ongoing survey designed to reach 3 million households each year nationwide. The goal of this survey is to allow the publication of detailed demographic and socioeconomic information more often than once a decade. Data for geographic units totaling more than 65,000 people will be released annually, while data for smaller geographic units will be based on either a three or five year moving average. It will replace the census long form, which will not be administered in 2010. There will undoubtedly be a challenging adjustment period as public health researchers begin to use ACS data.
At present, the level of information available for intercensal time points is quite limited, and derives from Census Bureau estimates at the state or county level. These estimates are used in the calculation of cancer rates by federal and state agencies, although some research has shown that they are not especially reliable, particularly county-level estimates for specific race groups [53]. Various private vendors publish intercensal estimates for areas smaller than counties, though it is impossible to verify their accuracy. Since many vendors use the Census Bureau estimates as controls (for example, vendor estimates of ZIP code populations in a county must add to the Census Bureau estimate for that county), vendor estimates necessarily suffer from the same limitations as the Census Bureau estimates. Finally, some state governments publish their own population estimates. Generally, these estimates are thought to represent improvements over the Census Bureau estimates because of higher levels of local knowledge and a broader use of data sources. We are unaware of any independent efforts to evaluate these claims, however. Examples include the population estimates and projections published by the California Department of Finance, and those by the Epidemiology Program of the Cancer Research Center of Hawaii. The latter population estimates were developed in response to a concern that the Native Hawaiian population was substantially undercounted in previous censuses, and are used by the NCI in calculating national cancer rates.
The 2000 census allowed respondents to select more than one race, although cancer data are only beginning to be collected in this manner. As a result, population data from 2000 must be "bridged" back to the earlier single-race categories to allow comparisons with earlier data. NCHS developed a sophisticated bridging algorithm taking into account age, sex, distribution of single-race groups within counties, and other covariates [54]. This algorithm is reflected in the 1991–2003 population projections and estimates that are published on the NCI web site and included in their statistical software. The Census Bureau itself uses a simpler algorithm in its estimates, allocating equal proportions of each multiple-race combination to the constituent single races [55]. Given the multiplicity of population estimates and methods for calculating them that are available, it is important to be aware of the sources of these data, and how they may influence the confidence associated with a particular research result. This is especially true for small-area analyses, where uncertainties are highest.
In addition to the issues noted above, it is important to realize that even the decennial census counts are not as accurate as popularly believed. The census represents an attempt to enumerate the population as of a single date, but invariably some people are missed or double-counted. These undercounts and overcounts are differential by race, socioeconomic status, and geographic area, potentially biasing cancer rates [56,57].
Countless epidemiologic and geographic studies make use of census data in some capacity, including most studies that report cancer rates for geographic areas. It is also quite common to use census data where individual-level data are not available, particularly for indicators of socioeconomic status [58-60], educational attainment [61] and housing characteristics [7]. Table 2 summarizes the population data sources described in this section.
Table 2 Sources of Population Data
Dataset name Source Agency URL Geographic Resolution
2000 Census Summary Files 1–4 US Census Bureau Census Tract, Block Group or Block (varies by data element)
American Community Survey US Census Bureau Areas with populations >65,000
E-1 City/County Population Estimates, with Annual Percent Change California Department of Finance City/County
US Population Data, 1969–2001 National Cancer Institute County
3. Surveys
In addition to the Census Bureau as a primary source of sociodemographic attribute data, special survey data can provide valuable information on these characteristics for population groups in some areas. Perhaps one of the best known such surveys is the CDC-sponsored Behavioral Risk Factor Surveillance System (BRFSS), which is touted as the "world's largest telephone survey". Designed in the 1980s to track trends in behavioral risk factors at the state level, this ongoing system of national surveys also provides subarea and subgroup information within some of the larger states. Some researchers have estimated county-level behavioral risk factor prevalence by combining the statewide BRFSS data with county-level demographic data [62,63]. A mapping application to view BRFSS response data at the state and metropolitan level is also available .
Another well-known national survey is the NCHS's National Health and Nutrition Examination Survey (NHANES), which has been in place since 1960 and combines questionnaire information with a national physical examination and biomonitoring program. NCHS also sponsors a National Health Care Survey (NHCS), a National Health Interview Survey (NHIS), a National Immunization Survey (NIS), and a National Survey of Family Growth (NSFG). Similarly designed large-scale efforts to track temporal and area differences for targeted health behaviors within a state include California's Tobacco Survey, Women's Health Survey, and Health Information Survey (Table 3).
Table 3 Sources of survey data. Survey data recorded at the ZIP code level are designed to give valid estimates of risk factor distributions at the State level.
Dataset name Source Agency URL Geographic Resolution
Behavioral Risk Factors Surveillance Survey (BRFSS) Centers for Disease Control ZIP code
National Health and Nutrition Examination Survey (NHANES), National Health Care Survey (NHCS), National Health Interview Survey (NHIS), National Immunization Survey (NIS), National Survey of Family Growth (NSFG). National Center for Health Statistics Metropolitan Statistical Area, National Region
California Tobacco Survey California Department of Health Services ZIP code
California Women's Health Survey California Department of Health Services
ZIP code
California Health Information Survey UCLA Center for Health Policy Research ZIP code
Although population survey data has not been extensively incorporated into GIS studies to date, these resources may in the future provide some opportunity to characterize regional differences in behavioral risk profiles targeted for specific health outcomes.
4. Environmental data
Over the past several decades there has been a large increase in the availability of spatially registered environmental data in the United States and other countries. Much of these data have been collected as a result of environmental regulations or government-funded research efforts. Examples of US programs to collect spatial data on concentrations or releases of pollutants in the environment include the United States Geological Survey (USGS) National Assessment of Water Quality program (NAWQA) , the Environmental Protection Agency (EPA) National Air Toxics Assessment database , and EPA's Toxic Release Inventory program . EPA has organized environmental data in an umbrella database called Envirofacts Data Warehouse . Some states have extensive efforts to collect additional environmental data. An example is California's Pesticide Use Reporting program ) that requires reporting of all agricultural pesticide use at the level of Public Land Survey System sections (a unit approximately one square mile in area).
There are several issues to consider in using these data for assigning "exposure" in epidemiologic studies. Monitoring data collected for regulatory purposes should be carefully evaluated for its usefulness for estimating individual exposures. The fate and transport of the chemicals in the environment should also be considered. Simple proximity measures to sites of chemical releases may not adequately describe the transport of the chemical in the environment. The likely route of exposure should be considered along with the biological plausibility for an association between the exposure and disease under study. Finally, much of the environmental monitoring data was collected within the past decade and reconstructing exposure over longer periods more relevant to cancer incidence will be challenging.
Environmental databases have begun to be used in epidemiology studies of cancer to determine if disease mortality or incidence rates are higher in areas with specific environmental exposures (i.e., ecologic study designs) or as a means of classifying individuals with respect to their potential exposure in an analytic epidemiologic study design (i.e., case-control, cohort studies). With few exceptions, the residence location is used as the geographic location for assigning exposure. Below we provide an overview of the various types of spatially registered exposure data and include examples of their use in epidemiologic studies of cancer.
a. Water quality data
The US EPA is responsible for regulating public drinking water supplies. A water supply is regulated if it has 5 or more connections or serves at least 25 people. Routine monitoring is required for a variety of contaminants and naturally occurring elements including disinfection by-products, arsenic, nitrate, certain pesticides and volatile organic chemicals. States are required to report violations of the Maximum Contaminant Levels (MCL) to EPA. Since 1996, EPA has been required to maintain a National Contaminant Occurrence Database (NCOD) using occurrence data for both regulated and unregulated contaminants in public water systems. The majority of historical public water supply measurement data, however, reside with the states. Some states record the latitude and longitude of the locations where the water samples were taken (location in the distribution system, point of entry to the distribution system, or water source location). The location information is typically not publicly available but may be available to researchers with appropriate approvals.
The water quality data are reported by utility and to be useful for epidemiologic studies a linkage to the towns served must be established. In larger metropolitan areas multiple utilities may serve a city or, conversely, one utility may serve multiple towns and subdivisions. Therefore, establishing an accurate linkage between the study participant's addresses and water utilities is essential to avoid misclassification of exposure. Long-term exposure metrics can be calculated when a lifetime water source history is collected. Examples of studies using public supply water quality monitoring data include studies of disinfection by-products [64-66]., nitrate [67,68]., radionuclides [69,70]., and arsenic [71,72]. Contaminants such as disinfection by-products and volatile organic compounds vary in concentration across a public supply distribution system. GIS-based modeling efforts have been used to improve estimates of exposure at individual residences [73,74].
In contrast to public water supplies, private domestic wells are not regulated and there are no monitoring requirements, although well owners may be required to provide some water quality information upon the sale of a property in some states. Some states have conducted representative surveys of private well water quality [75]. A nationwide survey was conducted by EPA in 1988–1990 [76,77]. The US Centers for Disease Control (CDC) conducted a survey of coliform bacteria, nitrate, and atrazine in private wells in nine Midwestern States . The paucity of historical water quality data for private wells limits the exposure assessment for epidemiologic studies of cancer in this population.
The USGS NAWQA program has been collecting information on nutrients, pesticides, volatile organic compounds, radionuclides, and major ions in more than 50 river basins and aquifers since 1991. All of the measurement data include spatial attributes. Because the goal of this research effort is to understand ambient water quality (not necessarily the same as drinking water quality) these data may not be of direct use in epidemiologic studies. However, the NAWQA data may be useful in modeling efforts to estimate contaminant levels in private wells. EPA also maintains two data management systems containing water quality information collected by federal, state, and private groups for surface and ground waters in all 50 states. The Legacy Data Center (LDC) is an archived database with data dating from the early 20th century up to the end of 1998. STORET contains data collected beginning in 1999, along with older data documented data from the LDC. Table 4 summarizes the sources of water quality data.
Table 4 Sources of Water Quality data
Database name Source Agency URL Geographic Resolution
National Contaminant Occurrence Database EPA Public water utility
National Water Quality Assessment (NAWQA) Data Warehouse USGS
Latitude and longitude
Legacy Data Center/STORET EPA Latitude and longitude
b. Air pollutants
The EPA collects and processes monitoring data from states on six criteria air pollutants (carbon monoxide, nitrogen dioxide, ozone, sulfur dioxide, particulate matter [PM10 and PM2.5], lead) and hazardous air pollutants, of which 188 have been identified. The hazardous air pollutants (HAP), also known as air toxics, are those for which there is some evidence of an increased risk for cancer or adverse reproductive outcomes. Routine monitoring of HAPs is not required and the monitoring data that exists is sparsely distributed compared with the criteria air pollutants. The data are maintained in the Air Quality Systems database.
EPA compiles HAP emissions from stationary sources (points and areas) and mobile sources in a National Toxics Inventory (NTI) database (now combined with the National Emissions Trends data in the National Emissions Inventory database), which is updated at three-year intervals. To do the updates, EPA obtains emissions inventories from state environmental agencies and supplemental data from other sources, including the Toxic Release Inventory. The first nationwide inventory was in 1996. The spatial scale of the emissions data varies by type of source. Location information for point sources emissions is available, whereas area-source emissions are estimated at the county level. Using a dispersion model EPA has estimated the annual average HAP concentrations for each census tract in the contiguous US [78]. These datasets are summarized in Table 5.
Table 5 Sources of Air Quality Data
Dataset name Source Agency URL Geographic Resolution
Air Quality System database EPA Monitoring stations (latitude, longitude)
National Emissions Inventory EPA Varies (point locations, county level)
Air pollutant monitoring data has been used in studies of lung cancer, which have generally employed some type of dispersion model to estimate exposure for metropolitan areas or census tracts [79-81]. Recently the modeled concentrations of HAP have been used to evaluate childhood cancer incidence [44]. Other studies have also evaluated traffic density and childhood cancer incidence [43].
c. Agricultural Pesticides
In the United States the U.S. Department of Agriculture (USDA) is the main federal agency responsible for collecting information on pesticide use on crops and livestock. The availability of historical agricultural pesticide use data in the US has been reviewed [82]. The first comprehensive survey of pesticide use on crops occurred in 1964 [83] and periodic surveys were conducted thereafter through the 1970s. These early surveys only provided national or regional estimates of crop-specific use for individual pesticides. From 1986 onwards, the USDA surveys produced state-specific estimates of pesticide use on field crops in the major producing states and from 1990 onwards, biannual state-specific estimates of pesticide use on fruits and vegetables were also available.
Several states have collected their own pesticide use information but most data collection efforts have been recent. Oregon enacted legislation requiring reporting of agricultural pesticide use beginning in 2002; however, insufficient funding was provided for additional years. State pesticide use data are most comprehensive for California, which has had some type of mandatory reporting for agricultural pesticides since the 1950s, currently overseen by the California Department of Pesticide Regulation. Beginning in 1969, information about restricted-use pesticides was made public. In 1990, a new law required growers to report all pesticide use on crops on a monthly basis, including the pesticide name and manufacturer, crop treated, the public land survey section where the pesticide was applied, the date and time of application, number of acres treated, method of application, and application rates. The availability of this detailed pesticide use data at the spatial scale of a section led to the development of methods to link the use data to cancer incidence data [84] for use in an ecologic study of childhood cancer at the census tract level [42]. The California data have also been used in a case-control study of pancreas cancer [85], cohort study of breast cancer [51], and an as-yet unpublished case-control study of childhood cancer. Methods have also been developed to estimate potential pesticide exposure at residences by linking pesticide use data to crop maps [86,87]. Pesticide "exposure" is assigned to homes that have crop fields within distances that reflect likely pesticide drift. Table 6 summarizes the sources of pesticide data.
Table 6 Sources of Pesticide Data
Dataset name Source Agency URL Geographic Resolution
Agricultural Chemical Use USDA State
California Pesticide Use Reporting database California Department of Pesticide Regulation Public Land Survey Section (approximately one square mile)
d. Industrial releases and hazardous waste
The Emergency Planning and Community Right to Know Act of 1986 in the United States requires certain industries to report to EPA annually their releases and waste management activities involving specific toxic chemicals. The data are available to the public in a database called the Toxics Release Inventory (TRI). Manufacturing, metal mining, coal mining, and electric generating facilities must report the estimated mass of toxic chemicals released into the environment (air, water, land, or underground injection), treated on-site, or shipped off-site for further waste treatment. Reporting is required only for facilities that meet certain minimum criteria in terms of the pounds of toxic chemical produced or processed; persistent chemicals that bioaccumulate are subject to lower minimum reporting requirements. The regulations do not require environmental monitoring, so much of the data are estimates of releases. Location information is reported by the business and is not verified by EPA. Some of the strengths and limitations of these data for environmental health studies has been described [88,89].
Canada also requires reporting of emissions of chemicals rated by the International Agency for Research on Cancer as likely, probable, and possible human carcinogens for 64 industrial sectors [90]. These data form part of the Canadian Environmental Quality Database, which also contains a national inventory of municipal waste disposal sites, municipal drinking water data, air quality data, and historical industrial location and productivity data [91]. A large multi-province case-control study of 18 cancer sites was conducted with the aim of linking residential histories by postal code to the environmental database for cancer surveillance. To date, one analysis of residential proximity to 7 types of heavy industries and risk of non-Hodgkin lymphoma (NHL) has been published. Residential proximity within 3.2 km of copper smelters and <0.8 km of sulfite pulp mills was associated with an increased risk of NHL [92] after adjusting for employment in the industries evaluated. Earlier case-control studies of NHL [93] and leukemia [94] found elevated risks for residing close to industrial sites but these studies relied on a self-reported assessment of the distance of the residence from industrial facilities which may be subject to recall bias.
The EPA maintains information on the location of waste handlers, waste treatment facilities and waste sites that are regulated under the Resource and Conservation Recovery Act (RCRA) and the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA), also known as the Superfund law in the RCRAInfo database available through the Envirofacts Data Warehouse. Information on the location of companies issued permits to discharge waste into rivers is maintained in the Permit Compliance System database (also available through Envirofacts). These data sources are summarized in Table 7.
Table 7 Sources of Hazardous Waste Data
Dataset name Source Agency URL Geographic Resolution
Toxics Release Inventory EPA Latitude, longitude
HazDat ATSDR Latitude, longitude
RCRAInfo EPA Latitude, longitude
Permit Compliance System (PCS)
The U.S. Agency for Toxic Substances and Disease Registry (ATSDR) was established by Congress in 1980 under CERCLA. Since 1986, ATSDR has been required to conduct a public health assessment at each of the sites on the EPA National Priorities List, waste sites deemed to be the most hazardous. The aim of these evaluations is evaluate exposure to hazardous substances and health effects among the population living in vicinity of the site [95]. The location of the sites and information on specific contaminants by the type of media (soil, air, water) in which they were measured are available from the ATSDR HazDat database web site. Limitations of these monitoring data for cancer studies include the limited historical measurement data. A few studies have evaluated cancer incidence among those potentially exposed to hazardous waste sites [96] or municipal waste sites and incinerators [97,98].
The reconstruction of historical exposure to releases from industries and waste sites is difficult for studies of cancers of long latency. A few studies have evaluated proximity and residence duration near sites. Long duration of residence within one-half mile of a chemical plant manufacturing PCBs was positively correlated with blood serum PCB concentrations [99]. However, none of the epidemiologic studies to date determined whether proximity resulted in meaningful exposure to chemicals from the sites. Confounding by socioeconomic status should also be evaluated because manufacturing and waste facilities are more likely to be located in neighborhoods of lower socioeconomic status [100] and socioeconomic status is associated with the incidence of some cancers.
5. Remote sensing/aerial imaging
Remotely sensed data include images of the earth and our atmosphere obtained by satellites or aircraft. The usefulness of the information depends largely on the technology used to obtain the imagery and the additional processing that has been done to georeference the data. The USGS Earth Resources Observation Systems Data Center (EDC) is the major U.S. storehouse of these data. Aerial photography has been available since the early part of the twentieth century. Digital Orthophoto Quadrangles (DOQs) which are digital images of aerial photos which combine the image characteristics of a photo with the georeferenced qualities of a map are available through EDC from 1987 through the present. DOQs are available in black and white, natural color, or color-infrared images and have 1-meter ground resolution. Satellite imagery useful for land cover characterization includes the multispectral Landsat imagery available as early as 1972. USGS has created historical land use and land cover data derived from 1970s and 1980s aerial photography (the Land Use and Land Cover Data). A national land cover datasets (NLCD) derived from Landsat multispectral imagery for 1992 is available. The Multi-resolution Land Characteristics (MRLC) national dataset which represents land cover in 2000 is currently being developed. Table 8 summarizes these data sources. Applications of these data to studies of cancer have included mapping residences on crop maps to estimate their probable exposure to agricultural pesticides [49,87,101].
Table 8 Sources of Remote Sensing Data
Dataset name Source Agency URL Geographic Resolution
Digital orthophoto quadrangles USGS 1:12,000
Satellite imagery USGS 1 meter to 1 km
National Landcover Dataset (NLCD) 1992 USGS 30 meters
Multi-resolution Land Characteristics (MRLC) 2000
Centralized geospatial data availability
The data sources we have described are available from a multitude of federal and state agencies. The National Cancer Institute's Geographic Information Systems web site offers links to many of these sources, as well as links to freely available geographical tools and resources. There have also been several initiatives to try and compile spatial data into a shared, centralized information system [102]. Such centralized systems offer the promise of standardized data coding systems, file formats and geographic boundary definitions. They also facilitate the sharing of metadata, or descriptive information about the data. The leader in this endeavor has been the Federal Geographic Data Committee . The FGDC is a consortium of federal agencies with the charge of developing the National Spatial Data Infrastructure (NSDI), a set of technologies, policies, standards and procedures that facilitate the creation and sharing of geospatial data. Among the achievements of the FGDC is the establishment of the National Spatial Data Clearinghouse, a central catalog of links to geospatial data and metadata. In 2003, an enhanced web portal was launched to further facilitate access to this data. Many states have echoed the national clearinghouse with clearinghouses of their own. The New York GIS Clearinghouse , for example, boasts over 400 member institutions providing links to thousands of datasets.
The cancer data collection community has yet to fully engage this resource. As of January 2004, no cancer incidence or mortality data was available through the national clearinghouse. The keyword "cancer" provided only a link to the Environmental Defense Scorecard, a web site from which various environmental data sets can be accessed, particularly those published by the EPA . Most of the very limited data in the "human health and disease" category accessible through the web portal consisted of hospital and other health facility locations for a smattering of states. In some cases, the steps required to make cancer data available through the national clearinghouse would be modest. For example, the NCI's mortality data, geographic boundary files, and associated metadata used in its Cancer Mortality Maps and Graphs web site are easily accessed and downloaded, and only minor modifications would be required to make them compliant with FGDC standards.
The DataWeb is another centralized online data resource, consisting of a network of online data libraries created in a collaboration between the CDC and the US Census Bureau. The libraries consist of both microdata and aggregate data in numerous categories. Available health data includes NHANES and NHIS survey data and county-level mortality. Information in DataWeb is accessed through DataFerret, an application that prepares data sets for the user to download. It allows users to select a "databasket" of variables and then recode those variables as needed. Users develop and customize data tables and may download them to their desktop in a variety of common formats.
Conclusion
In this article we have surveyed the distinctive characteristics of spatial data, along with commonly available sources of data relevant to etiologic cancer research. Spatial analysis is invaluable for data exploration, identification of geographic patterns, generation of new hypotheses, and providing supporting evidence about existing hypotheses. A geographic perspective will be increasingly relevant as GIS software, spatial analytic methods, and the availability and quality of geographically referenced data continues to improve.
==== Refs
Lee CV Irving JL Sources of spatial data for community health planning Journal of Public Health Management and Practice 1999 5 7 22 10538418
Burkitt DP Rothschild HR Geography of a disease: purpose and possibilities from geographical medicine Biocultural aspects of disease 1981 New York, Academic Press
Gould P Wallace R Spatial structures and scientific paradoxes in the AIDS pandemic Geografiska Annaler B 1994 76 105 116
Miller HJ Wentz EA Representation and spatial analysis in geographic information systems Annals of the Association of American Geographers 2003 93 574 594 10.1111/1467-8306.9303004
Jacquez GM Current practices in spatial analysis of cancer data: flies in the ointment, or, the limitations of spatial analysis International Journal of Health Geographics 2004 3 22 15479473 10.1186/1476-072X-3-22
Krieger N Place, space, and health: GIS and epidemiology Epidemiology 2003 14 384 385 12843759
Boscoe FP McLaughlin CC The effect of seasonal residence on cancer incidence rates Journal of Registry Management 2002 29 3 7
McElroy JA Remington PL Trentham-Dietz A Robert SA Newcomb PA Geocoding addresses from a large population-based study: lessons learned Epidemiology 2003 14 399 407 12843762
Bonner MR Han D Nie J Rogerson P Vena JE Freudenheim JL Positional accuracy of geocoded addresses in epidemiologic research Epidemiology 2003 14 408 412 12843763
Hurley SE Saunders TM Nivas R Hertz A Reynolds P Post office box addresses: a challenge for geographic information system-based studies Epidemiology 2003 14 386 391 12843760
Fotheringham AS Wong DWS The modifiable areal unit problem in multivariate statistical analysis Environment and Planning A 1991 23 1025 1044
Openshaw S The modifiable areal unit problem Concepts and techniques in modern geography 38 1984 Norwich, Geobooks
Holt D Steel DG Tranmer M Wrigley N Aggregation and ecological effects in geographically based data Geographical Analysis 1996 28 244 261
Cohen BL Lung cancer rate versus mean radon level in US counties of various characteristics Health Physics 1997 72 114 119 8972836
Lagarde F Pershagen G Parallel analyses of individual and ecologic data on residential radon, cofactors, and lung cancer in Sweden American Journal of Epidemiology 1999 149 268 274 9927223
Darby S Deo H Doll R Whitley E A parallel analysis of individual and ecological data on residential radon and lung cancer in south-west England Journal of the Royal Statistical Society A 2001 164 193 203
Lubin JH On the discrepancy between epidemiologic studies in individuals of lung cancer and residential radon and Cohen's ecologic regression Health Physics 1998 75 4 10 9645660
Darby S Hill D Doll R Radon: a likely carcinogen at all exposures Annals of Oncoogy 2001 12 1341 1351 10.1023/A:1012518223463
Gotway CA Young LJ Combining incompatible spatial data Journal of the American Statistical Association 2002 97 632 648 10.1198/016214502760047140
Dobson JE Bright EA Coleman PR Durfee RC Worley BA LandScan: a global population database for estimating populations at risk Photogrammetric Engineering and Remote Sensing 2000 66 849 857
Mennis J Generating surface models of population using dasymetric mapping Professional Geographer 2003 55 31 42
Mugglin AS Carlin BP Zhu L Conlon E Bayesian areal interpolation, estimation, and smoothing: an inferential approach for geographic information systems Environment and Planning A 1999 31 1337 1352
Mugglin AS Carlin BP Gelfand AE Fully model-based approaches for spatially misaligned data Journal of the American Statistical Association 2000 95 877 887
Twigg L Moon G Jones K Predicting small-area health-related behaviour: a comparison of smoking and drinking indicators Social Science and Medicine 2000 50 1109 1120 10714931 10.1016/S0277-9536(99)00359-7
Pearce J Boyle P Flowerdew R Predicting smoking behaviour in census output areas across Scotland Health and Place 2003 9 139 149 12753797 10.1016/S1353-8292(02)00036-9
Dubin RA Spatial autocorrelation: a primer Journal of Housing Economics 1998 7 304 327 10.1006/jhec.1998.0236
Griffith DA Spatial autocorrelation and spatial filtering: gaining understanding through theory and scientific visualization 2003 New York, Springer-Verlag
Fotheringham AS Brunsdon C Charlton M Geographically weighted regression: the analysis of spatially varying relationships 2002 Chichester, John Wiley & Sons
Brunsdon C McClatchey J Unwin DJ Spatial variation in the average rainfall-altitude relationship in Great Britain: an approach using geographically weighted regression International Journal of Climatology 2001 21 455 466 10.1002/joc.614
Fotheringham AS Brunsdon C Local forms of spatial analysis Geographical Analysis 1999 31 340 358
Anselin L Spatial dependence and spatial structural instability in applied regression analysis Journal of Regional Science 1990 30 185 207
Gelman A Price PN All maps of parameter estimates are misleading Statistics in Medicine 1999 18 3221 3234 10602147 10.1002/(SICI)1097-0258(19991215)18:23<3221::AID-SIM312>3.3.CO;2-D
Boscoe FP Pickle LW Choosing geographic units for choropleth rate maps, with an emphasis on public health applications Cartography and Geographic Information Science 2003 30 237 248 10.1559/152304003100011171
MacEachren AM Brewer CA Pickle LW Visualizing georeferenced data: representing reliability of health statistics Environment and Planning A 1998 30 1547 1561
Kafadar K Smoothing geographical data, particularly rates of disease Statistics in Medicine 1996 15 2539 2560 8961462 10.1002/(SICI)1097-0258(19961215)15:23<2539::AID-SIM379>3.3.CO;2-2
Talbot TO Kulldorff M Forand SP Haley VB Evaluation of spatial filters to create smoothed maps of health data Statistics in Medicine 2000 19 2399 2408 10960861 10.1002/1097-0258(20000915/30)19:17/18<2399::AID-SIM577>3.3.CO;2-I
Ghosh M Natarajan K Stroud TWF Carlin BP Generalized linear models for small-area estimation Journal of the American Statistical Association 1998 93 273 282
Mollié A Lawson AB Bayesian and empirical Bayes approaches to disease mapping In Disease mapping and risk assessment for public health 1999 Chichester, John Wiley & Sons 15 29
Semenciw RM Le ND Marrett LD Robson DL Turner D Walter SD Methodological issues in the development of the Canadian Cancer Incidence Atlas Statistics in Medicine 2000 19 2437 2449 10960864 10.1002/1097-0258(20000915/30)19:17/18<2437::AID-SIM580>3.0.CO;2-J
Toledano MB Jarup L Best NG Wakefield JC Elliott P Spatial variation and temporal trends of testicular cancer in Great Britain British Journal of Cancer 2001 84 1482 1487 11384098 10.1054/bjoc.2001.1739
Dockerty JD Sharples KJ Borman B An assessment of spatial clustering of leukaemias and lymphomas among young people in New Zealand Journal of Epidemiology and Community Health 1999 53 154 158 10396492
Reynolds P Von Behren J Gunier RB Goldberg DE Hertz A Harnly ME Childhood cancer and agricultural pesticide use: an ecologic study in California Environmental Health Perspectives 2002 110 319 324 11882484
Reynolds P Von Behren J Gunier RB Goldberg DE Hertz A Smith D Traffic patterns and childhood cancer incidence rates in California, United States Cancer Causes and Control 2002 13 665 673 12296514 10.1023/A:1019579430978
Reynolds P Von Behren J Gunier RB Goldberg DE Hertz A Smith DF Childhood cancer incidence rates and hazardous air pollutants in California: an exploratory analysis Environmental Health Perspectives 2003 111 663 668 12676632
Reynolds P Von Behren J Gunier RB Goldberg DE Hertz A Residential exposure to traffic in California and childhood cancer Epidemiology 2004 15 6 12 14712141 10.1097/01.ede.0000101749.28283.de
Crosignani P Tittarelli A Borgini A Codazzi T Rovelli A Porro E Contiero P Bianchi N Tagliabue G Fissi R Rossitto F Berrino F Childhood leukemia and road traffic: A population-based case-control study International Journal of Cancer 2004 108 596 599 10.1002/ijc.11597
Langholz B Ebi KL Thomas DC Peters JM London SJ Traffic density and the risk of childhood leukemia in a Los Angeles case-control study Annals of Epidemiology 2002 12 482 487 12377426 10.1016/S1047-2797(01)00317-9
Raaschou-Nielsen O Hertel O Thomsen BL Olsen JH Air pollution from traffic at the residence of children with cancer American Journal of Epidemiology 2001 153 433 443 11226975 10.1093/aje/153.5.433
Brody JG Vorhees DJ Melly SJ Swedis SR Drivas PJ Rudel RA Using GIS and historical records to reconstruct residential exposure to large-scale pesticide application Journal of Exposure Analysis and Environmental Epidemiology 2002 12 64 80 11859434 10.1038/sj.jea.7500205
Brody JG Aschengrau A McKelvey W Rudel RA Swartz CH Kennedy T Breast cancer risk and historical exposure to pesticides from wide-area applications assessed with GIS Environmental Health Perspectives 2004 112 889 897 15175178
Reynolds P Hurley SE Goldberg DE Yerabati S Gunier RB Hertz A Anton-Culver H Bernstein L Deapen D Horn-Ross PL Peel D Pinder R Ross RK West D Wright WE Ziogas A Residential proximity to agricultural pesticide use and incidence of breast cancer in the California Teachers Study cohort Environmental Research 2004 96 206 218 15325881 10.1016/j.envres.2004.03.001
Ralston B GIS and Public Data 2004 Clifton Park, Delmar Learning
Boscoe FP Miller BA Population estimation error and its impact on 1991–1999 cancer rates Professional Geographer 2004 56 516 529
Schenker N Parker JD From single-race reporting to multiple-race reporting: using imputation methods to bridge the transition Statistics in Medicine 2003 22 1571 1587 12704616 10.1002/sim.1512
United States Census Bureau Estimates and projections area methodology: county population estimates by age, sex, race, and Hispanic origin for July 1, 2002 2003 Washington DC, United States Census Bureau
Robinson JG West KK Adlakha A Coverage of the population in Census 2000: results from demographic analysis Population Research and Policy Review 2002 21 19 38 10.1023/A:1016537822148
United States Census Bureau Technical summary of ACE Revision II for the Committee on National Statistics 2003 Washington DC, United States Census Bureau
Krieger N Quesenberry C JrPeng T Horn-Ross P Stewart S Brown S Swallen K Guillermo T Suh D Alvarez-Martinez L Ward F Social class, race/ethnicity, and incidence of breast, cervix, colon, lung, and prostate cancer among Asian, Black, Hispanic, and White residents of the San Francisco Bay Area, 1988–92 (United States) Cancer Causes and Control 1999 10 525 537 10616822 10.1023/A:1008950210967
Robbins AS Whittemore AS Thom DH Differences in socioeconomic status and survival among White and Black men with prostate cancer American Journal of Epidemiology 2000 151 409 416 10695600
Ghori FY Gutterman-Litofsky DR Jamal A Yeung SCJ Arem R Sherman SI Socioeconomic factors and the presentation, management, and outcome of patients with differentiated thyroid carcinoma Thyroid 2002 12 1009 1016 12490079 10.1089/105072502320908358
Kwok RK Yankaskas BC The use of census data for determining race and education as SES indicators: A validation study Annals of Epidemiology 2001 11 171 177 11293403 10.1016/S1047-2797(00)00205-2
Pickle LW Su Y Within-state geographic patterns of health insurance coverage and health risk factors in the United States American Journal of Preventive Medicine 2002 22 75 83 11818175 10.1016/S0749-3797(01)00402-0
Jia H Muennig P Borawski E Comparison of small-area analysis techniques for estimating county-level outcomes American Journal of Preventive Medicine 2004 26 453 460 15165663 10.1016/j.amepre.2004.02.004
Cantor KP Lynch CF Hildesheim ME Dosemeci M Lubin J Alavanja M Craun G Drinking water source and chlorination byproducts. I. Risk of bladder cancer Epidemiology 1998 9 21 28 9430264 10.1097/00001648-199801000-00003
Hildesheim ME Cantor KP Lynch CF Dosemeci M Lubin J Alavanja M Craun G Drinking water source and chlorination byproducts. II. Risk of colon and rectal cancers Epidemiology 1998 9 29 35 9430265 10.1097/00001648-199801000-00004
King WD Marrett LD Case-control study of bladder cancer and chlorination by-products in treated water (Ontario, Canada) Cancer Causes and Control 1996 7 596 604 8932920 10.1007/BF00051702
Ward MH Mark SD Cantor KP Weisenburger DD Correa-Villasenor A Zahm SH Drinking water nitrate and the risk of non-Hodgkin's lymphoma Epidemiology 1996 7 465 471 8862975
Ward MH Cantor KP Riley D Merkle S Lynch CF Nitrate in public water supplies and risk of bladder cancer Epidemiology 2003 14 183 190 12606884 10.1097/00001648-200303000-00012
Lyman GH Lyman CG Johnson W Association of leukemia with radium groundwater contamination Journal of the American Medical Association 1985 254 621 626 4009896 10.1001/jama.254.5.621
Bean JA Isacson P Hahne RM Kohler J Drinking water and cancer incidence in Iowa. II. Radioactivity in drinking water American Journal of Epidemiology 1982 116 924 932 7148818
Ferreccio C Gonzalez C Milosavjlevic V Marshall G Sancha AM Smith AH Lung cancer and arsenic concentrations in drinking water in Chile Epidemiology 2000 11 673 679 11055628 10.1097/00001648-200011000-00010
Chiou HY Hsueh YM Liaw KF Horng SF Chiang MH Pu YS Lin JS Huang CH Chen CJ Incidence of internal cancers and ingested inorganic arsenic: a seven-year follow-up study in Taiwan Cancer Research 1995 55 1296 1300 7882325
Cohn P Savrin J Fagliano J Mapping of volatile organic chemicals in New Jersey water systems Journal of Exposure Analysis and Environmental Epidemiology 1999 9 171 180 10412666 10.1038/sj.jea.7500017
Gallagher MD Nuckols JR Stallones L Savitz DA Exposure to trihalomethanes and adverse pregnancy outcomes Epidemiology 1998 9 484 489 9730025 10.1097/00001648-199809000-00001
Johnson CJ Kross BC Continuing importance of nitrate contamination of groundwater and wells in rural areas American Journal of Industrial Medicine 1990 18 449 456 2248249
Environmental Protection Agency National survey of pesticides in drinking water wells Phase I report (EPA 570/9-90-015) 1990 Washington DC, Environmental Protection Agency
Environmental Protection Agency Another look: national survey of pesticides in drinking water wells Phase 2 report (EPA 570/9-91-020) 1992 Washington DC, Environmental Protection Agency
Rosenbaum AS Axelrad DA Woodruff TJ Wei YH Ligocki MP Cohen JP National estimates of outdoor air toxics concentrations Journal of the Air and Waste Management Association 1999 49 1138 1152 10616743
Pope CA IIIBurnett RT Thun MJ Calle EE Krewski D Ito K Thurston GD Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution Journal of the American Medical Association 2002 287 1132 1141 11879110 10.1001/jama.287.9.1132
Nafstad P Haheim LL Oftedal B Gram F Holme I Hjermann I Leren P Lung cancer and air pollution: a 27 year follow up of 16 209 Norwegian men Thorax 2003 58 1071 1076 14645978 10.1136/thorax.58.12.1071
Cohen AJ Outdoor air pollution and lung cancer Environmental Health Perspectives 2000 108 743 750 10931793
Ward MH Prince JR Stewart PA Zahm SH Determining the probability of pesticide exposures among migrant farmworkers: results from a feasibility study American Journal of Industrial Medicine 2001 40 538 553 11675623
USDA Economic Research Service Quantities of pesticides used by farmers in 1964 Agricultural Economic Report No AER-131 AER-131 1968 Washington DC, USDA Economic Research Service
Gunier RB Harnly ME Reynolds P Hertz A Von Behren J Agricultural pesticide use in California: pesticide prioritization, use densities, and population distributions for a childhood cancer study Environmental Health Perspectives 2001 109 1071 1078 11689348
Clary T Ritz B Pancreatic cancer mortality and organochlorine pesticide exposure in California, 1989–1996 American Journal of Industrial Medicine 2003 43 306 313 12594778 10.1002/ajim.10188
Ward MH Nuckols JR Weigel SJ Maxwell SK Cantor KP Miller RS Identifying populations potentially exposed to agricultural pesticides using remote sensing and a Geographic Information System Environmental Health Perspectives 2000 108 5 12 10622770
Rull RP Ritz B Historical pesticide exposure in California using pesticide use reports and land-use surveys: an assessment of misclassification error and bias Environmental Health Perspectives 2003 111 1582 1589 14527836
Neumann CM Improving the U.S. EPA Toxic Release Inventory database for environmental health research Journal of Toxicology and Environmental Health B 1998 1 259 270
Neumann CM Forman DL Rothlein JE Hazard screening of chemical releases and environmental equity analysis of populations proximate to toxic release inventory facilities in Oregon Environmental Health Perspectives 1998 106 217 226 9494125
Argo J Retrospective exposure assessment with emission inventories: a new approach to an old problem Environmetrics 1998 9 505 518 10.1002/(SICI)1099-095X(199809/10)9:5<505::AID-ENV319>3.0.CO;2-K
Johnson KC Mao Y Argo J Dubois S Semenciw R Lava J The National Enhanced Cancer Surveillance System: a case control approach to environment-related cancer surveillance in Canada Environmetrics 1998 9 495 504 10.1002/(SICI)1099-095X(199809/10)9:5<495::AID-ENV318>3.0.CO;2-H
Johnson KC Pan S Fry R Mao Y Residential proximity to industrial plants and non-Hodgkin lymphoma Epidemiology 2003 14 687 693 14569184
Linos A Blair A Gibson RW Everett G Van Lier S Cantor KP Schuman L Burmeister L Leukemia and non-Hodgkin's lymphoma and residential proximity to industrial plants Archives of Environmental Health 1991 46 70 74 2006896
Shore DL Sandler DP Davey FR McIntyre OR Bloomfield CD Acute leukemia and residential proximity to potential sources of environmental pollutants Archives of Environmental Health 1993 48 414 420 8250593
Amler RW Lybarger JA Research program for neurotoxic disorders and other adverse health outcomes at hazardous chemical sites in the United States of America Environmental Research 1993 61 279 284 8495669 10.1006/enrs.1993.1072
White E Aldrich TE Geographic studies of pediatric cancer near hazardous waste sites Archives of Environmental Health 1999 54 390 397 10634228
Goldberg MS al Homsi N Goulet L Riberdy H Incidence of cancer among persons living near a municipal solid waste landfill site in Montreal, Quebec Archives of Environmental Health 1995 50 416 424 8572719
Comba P Ascoli V Belli S Benedetti M Gatti L Ricci P Tieghi A Risk of soft tissue sarcomas and residence in the neighbourhood of an incinerator of industrial wastes Occupational and Environmental Medicine 2003 60 680 683 12937191 10.1136/oem.60.9.680
Orloff KG Dearwent S Metcalf S Kathman S Turner W Human exposure to polychlorinated biphenyls in a residential community Archives of Environmental Contamination and Toxicology 2003 44 125 131 12434227 10.1007/s00244-002-1301-5
Perlin SA Wong D Sexton K Residential proximity to industrial sources of air pollution: interrelationships among race, poverty, and age Journal of the Air and Waste Management Association 2001 51 406 421 11266104
Ward MH Giglierano J Nuckols JR Wolter C Colt JS Camann D Hartge P Proximity to crops and residential exposure to agricultural pesticides in Iowa Proceedings, EUROHEIS/SAHSU Conference 2003 140 145
Falke SR Environmental data: finding it, sharing it, and using it Journal of Urban Technology 2002 9 111 124 10.1080/1063073022000016504
| 15574197 | PMC539245 | CC BY | 2021-01-04 16:39:02 | no | Int J Health Geogr. 2004 Dec 1; 3:28 | utf-8 | Int J Health Geogr | 2,004 | 10.1186/1476-072X-3-28 | oa_comm |
==== Front
J Exp Clin Assist ReprodJournal of Experimental & Clinical Assisted Reproduction1743-1050BioMed Central London 1743-1050-1-31558828610.1186/1743-1050-1-3ResearchTissue engineering, stem cells, cloning, and parthenogenesis: new paradigms for therapy Hipp Jason [email protected] Anthony [email protected] Wake Forest Institute for Regenerative Medicine Wake Forest University School of Medicine Winston Salem, North Carolina USA2 Wake Forest University School of Medicine Medical Center Blvd. Winston Salem, North Carolina 27157 USA2004 8 12 2004 1 3 3 30 11 2004 8 12 2004 Copyright © 2004 Hipp and Atala; licensee BioMed Central Ltd.2004Hipp and Atala; 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.
Patients suffering from diseased and injured organs may be treated with transplanted organs. However, there is a severe shortage of donor organs which is worsening yearly due to the aging population. Scientists in the field of tissue engineering apply the principles of cell transplantation, materials science, and bioengineering to construct biological substitutes that will restore and maintain normal function in diseased and injured tissues. Both therapeutic cloning (nucleus from a donor cell is transferred into an enucleated oocyte), and parthenogenesis (oocyte is activated and stimulated to divide), permit extraction of pluripotent embryonic stem cells, and offer a potentially limitless source of cells for tissue engineering applications. The stem cell field is also advancing rapidly, opening new options for therapy. The present article reviews recent progress in tissue engineering and describes applications of these new technologies that may offer novel therapies for patients with end-stage organ failure.
==== Body
Introduction
The goal of tissue engineering is to repair organ pathologies such as those acquired congenitally or by cancer, trauma, infection, or inflammation. It is based upon the foundations of cell transplantation and materials science. Tissue can be engineered 1) in vivo- by stimulating the body's own regeneration response with the appropriate biomaterial, or 2) ex vivo- cells can be expanded in culture, attached to a scaffold and then reimplanted into the host. Cells may be heterologous (different species), allogeneic (same species, different individual), or autologous (same individual). Autologous cells are preferred because they will not evoke an immunologic response and thus the deleterious side effects of immunosuppressive agents can be avoided.
The ideal autologous cells can often be found within the organ itself. These cells (committed precursors) may be isolated, expanded and transplanted back into the same patient, thus representing an autologous transplantation resource. Previously, urothelial cells could be grown in the laboratory setting with only limited expansion. Several protocols were developed over the last 20 years which identified the undifferentiated cells and kept them undifferentiated during their growth phase [1-4]. Using such cell culture methods it is now possible to expand a urothelial strain from a single specimen which initially covered a surface area of 1 cm2 to one that covers a surface area of >4000 m2 (an area equivalent to one football field) within 8–14 weeks. These studies indicate the possibility of collecting autologous bladder cells from human patients, expanding them in culture, and returning them to the human donor in sufficient quantities for reconstructive purposes [1,3-11]. Major advances have been achieved within the past decade regarding possible expansion of several primary human cell types with specific techniques that employ autologous cells for clinical application.
While autologous cells are recognized as the ideal transplantation resource, many patients with end-stage organ disease are unable to yield sufficient cells for expansion and transplantation. Furthermore, some primary autologous human cells cannot be expanded from particular organs (i.e. pancreas, liver). Stem cells are envisioned as being an alternate source of cells from which the desired tissue can be derived. Human embryonic stem cells (HESC) can be derived from discarded non transferred embryos and have the advantage of being pluripotential (the ability to differentiate into all tissues of the embryo) and able to self-renew indefinitely. However, their clinical application is limited because they represent an allogeneic resource and thus their use would require high dose immunosuppressant therapy.
New stem cell technologies such as somatic cell nuclear transfer (therapeutic cloning) and parthenogenesis offer an exciting alternative to create an inexhaustible supply of ESC that can differentiate into all cell types of the embryo, while not being rejected by the patient's immune system. Although many tissues have been created with ESC, they are not used clinically because of an inability to control differentiation. Hence, their ability to form multiple tissue types also becomes their limitation. New genomics and bioinformatics technologies have and will continue to offer new insights into the understanding of ESC growth and differentiation and their application to engineering tissues. In the near future, these new technologies will allow for the generation of an unlimited supply of any cell type in the body.
Stem cells
The political and ethical controversy surrounding stem cells began in 1998 with the creation of HESC derived from discarded, non-transferred human embryos[12]. The HESC were isolated from the inner cell mass of a blastocyst (5 days post-fertilization embryo) using an immunosurgical technique whereby the blastocyst was incubated with antibodies specific to trophectoderm. Complement proteins then resulted in lysis of the trophectoderm so that the only surviving cells were the inner cell mass [13]. Given that some cells can not be expanded ex vivo, ESC can potentially be the ideal resource for tissue engineering because of two fundamental properties, 1) the ability to self-renew indefinitely, and 2) the ability to differentiate into all three germ layers.
With the current restrictions surrounding HESC work, many proponents of stem cell research have sought to modify the ban to incorporate the thousands of non-transferred frozen embryos resulting from IVF to be used for the creation of more HESC lines. A SART-RAND study identified approximately 400,000 frozen embryos in storage since the late 1970s [14]. However, only 2.8% of these have been designated for research. Of the 11,000 embryos designated for research, only 65% of these (n = 7,334) are expected to survive the freeze/thaw process. From this, 25% are expected to develop to blastocyst stage (n = 1, 834). If one assumes a 15% efficiency rate for establishment of a HESC line from blastocysts (as suggested by previous studies [12,15]), it may be estimated that approximately 275 HESC could be created from excess frozen embryos. However, the real number of HESC line generated would actually be much lower since not all frozen embryos allocated for research would be used to create HESC lines. Furthermore, even if the maximum possible number of HESC lines could be derived from human frozen embryos, the clinical application of such cells would be limited by the potential rejection from another individual's immune system. New stem cell technologies (such as somatic cell nuclear transfer and parthenogenesis) promise to overcome this limitation.
Somatic cell nuclear transfer (therapeutic cloning)
Somatic cell nuclear transfer (SCNT) entails the removal of an oocyte nucleus followed by its replacement with a nucleus derived from a somatic cell obtained from that patient. Activation with chemicals or electric shock stimulates cell division up to the blastocyst stage at which time the inner cell mass is isolated and cultured, resulting in ESC. This approach is distinct from reproductive cloning because the blasotcyst is not transplanted back to the uterus. Hence, development does not proceed beyond the 100 cell stage. This process also differs from fertilization since no sperm is used in this process. The resulting ESC are perfectly matched to the patients immune system and no immunosuppressants would therefore be required to prevent rejection.
While interest in the field of nuclear cloning remains high since the birth of Dolly (1997), the first successful nuclear transfer was actually reported over fifty years ago by Briggs and King [16]. Cloned frogs, which were the first vertebrates derived from nuclear transfer, were subsequently reported by Gurdon in 1962 [17] although the nuclei were derived from non-adult sources. Indeed, in just the past six years alone important advances in nuclear cloning technology have been reported – a pace of discovery that betokens the relative immaturity of this research arena. In fact Dolly was not the first cloned mammal to be produced from adult cells. Live lambs were produced in 1996 using nuclear transfer and differentiated epithelial cells, although these were derived from embryonic discs [18]. To be sure, the significance of the Dolly report was that this described the first mammal to be derived from an adult somatic cell using nuclear transfer [19]. Subsequently, animals from several species have been grown using nuclear transfer technology, including cattle [20], goats [21,22], mice [23], and pigs [24-27].
A better understanding of the differences between reproductive cloning and therapeutic cloning may help alleviate some of the controversy surrounding these technologies [28,29]. Banned in most countries for human applications, reproductive cloning is used to generate an embryo that has the identical genetic material as its cell source. Such an embryo could then be implanted into the uterus of a female to give rise to a liveborn infant that is a clone of the donor. In contrast, therapeutic cloning is used to generate only ESC lines whose genetic material is identical to that of its source. These autologous stem cells have the potential to become almost any type of cell in the adult body, and thus would be useful in tissue and organ replacement applications [30]. Therefore, therapeutic cloning (SCNT) may provide an alternative source of transplantable cells. Figure 1 shows the strategy of combining therapeutic cloning with tissue engineering to develop tissues and organs. It has been estimated that approximately 3,000 people die every day in USA of diseases that could have been treated with stem cells-derived tissues [31]. With current allogeneic tissue transplantation protocols, rejection is a frequent complication because of immunologic incompatibility and thus immunosuppressive drugs are generally required to manage host-versus-graft disease [30]. The use of transplantable tissue and organs derived from therapeutic cloning could obviate unwanted immune responses typically associated with transplantation of non-autologous tissues [32].
Figure 1 Strategy for therapeutic cloning and tissue engineering
While promising, somatic cell nuclear transfer technology has certain limitations requiring further improvement before it can be applied widely in clinical practice. Currently, the efficiency of the overall cloning process is quite low as the majority of embryos derived from animal cloning do not survive after implantation [33-35]. In practical terms, multiple nuclear transfers must be performed in order to produce one live offspring for animal cloning applications. The potential for cloned embryos to grow into live offspring ranges between <1 and 18% for sheep, pigs, and mice [36]. However, greater success (~ 80%) has been reported in cattle [37], a result which may in part be due to availability of advanced laboratory technologies specifically developed for this species for agricultural/breeding purposes. To improve cloning efficiencies, further improvements are required in the multiple complex steps of nuclear transfer such as enucleation and reconstruction, oocyte activation, and synchronization of cell cycle between donor cells and recipient oocytes [38].
It must be noted that abnormalities have been found in liveborn clones including macrosomia with an enlarged placenta ("large-offspring syndrome") [39], respiratory distress, defects of the kidney, liver, heart, and brain [40], obesity [41], and premature death [42]. These may be related to epigenetics of cloned cells which involve reversible modifications of DNA, while the original DNA (genetic) sequences remain intact. Faulty epigenetic modulation in clones may result from altered DNA methylation and/or histone modifications causing the overall chromatin structure of somatic nuclei not to be reprogrammed to an embryonic pattern of expression [30]. Reactivation of key embryonic genes at the blastocyst stage usually does not occur in embryos cloned from somatic cells, while embryos cloned from embryos consistently express early embryonic genes[43,44]. Proper epigenetic reprogramming to an embryonic state may help to improve the cloning efficiency and reduce the incidence of abnormal cloned cells.
Novel applications of somatic cell nuclear transfer (therapeutic cloning)
We applied principles of both tissue engineering and therapeutic cloning in an effort to produce genetically identical renal tissue in an animal model (Bos taurus) [45]. Bovine skin fibroblasts from adult Holstein steers were obtained by ear notch and single donor cells were isolated and microinjected into the perivitelline space of donor enucleated oocytes (nuclear transfer). The resulting blastocysts were transferred to the uterus of progestin-synchronized recipients permit further in vivo growth. After 12 weeks cloned renal cells were harvested, expanded in vitro, then seeded onto biodegradable scaffolds. The constructs (consisting of cells + scaffolds) were then implanted into the subcutaneous space of the same steer from which the cells were cloned to allow for tissue growth.
The kidney is a complex organ with multiple cell types and a complex functional anatomy rendering it one of the most difficult organs to reconstruct [46,47]. Previous efforts in tissue engineering of the kidney have been directed toward development of extracorporeal renal support systems made of biological and synthetic components [48-54]. Although ex vivo renal replacement devices are known to be life-sustaining, there are obvious benefits for patients with end-stage kidney disease if such devices could be implanted long-term without the need for an extracorporeal perfusion circuit or immunosuppressive drugs.
Cloned renal cells were seeded on scaffolds consisting of three collagen-coated cylindrical polycarbonate membranes (figure 2). The ends of the three membranes of each scaffold were connected to catheters terminating in a collecting reservoir. This created a renal neo-organ with a mechanism for collecting the excreted urinary fluid (figure 3). Scaffolds with the collecting devices were transplanted subcutaneously into the same steer from which the genetic material originated and retrieved 12 weeks after implantation.
Figure 2 Combining therapeutic cloning and tissue engineering to produce kidney tissue, an illustration of the tissue-engineered renal unit.
Figure 3 Renal unit seeded with cloned cells, three months after implantation, showing the accumulation of urinelike fluid.
Chemical analysis of the urine-like fluid (for urea nitrogen/creatinine levels, electrolyte levels, specific gravity, and glucose concentration) revealed that the implanted renal cells possessed filtration, reabsorption, and secretory capabilities. Histological examination of the retrieved implants revealed extensive vascularization and self-organization of the cells into glomeruli- and tubule-like structures. A clear continuity between glomeruli, tubules, and the polycarbonate membrane was noted that allowed the passage of urine into the collecting reservoir (figure 4). Immunohistochemical analysis with kidney-specific antibodies revealed the presence of renal proteins, and RT-PCR analysis confirmed the transcription of renal specific RNA in the cloned specimens. Western blot analysis confirmed the presence of elevated renal-specific protein levels.
Figure 4 Clear unidirectional continuity between the mature glomeruli, their tubules, and the polycarbonate membrane.
As previous studies have confirmed bovine clones harbor mitochondrial DNA (mtDNA) of strictly oocyte origin [55-57], the donor egg's mtDNA was thought to be a potential source of immunologic incompatibility. Differences in mtDNA-encoded proteins expressed by cloned cells could stimulate a T-cell response specific for mt-DNA-encoded minor histocompatibility antigens when cloned cells are implanted back into the original nuclear donor [58]. We used nucleotide sequencing of the mtDNA genomes of the clone and fibroblast nuclear donor to identify potential antigens in the muscle constructs. Only two amino acid substitutions were noted to distinguish cells from the clone and the nuclear donor. Since peptide-binding motifs for bovine MHC class I molecules remain poorly understood, there is no reliable method to predict the impact of these amino acid substitutions on bovine histocompatibility.
Oocyte-derived mtDNA was also considered to be a potential source of immunologic incompatibility in cloned renal cells. Maternally transmitted minor histocompatibility antigens in mice have been shown to stimulate both skin allograft rejection in vivo and cytotoxic T lymphocytes expansion in vitro [58] that could prevent the use of these cloned constructs in patients with chronic rejection of major histocompatibility-matched human renal transplants [59,60]. We tested for a possible T-cell response to the cloned renal devices using delayed-type hypersensitivity testing in vivo and Elispot analysis of interferon-gamma secreting T-cells in vitro. Both analyses revealed that the cloned renal cells showed no evidence of T-cell response, suggesting that rejection will not necessarily occur in the presence of oocyte-derived mtDNA (figure 5). This finding may represent a step forward in overcoming the histocompatibility problem of stem cell therapy [47].
Figure 5 Elispot analyses of the frequencies of T-cells that secrete IFN-gamma after primary and secondary stimulation with allogeneic renal cells, cloned renal cells, or nuclear donor fibroblasts.
These studies demonstrated that cells derived from nuclear transfer can be successfully harvested, expanded in culture, and transplanted in vivo with the use of biodegradable scaffolds on which the single suspended cells can organize into tissue structures that are genetically identical to that of the host. These studies were the first demonstration of the use of therapeutic cloning for regeneration of tissues in vivo. Others in the field have created mouse SCNT derived c-kit-positive stem cells to restore infarcted myocardium [61], dopaminergic neurons to correct the phenotype of a mouse model of Parkinson disease [62]. The first HESC line derived from SCNT was created in February, 2004 [63].
Parthenogenesis
Parthenogenesis (<Gr. "virgin birth") is production of offspring by a female with no genetic contribution from a male and without meiotic chromosome reduction. The process is common reproductive strategy among insects such as aphids, flies, ants, and honeybees, but is also known to occur in vertebrates including lizards, snakes, fish, birds, and amphibians. The first demonstration of artificially-stimulated parthenogenesis in vitro was made by Jacques Loeb (1899), who was able to activate oocytes from sea urchins and frogs by pricking them with a needle or by changing the ambient salt concentration. Pincus (1939) demonstrated parthenogenetic activation of mammalian eggs using temperature and chemical stimuli. Thus far, parthenogenetic activation of eggs has been studied in a variety of mammals including mice, goats, cows, monkeys, and humans. Plachot et al. described parthenogenesis in humans by examining 800 human oocytes and showed that 12 activated parthenogenetically and four underwent normal cleavage[64]. Although there have been no reports of naturally-occurring human parthenotes, a human parthenogenetic chimera has been described [65]. The juvenile patient presented with developmental delay, apparent sex reversal, and entirely parthenogenetic blood leukocytes. This finding confirmed the viability of chimeras in higher mammals as presaged by successful murine experiments over the previous two decades (see below).
There is no confirmed example of de novo mammalian parthenogenetic reproduction, but mammalian oocytes can be artificially induced to undergo parthenogenesis in vitro by a two-step protocol involving electroporation and/or treatment with a chemical agent (ionomycin, ethanol, or inositol 1,4,5-triphosphate) to elevate Ca2+ levels transiently, followed by application of an inhibitor of protein synthesis (cycloheximide) or protein phosphorylation (6-dimethylaminopurine). Success rates and viability appear to be organism dependent. Mouse parthenotes are capable of developing beyond the post-implantation stage in vivo [66,67]; porcine parthenotes have developed up to post-activation day 29 (limb bud stage, past the early heart beating stage); rabbit parthenotes until day 10–11 [68]; primates (Callithrix jacchus) have only been shown to implant [69]. The reason for this arrested development is believed to be due to genetic imprinting. In normal zygotes maternal and paternal haploid genomes are epigenetically distinct, and both sets are required for successful development [70,71]. Indeed, unstable chromosome modifications in the form of DNA methylation or histone modification are distinctly different in human sperm, compared to eggs. Therefore each gamete carries unique patterns of gene expression into the embryo. Since all genetic material in parthenotes is of maternal origin, there is no paternal imprinting component and this prevents proper development of extraembryonic tissues whose expression is regulated by the male genome [72]. In most mammals – including primates – oocytes are arrested at metaphase II just before ovulation. Cytogenetic microscopy shows the presence of a 2n polar body under the zona pellucida and a 2n protonucleus in the cytoplasm. After chemical activation to mimic the effects of sperm penetration on changes in cellular Ca2+ gradient, the cell fails to complete meiosis II. Instead, the second polar body is never extruded, resulting in a diploid protonucleus derived from two sets of sister chromatids. These chromatids then begin to undergo mitosis resulting in a parthenote manifesting uniparental disomy. Although the derivation of embryonic-like stem cells from oocytes (parthenogenetic stem cells, PSC) is relatively inefficient (perhaps due to complexities of genomic imprinting), when they are differentiated into adult tissues, they appear fully functional.
In spite of non-viability of monkey parthenotes, the extracted stem cells seem to assume the morphology and functional behavior of HESC and express appropriate ESC markers. They have embryonic-like replicative ability and have been propagated in vitro in an undifferentiated state for up to 14 months. In vitro, they have been differentiated into cardiomyocyte-like cells, smooth muscle, beating ciliated epithelia, adipocytes, several types of epithelial cells, as well as dopaminergic and serotoninergic neurons. Almost all of these neurons express TUJ1 (beta-tubulin III), and up to 25% of the TUJ1+ cells co-express tyrosine-hydroxylase. This latter enzyme marker is considered diagnostic for catecholaminergic neurons (dopamine, norepinephrine, and epinephrine [73]). Furthermore, HPLC analysis of culture media following a depolarizing KCl-buffer identifies the release of the neurotransmitters dopamine and serotonin from the cells. Ater two weeks of differentiation, about half of the cells demonstrate neuronal morphology and begin to express voltage-dependent sodium channels that can be blocked by tetrodotoxin.
These observations are recapitulated in vivo, since injection of monkey PSC into immunocompromised mice induces formation of benign teratomas containing tissue derivatives from all three germ layers (ectoderm, endoderm and mesoderm) including cartilage, muscle, bone, neurons, skin, hair follicles, and intestinal epithelia [74,75]. Of particular note is the apparent tendency of these cells to differentiate into neuronal tissues, as has been noted by chimera studies [67]. The reasons for this underlying preference are not well understood although one possible explanation is that it is a consequence of purely maternal genomic imprinting, reflecting a lack of epigenetic balance that would be conferred by paternally-imprinted genes.
To be sure, parthenotes are not free from ethical controversy and are viewed by some in society as artificial entities that in some sense represent 'tampering with nature.' Since a parthenote is analogous to a mature ovarian teratoma (a spontaneous in vivo tumorigenic event) the de facto acceptance of experiments using teratoma tumor tissue lends some legitimacy to experimentation on parthenotes. These contradictions await reconciliation in a comprehensive ethical framework.
Stem cell genomics
The pluripotentiality of stem cells is also their limitation, and explains why they are not used clinically today. Although ESC can be differentiated into skin, neurons, blood, cardiac cells, cartilage, endothelial cells, muscle, hepatocytes, and pancreatic cells, the efficiency can be quite limited for certain cell types. Another difficulty is studying the quality of differentiation: are the neurons derived from stem cells bona fide neurons, or merely neuronal-like cells? To address this question we developed high throughput methodologies using microarrays to evaluate new stem cell derivatives [76]. We differentiated HESC into retinal pigmented epithelial cells (RPE) (the site of lesions in macular degeneration and retinitis pigmentosa) and used microarrays to identify their genetic signature. We then compared their gene signature to those derived from two established RPE cell lines (one of which has been successfully used clinically). A bronchial epithelial cell line served as a negative control and a freshly isolated human RPE served as a positive control. We demonstrated similarity between our HESC derived RPE and the freshly isolated RPE. The bronchial epithelial and two other established RPE lines were less similar. Interestingly, the data set that represented the genes common to freshly isolated RPE and HESC derived RPE (but not in the two established lines), contained many retinal specific genes. This finding provided further support of the benefits of HESC: the ability to generate a limitless number of HESC with the potential to differentiate along specific lineages to allow creation of RPE cells in quantities necessary for clinical transplantation. The next step would be to couple this technology to ESC derived from SCNT (or parthenogenesis) to create the ideal treatment for macular degeneration and retinitis pigmentosa.
Another technology currently under development at our institution is "genomics guided tissue engineering." Here we perform microarrays periodically during stem cell differentiation. For example, microarrays are performed on undifferentiated monkey PSC, PSC derived neural precursors (PSC-NP), and NP that were further differentiated for 8 days (PSC-neurons). We have identified numerous targets such as receptors and ligands present at each of these distinct time points, and are modifying our culture system in order to improve the quality and quantity of differentiation. Furthermore, we are comparing the gene expression profiles of PSC derived neurons to gene expression profiles of reference neurons. Not only will this provide new insight into the type of neurons that may be generated, but it offers clues into what our stem cell derived neurons might be lacking. We can then go back to the culture system and try to target these specific genes/signaling pathways.
Further study of stem cell genomics will give additional insight into pluripotentiality. An understanding of pluripotentiality might allow for a somatic cell to be de-differentiated into an intermediate stage, which could then be expanded, differentiated and transplanted back into the patient. We are presently characterizing the genetic signature of pluripoteniality by analyzing gene expression among primate stem cells derived from a variety of methods (IVF, parthenogenesis, and adult stem cells). By identifying "stemness" genes by comparing undifferentiated stem cells to their differentiated counterpart, and comparing this to stem cells of different origins, a core set of pluripotential target genes may be mapped. Of particular interest are the 1,075 genes that are similarly down-regulated in IVF derived human ESC and monkey PSC. Furthermore, we have detected paternally imprinted genes in our HESC but not in our PSC data sets. From this we conclude that paternal imprinting might not be necessary for pluriopotentiality.
Conclusion
Our systems biology approach incorporates the fields of genomics, cell biology, nuclear transfer, and materials science, and utilizes personnel who have mastered the techniques of bioinformatics, cell harvest, culture, expansion, transplantation, as well as polymer design essential for the successful application of these technologies. Experimental efforts are currently underway involving virtually every type of tissue and organ of the human body. Various tissues are at different stages of development with some already being used clinically, a few in pre-clinical trials, and some in the discovery stage. Recent progress suggests that engineered tissues may have an expanded clinical applicability in the future and may represent a viable therapeutic option for those who require tissue replacement or repair.
==== Refs
Cilento BG Freeman MR Schneck FX Retik AB Atala A Phenotypic and cytogenetic characterization of human bladder urothelia expanded in vitro J Urol 1994 152 665 670 7517459
Scriven SD Booth C Thomas DF Trejdosiewicz LK Southgate J Reconstitution of human urothelium from monolayer cultures J Urol 1997 158 1147 1152 9258159 10.1097/00005392-199709000-00115
Liebert M Hubbel A Chung M Wedemeyer G Lomax MI Hegeman A Yuan TY Brozovich M Wheelock MJ Grossman HB Expression of mal is associated with urothelial differentiation in vitro: identification by differential display reverse-transcriptase polymerase chain reaction Differentiation 1997 61 177 185 9084136 10.1046/j.1432-0436.1997.6130177.x
Puthenveettil JA Burger MS Reznikoff CA Replicative senescence in human uroepithelial cells Adv Exp Med Biol 1999 462 83 91 10599415
Freeman MR Yoo JJ Raab G Soker S Adam RM Schneck FX Renshaw AA Klagsbrun M Atala A Heparin-binding EGF-like growth factor is an autocrine growth factor for human urothelial cells and is synthesized by epithelial and smooth muscle cells in the human bladder J Clin Invest 1997 99 1028 1036 9062361
Nguyen HT Park JM Peters CA Adam RM Orsola A Atala A Freeman MR Cell-specific activation of the HB-EGF and ErbB1 genes by stretch in primary human bladder cells In Vitro Cell Dev Biol Anim 1999 35 371 375 10462199
Liebert M Wedemeyer G Abruzzo LV Kunkel SL Hammerberg C Cooper KD Grossman HB Stimulated urothelial cells produce cytokines and express an activated cell surface antigenic phenotype Semin Urol 1991 9 124 130 1853009
Harriss DR Smooth muscle cell culture: a new approach to the study of human detrusor physiology and pathophysiology Br J Urol 1995 75 Suppl 1 18 26 7613854
Solomon LZ Jennings AM Sharpe P Cooper AJ Malone PS Effects of short-chain fatty acids on primary urothelial cells in culture: implications for intravesical use in enterocystoplasties J Lab Clin Med 1998 132 279 283 9794698 10.1016/S0022-2143(98)90040-3
Lobban ED Smith BA Hall GD Harnden P Roberts P Selby PJ Trejdosiewicz LK Southgate J Uroplakin gene expression by normal and neoplastic human urothelium Am J Pathol 1998 153 1957 1967 9846985
Rackley RR Bandyopadhyay SK Fazeli-Matin S Shin MS Appell R Immunoregulatory potential of urothelium: characterization of NF-kappaB signal transduction J Urol 1999 162 1812 1816 10524940 10.1097/00005392-199911000-00075
Thomson JA Itskovitz-Eldor J Shapiro SS Waknitz MA Swiergiel JJ Marshall VS Jones JM Embryonic stem cell lines derived from human blastocysts Science 1998 282 1145 1147 9804556 10.1126/science.282.5391.1145
Solter D Knowles BB Immunosurgery of mouse blastocyst Proc Natl Acad Sci U S A 1975 72 5099 5102 1108013
Hoffman DI Zellman GL Fair CC Mayer JF Zeitz JG Gibbons WE Turner JTG Cryopreserved embryos in the United States and their availability for research Fertility and Sterility 2003 79 1063 1069 12738496 10.1016/S0015-0282(03)00172-9
Lanzendorf SE Boyd CA Wright DL Muasher S Oehninger S Hodgen GD Use of human gametes obtained from anonymous donors for the production of human embryonic stem cell lines Fertil Steril 2001 76 132 137 11438331 10.1016/S0015-0282(01)01825-8
Briggs RKTJ Transplantation of living nuclei from blastula cells into enucleated frogs' eggs Proceedings of the National Academy of Sciences of the United States of America 1952 38 455 463 16589125
Gurdon JB Adult frogs derived from the nuclei of single somatic cells Dev Biol 1962 4 256 273 13903027 10.1016/0012-1606(62)90043-X
Campbell KH McWhir J Ritchie WA Wilmut I Sheep cloned by nuclear transfer from a cultured cell line Nature 1996 380 64 66 8598906 10.1038/380064a0
Wilmut I Schnieke AE McWhir J Kind AJ Campbell KH Viable offspring derived from fetal and adult mammalian cells Nature 1997 385 810 813 9039911 10.1038/385810a0
Cibelli JB Stice SL Golueke PJ Kane JJ Jerry J Blackwell C Ponce de Leon FA Robl JM Cloned transgenic calves produced from nonquiescent fetal fibroblasts Science 1998 280 1256 1258 9596577 10.1126/science.280.5367.1256
Baguisi A Behboodi E Melican DT Pollock JS Destrempes MM Cammuso C Williams JL Nims SD Porter CA Midura P Palacios MJ Ayres SL Denniston RS Hayes ML Ziomek CA Meade HM Godke RA Gavin WG Overstrom EW Echelard Y Production of goats by somatic cell nuclear transfer Nat Biotechnol 1999 17 456 461 10331804 10.1038/8632
Keefer CL Keyston R Lazaris A Bhatia B Begin I Bilodeau AS Zhou FJ Kafidi N Wang B Baldassarre H Karatzas CN Production of cloned goats after nuclear transfer using adult somatic cells Biol Reprod 2002 66 199 203 11751283
Wakayama T Perry AC Zuccotti M Johnson KR Yanagimachi R Full-term development of mice from enucleated oocytes injected with cumulus cell nuclei Nature 1998 394 369 374 9690471 10.1038/28615
Betthauser J Forsberg E Augenstein M Childs L Eilertsen K Enos J Forsythe T Golueke P Jurgella G Koppang R Lesmeister T Mallon K Mell G Misica P Pace M Pfister-Genskow M Strelchenko N Voelker G Watt S Thompson S Bishop M Production of cloned pigs from in vitro systems Nat Biotechnol 2000 18 1055 1059 11017042 10.1038/80242
Polejaeva IA Chen SH Vaught TD Page RL Mullins J Ball S Dai Y Boone J Walker S Ayares DL Colman A Campbell KH Cloned pigs produced by nuclear transfer from adult somatic cells Nature 2000 407 86 90 10993078 10.1038/35024082
Onishi A Iwamoto M Akita T Mikawa S Takeda K Awata T Hanada H Perry AC Pig cloning by microinjection of fetal fibroblast nuclei Science 2000 289 1188 1190 10947985 10.1126/science.289.5482.1188
De Sousa PA Dobrinsky JR Zhu J Archibald AL Ainslie A Bosma W Bowering J Bracken J Ferrier PM Fletcher J Gasparrini B Harkness L Johnston P Ritchie M Ritchie WA Travers A Albertini D Dinnyes A King TJ Wilmut I Somatic cell nuclear transfer in the pig: control of pronuclear formation and integration with improved methods for activation and maintenance of pregnancy Biol Reprod 2002 66 642 650 11870070
Colman A Kind A Therapeutic cloning: concepts and practicalities Trends Biotechnol 2000 18 192 196 10758513 10.1016/S0167-7799(00)01434-7
Vogelstein B Alberts B Shine K Genetics. Please don't call it cloning! Science 2002 295 1237 11847324 10.1126/science.1070247
Hochedlinger K Jaenisch R Nuclear transplantation, embryonic stem cells, and the potential for cell therapy N Engl J Med 2003 349 275 286 12867612 10.1056/NEJMra035397
Lanza RP Cibelli JB West MD Dorff E Tauer C Green RM The ethical reasons for stem cell research Science 2001 292 1299 11360981 10.1126/science.292.5520.1299b
Lanza RP Cibelli JB West MD Prospects for the use of nuclear transfer in human transplantation Nat Biotechnol 1999 17 1171 1174 10585712 10.1038/70709
Solter D Mammalian cloning: advances and limitations Nat Rev Genet 2000 1 199 207 11252749 10.1038/35042066
Rideout WM Eggan K Jaenisch R Nuclear cloning and epigenetic reprogramming of the genome Science 2001 293 1093 1098 11498580 10.1126/science.1063206
Hochedlinger K Jaenisch R Nuclear transplantation: lessons from frogs and mice Curr Opin Cell Biol 2002 14 741 748 12473349 10.1016/S0955-0674(02)00380-0
Tsunoda Y Kato Y Recent progress and problems in animal cloning Differentiation 2002 69 158 161 11841470 10.1046/j.1432-0436.2002.690405.x
Kato Y Tani T Sotomaru Y Kurokawa K Kato J Doguchi H Yasue H Tsunoda Y Eight calves cloned from somatic cells of a single adult Science 1998 282 2095 2098 9851933 10.1126/science.282.5396.2095
Dinnyes A De Sousa P King T Wilmut I Somatic cell nuclear transfer: recent progress and challenges Cloning Stem Cells 2002 4 81 90 12006159 10.1089/153623002753632075
Young LE Sinclair KD Wilmut I Large offspring syndrome in cattle and sheep Rev Reprod 1998 3 155 163 9829550 10.1530/ror.0.0030155
Cibelli JB Campbell KH Seidel GE West MD Lanza RP The health profile of cloned animals Nat Biotechnol 2002 20 13 14 11753346 10.1038/nbt0102-13
Tamashiro KL Wakayama T Akutsu H Yamazaki Y Lachey JL Wortman MD Seeley RJ D'Alessio DA Woods SC Yanagimachi R Sakai RR Cloned mice have an obese phenotype not transmitted to their offspring Nat Med 2002 8 262 267 11875497 10.1038/nm0302-262
Ogonuki N Inoue K Yamamoto Y Noguchi Y Tanemura K Suzuki O Nakayama H Doi K Ohtomo Y Satoh M Nishida A Ogura A Early death of mice cloned from somatic cells Nat Genet 2002 30 253 254 11836501 10.1038/ng841
Bortvin A Eggan K Skaletsky H Akutsu H Berry DL Yanagimachi R Page DC Jaenisch R Incomplete reactivation of Oct4-related genes in mouse embryos cloned from somatic nuclei Development 2003 130 1673 1680 12620990 10.1242/dev.00366
Boiani M Eckardt S Scholer HR McLaughlin KJ Oct4 distribution and level in mouse clones: consequences for pluripotency Genes Dev 2002 16 1209 1219 12023300 10.1101/gad.966002
Lanza RP Chung HY Yoo JJ Wettstein PJ Blackwell C Borson N Hofmeister E Schuch G Soker S Moraes CT West MD Atala A Generation of histocompatible tissues using nuclear transplantation Nat Biotechnol 2002 20 689 696 12089553 10.1038/nbt703
Amiel GE Atala A Current and future modalities for functional renal replacement Urol Clin North Am 1999 26 235 46, xi 10086063
Auchincloss H Bonventre JV Transplanting cloned cells into therapeutic promise Nat Biotechnol 2002 20 665 666 12089546 10.1038/nbt0702-665
Aebischer P Ip TK Panol G Galletti PM The bioartificial kidney: progress towards an ultrafiltration device with renal epithelial cells processing Life Support Syst 1987 5 159 168 3669723
Amiel GE Yoo JJ Atala A Renal therapy using tissue-engineered constructs and gene delivery World J Urol 2000 18 71 79 10766048 10.1007/s003450050013
Humes HD Buffington DA MacKay SM Funke AJ Weitzel WF Replacement of renal function in uremic animals with a tissue-engineered kidney Nat Biotechnol 1999 17 451 455 10331803 10.1038/8626
Ip TK Aebischer P Galletti PM Cellular control of membrane permeability. Implications for a bioartificial renal tubule ASAIO Trans 1988 34 351 355 3196532
Joki T Machluf M Atala A Zhu J Seyfried NT Dunn IF Abe T Carroll RS Black PM Continuous release of endostatin from microencapsulated engineered cells for tumor therapy Nat Biotechnol 2001 19 35 39 11135549 10.1038/83481
Lanza RP Hayes JL Chick WL Encapsulated cell technology Nat Biotechnol 1996 14 1107 1111 9631060 10.1038/nbt0996-1107
MacKay SM Funke AJ Buffington DA Humes HD Tissue engineering of a bioartificial renal tubule Asaio J 1998 44 179 183 9617948
Evans MJ Gurer C Loike JD Wilmut I Schnieke AE Schon EA Mitochondrial DNA genotypes in nuclear transfer-derived cloned sheep Nat Genet 1999 23 90 93 10471506 10.1038/12696
Hiendleder S Schmutz SM Erhardt G Green RD Plante Y Transmitochondrial differences and varying levels of heteroplasmy in nuclear transfer cloned cattle Mol Reprod Dev 1999 54 24 31 10423294 10.1002/(SICI)1098-2795(199909)54:1<24::AID-MRD4>3.0.CO;2-S
Steinborn R Schinogl P Zakhartchenko V Achmann R Schernthaner W Stojkovic M Wolf E Muller M Brem G Mitochondrial DNA heteroplasmy in cloned cattle produced by fetal and adult cell cloning Nat Genet 2000 25 255 257 10888867 10.1038/77000
Fischer Lindahl K Hermel E Loveland BE Wang CR Maternally transmitted antigen of mice: a model transplantation antigen Annu Rev Immunol 1991 9 351 372 1910682 10.1146/annurev.iy.09.040191.002031
Hadley GA Linders B Mohanakumar T Immunogenicity of MHC class I alloantigens expressed on parenchymal cells in the human kidney Transplantation 1992 54 537 542 1412735
Yard BA Kooymans-Couthino M Reterink T van den Elsen P Paape ME Bruyn JA van Es LA Daha MR van der Woude FJ Analysis of T cell lines from rejecting renal allografts Kidney Int Suppl 1993 39 S133 8 8468915
Lanza R Moore MA Wakayama T Perry AC Shieh JH Hendrikx J Leri A Chimenti S Monsen A Nurzynska D West MD Kajstura J Anversa P Regeneration of the infarcted heart with stem cells derived by nuclear transplantation Circ Res 2004 94 820 827 14764454 10.1161/01.RES.0000120863.53562.DF
Barberi T Klivenyi P Calingasan NY Lee H Kawamata H Loonam K Perrier AL Bruses J Rubio ME Topf N Tabar V Harrison NL Beal MF Moore MA Studer L Neural subtype specification of fertilization and nuclear transfer embryonic stem cells and application in parkinsonian mice Nat Biotechnol 2003 21 1200 1207 14502203 10.1038/nbt870
Hwang WS Ryu YJ Park JH Park ES Lee EG Koo JM Jeon HY Lee BC Kang SK Kim SJ Ahn C Hwang JH Park KY Cibelli JB Moon SY Evidence of a pluripotent human embryonic stem cell line derived from a cloned blastocyst Science 2004 303 1669 1674 14963337 10.1126/science.1094515
Plachot M Mandelbaum J de Grouchy J [In vitro parthenogenesis in the human species] Ann Genet 1984 27 158 161 6334479
Strain L Warner JP Johnston T Bonthron DT A human parthenogenetic chimaera Nat Genet 1995 11 164 169 7550344 10.1038/ng1095-164
Kaufman MH Barton SC Surani MA Normal postimplantation development of mouse parthenogenetic embryos to the forelimb bud stage Nature 1977 265 53 55 556803
Surani MA Barton SC Norris ML Development of reconstituted mouse eggs suggests imprinting of the genome during gametogenesis Nature 1984 308 548 550 6709062 10.1038/308548a0
Ozil JP The parthenogenetic development of rabbit oocytes after repetitive pulsatile electrical stimulation Development 1990 109 117 127 2209460
Marshall VS Wilton LJ Moore HD Parthenogenetic activation of marmoset (Callithrix jacchus) oocytes and the development of marmoset parthenogenones in vitro and in vivo Biol Reprod 1998 59 1491 1497 9828197
Surani MA Kothary R Allen ND Singh PB Fundele R Ferguson-Smith AC Barton SC Genome imprinting and development in the mouse Dev Suppl 1990 89 98 2090435
Sasaki H Jones PA Chaillet JR Ferguson-Smith AC Barton SC Reik W Surani MA Parental imprinting: potentially active chromatin of the repressed maternal allele of the mouse insulin-like growth factor II (Igf2) gene Genes Dev 1992 6 1843 1856 1383088
Surani MA Barton SC Development of gynogenetic eggs in the mouse: implications for parthenogenetic embryos Science 1983 222 1034 1036 6648518
Kumer SC Vrana KE Intricate regulation of tyrosine hydroxylase activity and gene expression J Neurochem 1996 67 443 462 8764568
Cibelli JB Grant KA Chapman KB Cunniff K Worst T Green HL Walker SJ Gutin PH Vilner L Tabar V Dominko T Kane J Wettstein PJ Lanza RP Studer L Vrana KE West MD Parthenogenetic stem cells in nonhuman primates Science 2002 295 819 11823632 10.1126/science.1065637
Vrana KE Hipp JD Goss AM McCool BA Riddle DR Walker SJ Wettstein PJ Studer LP Tabar V Cunniff K Chapman K Vilner L West MD Grant KA Cibelli JB Nonhuman primate parthenogenetic stem cells Proc Natl Acad Sci U S A 2003 100 Suppl 1 11911 11916 14504386 10.1073/pnas.2034195100
Klimanskaya IHJRKAWMAALR Derivation and Comparative Assessment of Retinal Pigment Epithelium from Human Embryonic Stem Cells Using Transcriptomics Cloning Stem Cells 2004 6 217 245 15671670
| 15588286 | PMC539246 | CC BY | 2021-01-04 16:39:57 | no | J Exp Clin Assist Reprod. 2004 Dec 8; 1:3 | utf-8 | J Exp Clin Assist Reprod | 2,004 | 10.1186/1743-1050-1-3 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-4-571558829510.1186/1471-2334-4-57Case ReportThe unmasking of Pneumocystis jiroveci pneumonia during reversal of immunosuppression: case reports and literature review Wu Alan KL [email protected] Vincent CC [email protected] Bone SF [email protected] Ivan FN [email protected] Rodney A [email protected] David S [email protected] Kwok Y [email protected] Division of Infectious Diseases, Centre of Infection, Queen Mary Hospital, The University of Hong Kong; Hong Kong Special Administrative Region, China2 Division of Pulmonary Medicine, Department of Medicine and Therapeutics, Prince of Wales Hospital, The Chinese University of Hong Kong; Hong Kong Special Administrative Region, China2004 9 12 2004 4 57 57 1 9 2004 9 12 2004 Copyright © 2004 Wu 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
Pneumocystis jiroveci pneumonia (PCP) is an important opportunistic infection among immunosuppressed patients, especially in those infected with human immunodeficiency virus (HIV). The clinical presentation of PCP in immunosuppressed patients have been well-reported in the literature. However, the clinical importance of PCP manifesting in the setting of an immunorestitution disease (IRD), defined as an acute symptomatic or paradoxical deterioration of a (presumably) preexisting infection, which is temporally related to the recovery of the immune system and is due to immunopathological damage associated with the reversal of immunosuppressive processes, has received relatively little attention until recently.
Case presentation
We aim to better define this unique clinical syndrome by reporting two cases of PCP manifesting acutely with respiratory failure during reversal of immunosuppression in non-HIV infected patients, and reviewed the relevant literature. We searched our databases for PCP cases manifesting in the context of IRD according to our predefined case definition, and reviewed the case notes retrospectively. A comprehensive search was performed using the Medline database of the National Library of Medicine for similar cases reported previously in the English literature in October 2003. A total of 28 non-HIV (excluding our present case) and 13 HIV-positive patients with PCP manifesting as immunorestitution disease (IRD) have been reported previously in the literature. During immunorestitution, a consistent rise in the median CD4 lymphocyte count (28/μL to 125/μL), with a concomitant fall in the median HIV viral load (5.5 log10 copies/ml to 3.1 log10 copies/ml) was observed in HIV-positive patients who developed PCP. A similar upsurge in peripheral lymphocyte count was observed in our patients preceding the development of PCP, as well as in other non-HIV immunosuppressed patients reported in the literature.
Conclusions
PCP manifesting as IRD may be more common than is generally appreciated. Serial monitoring of total lymphocyte or CD4 count could serve as a useful adjunct to facilitate the early diagnosis and pre-emptive treatment of this condition in a wide range of immunosuppressed hosts, especially in the presence of new pulmonary symptoms and/or radiographic abnormalities compatible with the diagnosis.
==== Body
Background
Pneumocystis jiroveci (Pj) (previously known as Pneumocystis carinii f. sp. hominis) was first identified as a pathogen in premature infants suffering from interstitial plasma cell pneumonia in European countries during and after World War II, occasionally occurring in epidemics [1-3]. Since then Pneumocystis pneumonia (PCP) had only been reported sporadically in patients with malignancies and solid organ transplantations until the HIV epidemic [4]. The incidence of PCP increased significantly after the emergence of human immunodeficiency virus (HIV) infection. However, with the identification of CD4 T lymphocyte depletion as an independent risk factor for the development of PCP [5], widespread use of antimicrobial prophylaxis [4], and the introduction of highly active antiretroviral therapy (HAART), there has been a steady decline in the incidence of PCP among HIV-infected patients [7,8].
Nevertheless, with the rising number of patients receiving immunosuppressive therapies for malignancies, solid organ transplantations and autoimmune diseases, PCP has been increasingly recognized in non-HIV immunosuppressed hosts [9-15]. For instance, PCP occurs in 3.4% to 43% of solid organ transplant recipients [16], and it is particularly prevalent among those patients who are put on long-term steroids. In a non-HIV immunosuppressed cohort with PCP, the use of steroids was found to be a contributing factor in 87% of patients [17]. In another similar cohort of immunosuppressed patients, steroids had been administered systemically in 90.5% within one month before the diagnosis of PCP. Although a median daily dose equivalent to 30 mg of prednisone was administered in most of these patients prior to the development of PCP, up to 25% had received as little as 16 mg of prednisone daily [18]. Interestingly, PCP has also been reported in patients with endogenous steroid excess due to Cushing's disease [19,20].
Paradoxically, the clinical symptoms of PCP were often unmasked in HIV-negative immunosuppressed patients during the reversal of immunosuppression, often at the time when the dose of steroids was tapered [11,17,21-24], or when the endogenous steroid production was reduced [25-27]. However, serial changes in the absolute lymphocyte count before and during reversal of immunosuppression were not mentioned in these patients. Recently, paradoxical worsening of clinical symptoms and signs of PCP after initiation of HAART has also been reported in HIV-positive patients [28-31]. The onset of clinical deterioration was associated with an upsurge in the CD4 lymphocyte count and a reduction in the HIV viral load [28-31]. Tissue damage is thought to occur as a result of immune reconstitution in HIV-positive patients. Here, we report two cases and review the literature on this topic from the perspective of immunorestitution disease.
Case presentation
Case 1
This is a fifty-one year old female patient with history of diabetes mellitus and systemic lupus erythematoses (SLE) complicated by lupus nephritis. Although we have included her case in our previous publication [32], we have not reported her clinical details at that time. She was put on prednisolone 30 mg and azathioprine 100 mg daily since end of June and mid-July 2002, respectively. She was admitted to Queen Mary Hospital on 11th August 2002 for investigation of jaundice. Investigations revealed deranged liver function tests with cholestatic pattern. A diagnosis of drug-induced hepatotoxicity was entertained, and azathioprine was stopped after admission. As her autoimmune disease was under control, her steroid dosage was reduced from 25 mg to 15 mg daily within the next 14 days. Her CXR taken on admission was normal.
Soon after her immunosuppressive therapy was tapered, she developed fever and non-productive cough. A repeat CXR performed on 9th Sept revealed new infiltrate over the left mid-zone, suggestive of pneumonia. She was started on intravenous ceftazidime 1 gram eight hourly and oral clarithromycin 500 mg twice daily. Serial CXR performed three days later showed increasing bilateral pulmonary infiltrates and worsening hypoxemia. There was an upsurge of total lymphocyte count from 0.7 × 109/L (total white cell count 7.2 × 109/L) at the time of admission to 5.6 × 109/L (total white cell count 10.8 × 109/L) at the time of clinical deterioration. Bronchoscopy with transbronchial biopsy performed on the same day revealed Pneumocystitis jiroveci by methenamine sliver stain. Workup for other opportunistic pathogens including cytomegalovirus and aspergillus was negative. She was commenced on intravenous pentamidine (4 mg/kg/day) and corticosteroids for severe PCP infection. Despite active treatment she developed progressive respiratory failure and required admission to intensive care unit. She subsequently recovered after a stormy hospital course, and upon discharge from hospital, her total lymphocyte count had returned to her baseline of 0.86 × 109/L.
Case 2
A thirty-three year old gentleman initially presented to Prince of Wales Hospital with a diagnosis of SLE/dermatomyositis overlap syndrome. He was treated with steroid and hydroxychloroquine 200 mg twice daily since 1997. He had a flare up of disease in May 1998 with active vasculitis and myositis, for which he was put on prednisolone and azthioprine 50 mg and 100 mg daily respectively. Upon reassessment one month later, disease activity was under control, and the dosage of prednisolone was reduced to 45 mg daily.
Twelve day after reducing the immunosuppressive regimen, he was admitted to hospital for treatment of left buttock abscess. The CXR taken on admission was unremarkable. An aspirate of the pus from the lesion grew methicillin-sensitive staphylococcus aureus; he was treated with cloxacillin 1 g intravenously every 6 hourly, together with incision and drainage of the buttock abscess. In view of the underlying active pyogenic infection, the steroid dosage was rapidly tapered from 45 mg to 15 mg daily within the next four days. However, he was noted to have persistent fever associated with mild unproductive cough. A repeat chest radiograph showed new infiltrates over the right upper and left lower zones, and he was empirically treated with intravenous ceftazidime 1 gram every 8 hours, cloxacillin 1 gram every 6 hours and netimicin 100 mg every 8 hours. As there was no clinical response after 5 days of treatment, bronchoscopy and bronchoalveolar lavage (BAL) was performed, which was positive for Pneumocystis jiroveci. Investigation for the presence of co-existing opportunistic pathogens such as cytomegalovirus and aspergillus was negative. On the day after bronchoscopy, he was commenced on intravenous cotrimoxazole 1.3 grams every 6 hours. He remained stable initially with fever on downward trend. However, on the 3rd day of treatment, he developed sudden desaturation with resurgence of high fever, and required supplemental oxygen therapy. Repeat chest radiograph showed increased perihilar hazziness in both lung fields. There was also an upsurge of total lymphocyte count from 0.6 × 109/L (total white cell count 11.2 × 109/L) on admission, to 1.3 × 109/L (total white cell count 10.4 × 109/L) at the time of clinical deterioration. He was treated with high dose prednisolone (80 mg daily), and his condition improved promptly afterwards. He was subsequently discharged, and on follow up at the clinic one month later, his total lymphocyte count had returned to his baseline level of 0.6 × 109/L.
Immunorestitution disease (IRD) has been described in both HIV and non-HIV immunosuppressed hosts previously [27-31]. In the setting of PCP, it is defined as an acute symptomatic presentation of the disease that is related temporally to the recovery of the immune system, associated with reversal of immunosuppressive processes such as reduction in the dosage of corticosteroids and/or cytotoxic agents or a reduction of HIV viral load due to HAART, which results in the development of immunopathological damage. The preexisting microbial infection could be either asymptomatic or mildly symptomatic. Using this case definition, we attempted to review the English literature for other reported cases of PCP manifesting as IRD. The English-language literature (1966 – 2003) was searched in the Medline database of the National Library of Medicine in October 2003. The keywords "Pneumocystis carinii", "Pneumocystis jiroveci", "HIV", immunosuppression", "immunosuppressive", "steroid", and "corticosteroid" were used to select cases. All the case reports and case series with clinical details were included in this study if they fulfilled the above definition of IRD. When appropriate, the cited bibliographies were also retrieved for further analysis. As for statistical analysis, we used the Wilcoxon Signed Rank test, a non-parametric test for comparing paired samples, to analyze the serial changes in lymphocyte counts and HIV viral loads before and during the development of IRD. A two-tailed p-value of less than 0.05 was considered significant. All statistical analyses were performed using SPSS version 11.5 for Windows.
Including our present case, a total of 29 cases of PCP in non-HIV immunosuppressed hosts fulfilling our definition of IRD have been reported in the literature (table 1) [22-27,32]. There were altogether 13 males and 8 females, with a median age of 38 years (range 2 to 75 years). The age and sex were not mentioned in 8 cases. Their underlying immunosuppressive conditions included solid organ tumours (13 cases), haematological diseases (8), autoimmune diseases (4), endogenous Cushing's disease (3), and a solid organ transplant recipient (1). All patients had received steroids or had excessive endogenous steroid production, whereas 18 (62.1%) of them had concomitant cytotoxic therapy for the underlying diseases. The median duration between steroid tapering and clinical manifestations of PCP was 21 days (range 1 to 83 days). Steroids were completely withdrawn at a median of 7.5 days (range 1 to 21 days) before the onset of symptoms in eight patients. Serial lymphocyte counts were only available in eight patients. An upsurge of the absolute lymphocyte counts was observed from the time of reduction of immunosuppression (median 300/μL, range 290 to 600/μL at baseline) to the time of occurrence of IRD (median 1200/μL, range 600 to 5620/μL); the median increase in total lymphocyte count was 800/μL, with a range of 300 to 4880/μL. Comparing the lymphocyte counts before and after reversal of immunosuppressive therapy, the difference was statistically significant (Wilcoxon Signed Rank Test for paired samples; p = 0.012). In addition to our patient, reintroduction or increasing doses of steroids were required in 7 (53.8%) of 13 patients in the acute management of PCP in the literature, at the time when they developed clinical deterioration during antimicrobial therapy [24-27]. Seven (53.8%) of 13 cases had respiratory failure requiring mechanical ventilation. Among these 29 cases, 13 (44.8%) subsequently died of PCP.
Among HIV-positive patients, 13 cases with newly diagnosed PCP were reported in the literature, in which IRD occurred shortly after the introduction of HAART (table 2) [28-31]. Seven (53.8%) out of 13 cases received steroids as adjunctive therapy in addition to antimicrobials. HAART was given in all cases at a median 18 days (range 1 to 35 days) after the initiation of treatment for PCP. During IRD, recurrence of fever (100%), dyspnoea (100%), and paradoxical worsening of pulmonary infiltrates (58.3%) were observed in these patients [28-31]. IRD occurred at a median of 14 days (range 5 to 17 days) after HAART. An upsurge of the CD4 lymphocyte count was observed before (median 28/μL, range 4 to 290/μL) and during IRD (median 125/μL, range 30 to 564/μL); this was associated with a concomitant reduction of the median HIV viral load from 5.5 log10 copies/ml (range 5.0 to 5.9 log10 copies/ml) to 3.1 log10 copies/ml (range 2.9 to 4.5 log10 copies/ml) before and during IRD respectively, and the differences observed in both the CD4 counts and viral loads before and during IRD reached statistical significance (Wilcoxon Signed Rank Test for related samples; p = 0.001 and 0.017, respectively). Antimicrobials, steroids, or both for PCP were reintroduced for IRD in 4, 1, and 6 cases respectively. Only 2 cases were treated conservatively. One case required mechanical ventilation for severe respiratory distress. None of the patients died.
PCP manifesting as a form of IRD is not a rare phenomenon. As shown in our previous study, it happens in 7 out of 10 (70%) of HIV-negative immunosuppressed hosts infected with Pj [32]. However, the diagnosis of PCP is usually delayed in this group of patients because of atypical presentation. In this clinical setting, PCP manifesting as IRD often runs an acute and fulminant course, with nonspecific lesions on chest radiographs, and high absolute lymphocyte counts [32]. In our own reported series, despite the administration of steroid therapy to suppress the immunopathological damage, more than 80% of patients developed acute respiratory failure and required mechanical ventilation. Patients who developed PCP during reversal of immunosuppressive therapy in our series tended to be older, and this might partially explain the increased mortality observed in this group [32].
Rapid reduction of immunosuppressive therapy such as steroids has been implicated as a predisposing factor for the development of PCP in HIV-negative patients [11,17,23,24]. In one study, PCP occurred in 79 (70%) of 113 patients during steroid tapering [17]. Another study suggested that 8 (72.2%) out of 11 episodes of PCP developed when steroid therapy was tapered [23]. A subsequent study also demonstrated that 43% of patients had a rapid reduction of steroid dosing before the clinical manifestations of PCP [11]. A similar experience was reported in children, and 17 (89.5%) out of 19 children were diagnosed to have PCP during steroid tapering according to a previous report [21]. Another series revealed that 7 of 11 patients experienced acutely symptomatic PCP when the dose of steroids was decreased or terminated 5 days to 3 weeks before the diagnosis of PCP [22]. However, all these cases were not analyzed from a perspective of IRD. Serial changes of the absolute lymphocyte counts or their subsets were either not noted or reported [11,17,21-24]. Hence we have not included these cases for further analysis in this review.
Among HIV positive patients, PCP manifesting acutely during the initiation anti-retroviral therapy is a well-recognized phenomenon. The underlying immunopathological nature of this condition, which is reminiscent to IRD occurring in non-HIV infected patients, has been confirmed by histological examination of the lungs and transbronchial biopsy specimens, which demonstrated mixed inflammatory infiltrates including macrophages, neutrophils, lymphocytes, and plasma cells. Almost all infiltrating lymphocytes found in the tissues were of the T cell lineage, shown by immunophenotyping to be predominantly CD4 and CD8 cells [28]. In another study [30], the BAL fluid obtained from one of six patients with an IRD-type presentation of PCP was analyzed. Infiltration of predominantly CD4 and CD8 lymphocytes with the proliferative marker (Ki67) and perforin-positive cell were seen in the BAL specimen. Therefore, it is likely that the phagocytosed Pj is presented by alveolar macrophage to T cells, which trigger the inflammatory response [30].
In our own experience, as well as from the review of published literature, it appears that a surge of absolute lymphocyte count, especially the CD4 lymphocyte count in HIV-positive patients, could potentially act as a surrogate marker for immunopathological damage during IRD in both HIV-negative and HIV-positive patients. In our recent publication [32], 7 out of 10 non-HIV immunosuppressed patients demonstrated a consistent rise in the absolute lymphocyte count during tapering of immunosuppression prior to the onset of symptomatic PCP. In this group of patients, the surge in lymphocyte count is likely the result of withdrawal of lymphocytotoxic immunosuppressants such as corticosteroids. Similarly, a rising trend of the CD4 lymphocyte count, consistent with immune reconstitution after HAART, was also observed in 13 HIV-positive cases before and during the development of symptomatic PCP [28-31]. In fact, an upsurge in the absolute lymphocyte count has been shown to be a marker of IRD in our previous publications involving viral and tuberculous infections [33-36]. However, it must be emphasized that the number of circulating lymphocytes may not always correlate with their number in the affected tissues or their in vivo functional activity. This can be exemplified by a case of PCP occurring during steroid withdrawal, in which the lymphocyte counts surged to a very high level and then rapidly dropped to a low level within one day. The migration of lymphocytes from the circulation to tissue might explain this rapid drop in lymphocyte count and the resulting immunopathological damage [27]. In the future, further studies on lymphocyte subsets and cytokine profiles of susceptible hosts during the development of IRD should be performed to elucidate the underlying immunopathological mechanisms behind this interesting phenomenon.
From the result of this review, it appears that HIV-positive patients with PCP are at risk of clinical deterioration due to IRD if HAART therapy is started within 1 to 2 weeks after the initiation of treatment for PCP (table 2). With a better understanding of the pathogenetic mechanisms resulting in IRD, we may be able to prevent the occurrence of IRD by delaying the initiation of HAART in HIV-positive patients with PCP. However, in non-HIV immunosuppressed patients, it is even more important to recognize the atypical presentations of PCP in the context of IRD. Since the clinical and/or radiological features alone may not be sufficient for diagnosis, analysis of serial changes in lymphocyte counts in patients undergoing a reduction of immunosuppression can alert the clinician to the possibility of IRD due to occult pathogens such as Pj. To prevent IRD in non-HIV immunosuppressed patients, the use of prophylactic antibiotics against Pj to reduce the microbial load in selected patients remains an important issue. Recently, a multi-center study showed that the CD4 lymphocyte count may be a useful marker to monitor the risk of development of PCP in non-HIV immunosuppressed hosts [37], and patients with low CD4 lymphocyte counts of less than 300 or 400 may require prophylaxis. In fact, asymptomatic colonization of Pj has been demonstrated in HIV-negative patients when the CD4 lymphocyte count was less than 400 [38]. Nested polymerase chain reaction (PCR) identified a significant percentage of clinically silent Pj colonization in 20% of non-HIV immunosuppressed patients [39]. Therefore, early detection of asymptomatic infection of Pj in blood and respiratory specimens before, and during intense immunosuppression may enable selection of cases for pre-emptive treatment of Pj infection in order to prevent the development of IRD during reversal of immunosuppression [40,41].
Conclusions
PCP occurring in the context of IRD is not a rare phenomenon and is likely to be under-reported in the literature. In this setting, it may be more common for PCP to manifest acutely with a fulminant clinical course. Clinicians caring for immunosuppressed patients should be alert to this unique phenomenon so as to initiate timely and appropriate investigations and treatment for their patients. Serial monitoring of lymphocyte count, or if possible CD4 count, could serve as a useful adjunct to facilitate the diagnosis and management of this condition in a wide range of immunosuppressed hosts, especially in the presence of new pulmonary symptoms and/or radiographic abnormalities compatible with the diagnosis.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RAL and DSH were involved in the clinical evaluation and treatment of patients. BST and IFH helped with literature searching and review. AKW and VCC drafted and refined the manuscript. KYY conceived the study, participated in its design and coordination, and supervised the preparation of the manuscript. All authors have read and approved the final draft of the manuscript before submission.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Table 1 Summary of literature reported cases of HIV-negative immunocompromised patients with PCP manifested as IRD
Case [Ref.] Sex/Age (years) Underlying disease (s) Reduction of IS level before symptoms onset of IRD Symptoms & signs at IRD; change of lymphocyte count before & during IRD (if mentioned) Treatment, clinical progress & outcome
1–7 [22] M/F: 4:3 Median age 12, range 2–25 Acute leukemia in remission (4), acute leukemia in relapse (1), Hodgkin's disease (1), embryonal carcinoma of testes (1) P ↓ from 100 mg to 40 mg over 3 weeks in 1 patients;
In another 6 patients, P stopped in a median of 10.5 days, range (1–21 days) before symptoms onset NM Died (5) & survived (2)
8–15 [23] NM Primary brain tumour (8) Dexa ↓ over a median of 12.5 days, range (1–63 days) Fever (4), nonproductive cough (4), productive cough (2), dyspnoea (7), chest pain (4); CXR: bilateral infiltrates (3), diffuse infiltrates (3), focal infiltrates (1), clear (1) Died (3) & survived (5)
16 [24] M/55 Primary brain tumour (glioblastoma multiforme) Dexa ↓ from 16 mg qd to 2 mg qd over 8 weeks Intermittent fever, nonproductive cough, progressive dyspnoea; CXR: bilateral interstitial infiltrates; PaO2 (RA): 51 mmHg Treated with intravenous cotrimoxazole; survived
17 [24] F/74 Primary brain tumour (meningioma) Dexa ↓ from 12 mg qd to 4 mg qd over 2 weeks Intermittent fever, nonproductive cough; CXR: bilateral interstitial infiltrates; PaO2 (RA): 45 mmHg Treated with intravenous cotrimoxazole; survived
18 [24] M/50 Primary brain tumour (astrocytoma) Dexa ↓ from 16 mg qd to 1 mg qd over 8 weeks Fever, nonproductive cough, dyspnoea; CXR: bilateral interstitial infiltrates; PaO2 (RA): 73 mmHg Treated with intravenous cotrimoxazole; mechanical ventilation; survived
19 [24] M/75 Primary brain tumour (glioblastoma multiforme) Dexa ↓ from 16 mg qd to 4 mg qd over 6 weeks Fever, nonproductive cough, bloody diarrhoea; CXR: clear; PaO2 (RA): 89 mmHg Treated with intravenous cotrimoxazole; survived
20 [25] M/24 ACTH- producing metastatic bronchial carcinoid Serum cortisol ↓ from 138 pg/ml to 18 pg/ml over 54 days Fever, nonproductive cough, weakness, sweats; CXR: bilateral fluffy infiltrates; PaO2 (RA): 40 mmHg Treated with intravenous cotrimoxazole; mechanical ventilation; died of malignancy
21 [26] F/38 Endogenous Cushing's syndrome Metyrapone 750 mg qd added 1 day before symptoms onset Productive cough, dyspnoea; CXR: right lower upper lobe infiltrates; PaO2 (RA):31 mmHg Treated with intravenous cotrimoxazole; mechanical ventilation; died
22–28 [32] M/F 4:3 Mean (SD) age 53.1 (13.6) ITP (2), GN (2), bullous pemphigoid (1), endogenous Cushing's syndrome (1), and renal transplantation (1) Reduction of steroid but details of tailing regimen was not mentioned An upsurge of lymphocyte counts from the reduction of immunosuppression (median 300/μL, range 290 to 740/μL) to the onset of IRD (median 1500/μL, range 600 to 5620/μL) Treated with steroid as anti-PJP therapy in 7 (100%); mechanical ventilation in 6 (85.7%), died in 3 (42.9%)
29 M/33 (Our patient) Systemic lupus erythematosus/dermato-myositis overlapping syndrome P ↓ from 45 mg to 15 mg over 4 days Fever, dyspnoea; CXR: increased perihilar infilitrates; lymphocyte count increased from 600 to 1300/μL Treated with intravenous cotrimoxazole and steroid; survived
Note. Aza, azathioprine; CXR, chest radiograph; Dexa, dexamethasone; IRD, immunorestitution disease; ITP, immune thrombocytopenia purpura; IS, immunosuppression; GN, glomerulonephritis; P, prednisolone; PCP, Pneumocystis jiroveci pneumonia; RA, room air.
Table 2 Summary of literature reported cases of HIV-positive patients with IRD to PCP after HAART
Case [Ref.] Sex/Age CD4 (/μL) & HIVRNA (log10 copies/ml) before HAART Therapy of PJP & HAART regimen Day of HAART after initiation of PCP treatment Symptoms & signs during IRD Day of IRD after initiation of HAART Day of steroid withdrawal before the onset of IRD CD4 (/μL) & HIVRNA (log10 copies/ml) during IRD Therapy of IRD & clinical outcome
1 [28] M/37 7 & 5.1 Cotrimoxazole & MP; zidovudine, lamivudine, & indinavir 16 days High fever, acute respiratory failure; CXR: patchy alveolar opacities in both upper lobes 7 days 7 days 38 & UD Restart cotri-moxazole & stop HAART; survived
2 [28] M/47 28 & 5.0 Cotrimoxazole & MP, then aerosolized pentamidine; viramune, stavudine, & didanosine 1 day High fever, acute respiratory failure requiring intubation; CXR: diffuse alveolar opacities 17 days 2 days 40 & 4.5 Restart MP & stop HAART; survived
3 [28] F/50 230 & 5.8 Cotrimoxazole & MP; zidovudine, lamivudine, & indinavir 16 days High fever, acute respiratory failure; CXR: patchy alveolar opacities in both upper lobes 7 days 7 days 564 & 3.1 Start Atovaquone, aerosolized pentamidine, & steroid; survived
4–6 [29] NM 26 & 5.5 (median) Cotrimoxazole & high dose steroid; NM 15 – 18 days (range) Swinging fever, acute respiratory failure, & radiological deterioration 5 days (median); 3–17 days (range) NM 62 & 2.87 (median) Re-introducing high dose steroids & alternative PJP therapy; all three patients survived
7 [30] M/38 4 & 5.5 Atovaquone; didanosine, efavirenz, nelfinavir, & stavudine 35 days Fever, cough, dyspnoea, & night sweats; CXR: bilateral mid & lower zone airspace shadow 14 days NA 125 & 3.6 Intravenous pentamidine & hydrocortisone; survived
8 [30] NM 70 & NM Cotrimoxazole; zidovudine 182 & NM Cotrimoxazole; survived
9 [30] NM 10 & NM Cotrimoxazole; zidovudine 21 days (median) 17–24 days (range) Fever, dyspnoea, with or without cough 15 days (median) 5–30 days (range) NM 30 & NM Supportive therapy
10 [30] NM 216 & NM Cotrimoxazole & steroid; zidovudine 340 & NM Cotrimoxazole & steroid
11 [30] NM 290 & NM Cotrimoxazole; zidovudine, & didanosine 430 & NM Cotrimoxazole
12 [30] NM 60 & NM Cotrimoxazole; zidovudine 130 & NM Supportive therapy
13 [31] M/34 46 & > 5.9 Cotrimoxazole; zidovudine, lamivudine, lopinavir-ritonavir 18 days Recurrent fever, chest discomfort, cough, & dyspneoa; CXR showed diffuse bilateral reticulonodular infiltrates 14 days NA 435 & 4.5 Cotrimoxazole (pro-phylactic dose) & keeping HAART; survived
Note. CXR, chest radiograph; HAART, highly active antiretroviral therapy; IRD, immunorestitution disease; MP, methylprednisolone; NA, not applicable; NM, not mentioned; PCP, pneumocystis jiroveci pneumonia
==== Refs
Gajdusek DC Pneumocystis carinii; etiologic agent of interstitial plasma cell pneumonia of premature and young infants Pediatrics 1957 19 543 65 13419426
Dutz W Pneumocystis carinii pneumonia Pathol Annu 1970 5 309 41 5000423
Burke BA Good RA Pneumocystis carinii infection Medicine (Baltimore) 1973 52 23 51 4540153
Hughes WT Current issues in the epidemiology, transmission, and reactivation of Pneumocystis carinii Semin Respir Infect 1998 13 283 8 9872624
Masur H Ognibene FP Yarchoan R Shelhamer JH Baird BF Travis W Suffredini AF Deyton L Kovacs JA Falloon J CD4 counts as predictors of opportunistic pneumonias in human immunodeficiency virus (HIV) infection Ann Intern Med 1989 111 223 31 2546472
Chien SM Rawji M Mintz S Rachlis A Chan CK Changes in hospital admissions pattern in patients with human immunodeficiency virus infection in the era of Pneumocystis carinii prophylaxis Chest 1992 102 1035 9 1395739
Palella FJ JrDelaney KM Moorman AC Loveless MO Fuhrer J Satten GA Aschman DJ Holmberg SD Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators N Engl J Med 1998 338 853 60 9516219 10.1056/NEJM199803263381301
Pulvirenti J Herrera P Venkataraman P Ahmed N Pneumocystis carinii pneumonia in HIV-infected patients in the HAART era AIDS Patient Care STDS 2003 17 261 5 12880489 10.1089/108729103322108139
Sepkowitz KA Pneumocystis carinii pneumonia in patients without AIDS Clin Infect Dis 1993 S416 22 8274607
Varthalitis I Aoun M Daneau D Meunier F Pneumocystis carinii pneumonia in patients with cancer. An increasing incidence Cancer 1993 71 481 5 8422642
Arend SM Kroon FP van't Wout JW Pneumocystis carinii pneumonia in patients without AIDS, 1980 through 1993. An analysis of 78 cases Arch Intern Med 1995 155 2436 41 7503602 10.1001/archinte.155.22.2436
Lufft V Kliem V Behrend M Pichlmayr R Koch KM Brunkhorst R Incidence of Pneumocystis carinii pneumonia after renal transplantation. Impact of immunosuppression Transplantation 1996 62 421 3 8779695 10.1097/00007890-199608150-00022
Kulke MH Vance EA Pneumocystis carinii pneumonia in patients receiving chemotherapy for breast cancer Clin Infect Dis 1997 25 215 8 9332513
Bachelez H Schremmer B Cadranel J Mouly F Sarfati C Agbalika F Schlemmer B Mayaud CM Dubertret L Fulminant Pneumocystis carinii pneumonia in 4 patients with dermatomyositis Arch Intern Med 1997 157 1501 3 9224230 10.1001/archinte.157.13.1501
Varela JM Fernandez-Alonso J Wichmann I Calderon EJ Pneumocystis carinii investigation in patients with Wegener's granulomatosis Br J Rheumatol 1998 37 349 50 9566687 10.1093/rheumatology/37.3.349
Sepkowitz KA Brown AE Armstrong D Pneumocystis carinii pneumonia without acquired immunodeficiency syndrome. More patients, same risk Arch Intern Med 1995 155 1125 8 7763117 10.1001/archinte.155.11.1125
Sepkowitz KA Brown AE Telzak EE Gottlieb S Armstrong D Pneumocystis carinii pneumonia among patients without AIDS at a cancer hospital JAMA 1992 267 832 7 1732656 10.1001/jama.267.6.832
Yale SH Limper AH Pneumocystis carinii pneumonia in patients without acquired immunodeficiency syndrome: associated illness and prior corticosteroid therapy Mayo Clin Proc 1996 71 5 13 8538233
Anthony LB Greco FA Pneumocystis carinii pneumonia: a complication of Cushing's syndrome Ann Intern Med 1981 94 488 9 6971068
Graham BS Tucker WS Jr Opportunistic infections in endogenous Cushing's syndrome Ann Intern Med 1984 101 334 8 6331781
Johnson HD Johnson WW Pneumocystic carinii pneumonia in children with cancer. Diagnosis and treatment JAMA 1970 214 1067 73 5312183 10.1001/jama.214.6.1067
Singer C Armstrong D Rosen PP Schottenfeld D Pneumocystis carinii pneumonia: a cluster of eleven cases Ann Intern Med 1975 82 772 7 1094880
Henson JW Jalaj JK Walker RW Stover DE Fels AO Pneumocystis carinii pneumonia in patients with primary brain tumors Arch Neurol 1991 48 406 9 2012515
Slivka A Wen PY Shea WM Loeffler JS Pneumocystis carinii pneumonia during steroid taper in patients with primary brain tumors Am J Med 1993 94 216 9 8381584 10.1016/0002-9343(93)90187-T
Natale RB Yagoda A Brown A Singer C Stover D Bajorunas D Combined Pneumocystis carinii and Nocardia asteroides pneumonitis in a patient with an ACTH-producing carcinoid Cancer 1981 47 2933 5 7020921
Fulkerson WJ Newman JH Endogenous Cushing's syndrome complicated by Pneumocystis carinii pneumonia Am Rev Respir Dis 1984 129 188 9 6608298
Cheng VC Yuen KY Chan WM Wong SS Ma ES Chan RM Immunorestitution disease involving the innate and adaptive response Clin Infect Dis 2000 30 882 92 10880300 10.1086/313809
Wislez M Bergot E Antoine M Parrot A Carette MF Mayaud C Cadranel J Acute respiratory failure following HAART introduction in patients treated for Pneumocystis carinii pneumonia Am J Respir Crit Care Med 2001 164 847 51 11549544
Dean GL Williams DI Churchill DR Fisher MJ Transient clinical deterioration in HIV patients with Pneumocystis carinii pneumonia after starting highly active antiretroviral therapy: another case of immune restoration inflammatory syndrome Am J Respir Crit Care Med 2002 165 1670 12070070
Barry SM Lipman MC Deery AR Johnson MA Janossy G Immune reconstitution pneumonitis following Pneumocystis carinii pneumonia in HIV-infected subjects HIV Med 2002 3 207 11 12139660 10.1046/j.1468-1293.2002.00115.x
Koval CE Gigliotti F Nevins D Demeter LM Immune reconstitution syndrome after successful treatment of Pneumocystis carinii pneumonia in a man with human immunodeficiency virus type 1 infection Clin Infect Dis 2002 35 491 3 12145736 10.1086/341974
Cheng VC Hung IF Wu AK Tang BS Chu CM Yuen KY Lymphocyte surge as a marker for immunorestitution caused by Pneumocystis carinii pneumonia in non-HIV immunosuppressed hosts Eur J Clin Microbiol Infect Dis 2004 23 512 4 15141339 10.1007/s10096-004-1140-6
Cheng VC Yuen KY Wong SS Woo PC Ho PL Lee R Chan RM Immunorestitution diseases in patients not infected with HIV Eur J Clin Microbiol Infect Dis 2001 20 402 6 11476440 10.1007/s100960100507
Cheng VC Woo PC Lau SK Cheung CH Yung RW Yam LY Yuen KY Peripartum tuberculosis as a form of immunorestitution disease Eur J Clin Microbiol Infect Dis 2003 22 313 7 12743831 10.1007/s10096-003-0998-z
Cheng VC Ho PL Lee RA Chan KS Chan KK Woo PC Lau SK Yuen KY Clinical spectrum of paradoxical deterioration during antituberculosis therapy in non-HIV-infected patients Eur J Clin Microbiol Infect Dis 2002 21 803 9 12461590 10.1007/s10096-002-0821-2
Cheng VC Yam WC Woo PC Lau SK Hung IF Wong SP Cheung WC Yuen KY Risk factors for development of paradoxical response during antituberculosis therapy in HIV-negative patients Eur J Clin Microbiol Infect Dis 2003 22 597 602 14508660 10.1007/s10096-003-0998-z
Mansharamani NG Balachandran D Vernovsky I Garland R Koziel H Peripheral blood CD4 + T-lymphocyte counts during Pneumocystis carinii pneumonia in immunocompromised patients without HIV infection Chest 2000 118 712 20 10988193 10.1378/chest.118.3.712
Nevez G Raccurt C Vincent P Jounieaux V Dei-Cas E Pulmonary colonization with Pneumocystis carinii in human immunodeficiency virus-negative patients: assessing risk with blood CD4+ T cell counts Clin Infect Dis 1999 29 1331 2 10524989 10.1086/313478
Sing A Trebesius K Roggenkamp A Russmann H Tybus K Pfaff F Bogner JR Emminger C Heesemann J Evaluation of diagnostic value and epidemiological implications of PCR for Pneumocystis carinii in different immunosuppressed and immunocompetent patient groups J Clin Microbiol 2000 38 1461 7 10747126
Rabodonirina M Cotte L Boibieux A Kaiser K Mayencon M Raffenot D Trepo C Peyramond D Picot S Detection of Pneumocystis carinii DNA in blood specimens from human immunodeficiency virus-infected patients by nested PCR J Clin Microbiol 1999 37 127 31 9854076
Ribes JA Limper AH Espy MJ Smith TF PCR detection of Pneumocystis carinii in bronchoalveolar lavage specimens: analysis of sensitivity and specificity J Clin Microbiol 1997 35 830 5 9157136
| 15588295 | PMC539247 | CC BY | 2021-01-04 16:03:31 | no | BMC Infect Dis. 2004 Dec 9; 4:57 | utf-8 | BMC Infect Dis | 2,004 | 10.1186/1471-2334-4-57 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-4-571558829510.1186/1471-2334-4-57Case ReportThe unmasking of Pneumocystis jiroveci pneumonia during reversal of immunosuppression: case reports and literature review Wu Alan KL [email protected] Vincent CC [email protected] Bone SF [email protected] Ivan FN [email protected] Rodney A [email protected] David S [email protected] Kwok Y [email protected] Division of Infectious Diseases, Centre of Infection, Queen Mary Hospital, The University of Hong Kong; Hong Kong Special Administrative Region, China2 Division of Pulmonary Medicine, Department of Medicine and Therapeutics, Prince of Wales Hospital, The Chinese University of Hong Kong; Hong Kong Special Administrative Region, China2004 9 12 2004 4 57 57 1 9 2004 9 12 2004 Copyright © 2004 Wu 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
Pneumocystis jiroveci pneumonia (PCP) is an important opportunistic infection among immunosuppressed patients, especially in those infected with human immunodeficiency virus (HIV). The clinical presentation of PCP in immunosuppressed patients have been well-reported in the literature. However, the clinical importance of PCP manifesting in the setting of an immunorestitution disease (IRD), defined as an acute symptomatic or paradoxical deterioration of a (presumably) preexisting infection, which is temporally related to the recovery of the immune system and is due to immunopathological damage associated with the reversal of immunosuppressive processes, has received relatively little attention until recently.
Case presentation
We aim to better define this unique clinical syndrome by reporting two cases of PCP manifesting acutely with respiratory failure during reversal of immunosuppression in non-HIV infected patients, and reviewed the relevant literature. We searched our databases for PCP cases manifesting in the context of IRD according to our predefined case definition, and reviewed the case notes retrospectively. A comprehensive search was performed using the Medline database of the National Library of Medicine for similar cases reported previously in the English literature in October 2003. A total of 28 non-HIV (excluding our present case) and 13 HIV-positive patients with PCP manifesting as immunorestitution disease (IRD) have been reported previously in the literature. During immunorestitution, a consistent rise in the median CD4 lymphocyte count (28/μL to 125/μL), with a concomitant fall in the median HIV viral load (5.5 log10 copies/ml to 3.1 log10 copies/ml) was observed in HIV-positive patients who developed PCP. A similar upsurge in peripheral lymphocyte count was observed in our patients preceding the development of PCP, as well as in other non-HIV immunosuppressed patients reported in the literature.
Conclusions
PCP manifesting as IRD may be more common than is generally appreciated. Serial monitoring of total lymphocyte or CD4 count could serve as a useful adjunct to facilitate the early diagnosis and pre-emptive treatment of this condition in a wide range of immunosuppressed hosts, especially in the presence of new pulmonary symptoms and/or radiographic abnormalities compatible with the diagnosis.
==== Body
Background
Pneumocystis jiroveci (Pj) (previously known as Pneumocystis carinii f. sp. hominis) was first identified as a pathogen in premature infants suffering from interstitial plasma cell pneumonia in European countries during and after World War II, occasionally occurring in epidemics [1-3]. Since then Pneumocystis pneumonia (PCP) had only been reported sporadically in patients with malignancies and solid organ transplantations until the HIV epidemic [4]. The incidence of PCP increased significantly after the emergence of human immunodeficiency virus (HIV) infection. However, with the identification of CD4 T lymphocyte depletion as an independent risk factor for the development of PCP [5], widespread use of antimicrobial prophylaxis [4], and the introduction of highly active antiretroviral therapy (HAART), there has been a steady decline in the incidence of PCP among HIV-infected patients [7,8].
Nevertheless, with the rising number of patients receiving immunosuppressive therapies for malignancies, solid organ transplantations and autoimmune diseases, PCP has been increasingly recognized in non-HIV immunosuppressed hosts [9-15]. For instance, PCP occurs in 3.4% to 43% of solid organ transplant recipients [16], and it is particularly prevalent among those patients who are put on long-term steroids. In a non-HIV immunosuppressed cohort with PCP, the use of steroids was found to be a contributing factor in 87% of patients [17]. In another similar cohort of immunosuppressed patients, steroids had been administered systemically in 90.5% within one month before the diagnosis of PCP. Although a median daily dose equivalent to 30 mg of prednisone was administered in most of these patients prior to the development of PCP, up to 25% had received as little as 16 mg of prednisone daily [18]. Interestingly, PCP has also been reported in patients with endogenous steroid excess due to Cushing's disease [19,20].
Paradoxically, the clinical symptoms of PCP were often unmasked in HIV-negative immunosuppressed patients during the reversal of immunosuppression, often at the time when the dose of steroids was tapered [11,17,21-24], or when the endogenous steroid production was reduced [25-27]. However, serial changes in the absolute lymphocyte count before and during reversal of immunosuppression were not mentioned in these patients. Recently, paradoxical worsening of clinical symptoms and signs of PCP after initiation of HAART has also been reported in HIV-positive patients [28-31]. The onset of clinical deterioration was associated with an upsurge in the CD4 lymphocyte count and a reduction in the HIV viral load [28-31]. Tissue damage is thought to occur as a result of immune reconstitution in HIV-positive patients. Here, we report two cases and review the literature on this topic from the perspective of immunorestitution disease.
Case presentation
Case 1
This is a fifty-one year old female patient with history of diabetes mellitus and systemic lupus erythematoses (SLE) complicated by lupus nephritis. Although we have included her case in our previous publication [32], we have not reported her clinical details at that time. She was put on prednisolone 30 mg and azathioprine 100 mg daily since end of June and mid-July 2002, respectively. She was admitted to Queen Mary Hospital on 11th August 2002 for investigation of jaundice. Investigations revealed deranged liver function tests with cholestatic pattern. A diagnosis of drug-induced hepatotoxicity was entertained, and azathioprine was stopped after admission. As her autoimmune disease was under control, her steroid dosage was reduced from 25 mg to 15 mg daily within the next 14 days. Her CXR taken on admission was normal.
Soon after her immunosuppressive therapy was tapered, she developed fever and non-productive cough. A repeat CXR performed on 9th Sept revealed new infiltrate over the left mid-zone, suggestive of pneumonia. She was started on intravenous ceftazidime 1 gram eight hourly and oral clarithromycin 500 mg twice daily. Serial CXR performed three days later showed increasing bilateral pulmonary infiltrates and worsening hypoxemia. There was an upsurge of total lymphocyte count from 0.7 × 109/L (total white cell count 7.2 × 109/L) at the time of admission to 5.6 × 109/L (total white cell count 10.8 × 109/L) at the time of clinical deterioration. Bronchoscopy with transbronchial biopsy performed on the same day revealed Pneumocystitis jiroveci by methenamine sliver stain. Workup for other opportunistic pathogens including cytomegalovirus and aspergillus was negative. She was commenced on intravenous pentamidine (4 mg/kg/day) and corticosteroids for severe PCP infection. Despite active treatment she developed progressive respiratory failure and required admission to intensive care unit. She subsequently recovered after a stormy hospital course, and upon discharge from hospital, her total lymphocyte count had returned to her baseline of 0.86 × 109/L.
Case 2
A thirty-three year old gentleman initially presented to Prince of Wales Hospital with a diagnosis of SLE/dermatomyositis overlap syndrome. He was treated with steroid and hydroxychloroquine 200 mg twice daily since 1997. He had a flare up of disease in May 1998 with active vasculitis and myositis, for which he was put on prednisolone and azthioprine 50 mg and 100 mg daily respectively. Upon reassessment one month later, disease activity was under control, and the dosage of prednisolone was reduced to 45 mg daily.
Twelve day after reducing the immunosuppressive regimen, he was admitted to hospital for treatment of left buttock abscess. The CXR taken on admission was unremarkable. An aspirate of the pus from the lesion grew methicillin-sensitive staphylococcus aureus; he was treated with cloxacillin 1 g intravenously every 6 hourly, together with incision and drainage of the buttock abscess. In view of the underlying active pyogenic infection, the steroid dosage was rapidly tapered from 45 mg to 15 mg daily within the next four days. However, he was noted to have persistent fever associated with mild unproductive cough. A repeat chest radiograph showed new infiltrates over the right upper and left lower zones, and he was empirically treated with intravenous ceftazidime 1 gram every 8 hours, cloxacillin 1 gram every 6 hours and netimicin 100 mg every 8 hours. As there was no clinical response after 5 days of treatment, bronchoscopy and bronchoalveolar lavage (BAL) was performed, which was positive for Pneumocystis jiroveci. Investigation for the presence of co-existing opportunistic pathogens such as cytomegalovirus and aspergillus was negative. On the day after bronchoscopy, he was commenced on intravenous cotrimoxazole 1.3 grams every 6 hours. He remained stable initially with fever on downward trend. However, on the 3rd day of treatment, he developed sudden desaturation with resurgence of high fever, and required supplemental oxygen therapy. Repeat chest radiograph showed increased perihilar hazziness in both lung fields. There was also an upsurge of total lymphocyte count from 0.6 × 109/L (total white cell count 11.2 × 109/L) on admission, to 1.3 × 109/L (total white cell count 10.4 × 109/L) at the time of clinical deterioration. He was treated with high dose prednisolone (80 mg daily), and his condition improved promptly afterwards. He was subsequently discharged, and on follow up at the clinic one month later, his total lymphocyte count had returned to his baseline level of 0.6 × 109/L.
Immunorestitution disease (IRD) has been described in both HIV and non-HIV immunosuppressed hosts previously [27-31]. In the setting of PCP, it is defined as an acute symptomatic presentation of the disease that is related temporally to the recovery of the immune system, associated with reversal of immunosuppressive processes such as reduction in the dosage of corticosteroids and/or cytotoxic agents or a reduction of HIV viral load due to HAART, which results in the development of immunopathological damage. The preexisting microbial infection could be either asymptomatic or mildly symptomatic. Using this case definition, we attempted to review the English literature for other reported cases of PCP manifesting as IRD. The English-language literature (1966 – 2003) was searched in the Medline database of the National Library of Medicine in October 2003. The keywords "Pneumocystis carinii", "Pneumocystis jiroveci", "HIV", immunosuppression", "immunosuppressive", "steroid", and "corticosteroid" were used to select cases. All the case reports and case series with clinical details were included in this study if they fulfilled the above definition of IRD. When appropriate, the cited bibliographies were also retrieved for further analysis. As for statistical analysis, we used the Wilcoxon Signed Rank test, a non-parametric test for comparing paired samples, to analyze the serial changes in lymphocyte counts and HIV viral loads before and during the development of IRD. A two-tailed p-value of less than 0.05 was considered significant. All statistical analyses were performed using SPSS version 11.5 for Windows.
Including our present case, a total of 29 cases of PCP in non-HIV immunosuppressed hosts fulfilling our definition of IRD have been reported in the literature (table 1) [22-27,32]. There were altogether 13 males and 8 females, with a median age of 38 years (range 2 to 75 years). The age and sex were not mentioned in 8 cases. Their underlying immunosuppressive conditions included solid organ tumours (13 cases), haematological diseases (8), autoimmune diseases (4), endogenous Cushing's disease (3), and a solid organ transplant recipient (1). All patients had received steroids or had excessive endogenous steroid production, whereas 18 (62.1%) of them had concomitant cytotoxic therapy for the underlying diseases. The median duration between steroid tapering and clinical manifestations of PCP was 21 days (range 1 to 83 days). Steroids were completely withdrawn at a median of 7.5 days (range 1 to 21 days) before the onset of symptoms in eight patients. Serial lymphocyte counts were only available in eight patients. An upsurge of the absolute lymphocyte counts was observed from the time of reduction of immunosuppression (median 300/μL, range 290 to 600/μL at baseline) to the time of occurrence of IRD (median 1200/μL, range 600 to 5620/μL); the median increase in total lymphocyte count was 800/μL, with a range of 300 to 4880/μL. Comparing the lymphocyte counts before and after reversal of immunosuppressive therapy, the difference was statistically significant (Wilcoxon Signed Rank Test for paired samples; p = 0.012). In addition to our patient, reintroduction or increasing doses of steroids were required in 7 (53.8%) of 13 patients in the acute management of PCP in the literature, at the time when they developed clinical deterioration during antimicrobial therapy [24-27]. Seven (53.8%) of 13 cases had respiratory failure requiring mechanical ventilation. Among these 29 cases, 13 (44.8%) subsequently died of PCP.
Among HIV-positive patients, 13 cases with newly diagnosed PCP were reported in the literature, in which IRD occurred shortly after the introduction of HAART (table 2) [28-31]. Seven (53.8%) out of 13 cases received steroids as adjunctive therapy in addition to antimicrobials. HAART was given in all cases at a median 18 days (range 1 to 35 days) after the initiation of treatment for PCP. During IRD, recurrence of fever (100%), dyspnoea (100%), and paradoxical worsening of pulmonary infiltrates (58.3%) were observed in these patients [28-31]. IRD occurred at a median of 14 days (range 5 to 17 days) after HAART. An upsurge of the CD4 lymphocyte count was observed before (median 28/μL, range 4 to 290/μL) and during IRD (median 125/μL, range 30 to 564/μL); this was associated with a concomitant reduction of the median HIV viral load from 5.5 log10 copies/ml (range 5.0 to 5.9 log10 copies/ml) to 3.1 log10 copies/ml (range 2.9 to 4.5 log10 copies/ml) before and during IRD respectively, and the differences observed in both the CD4 counts and viral loads before and during IRD reached statistical significance (Wilcoxon Signed Rank Test for related samples; p = 0.001 and 0.017, respectively). Antimicrobials, steroids, or both for PCP were reintroduced for IRD in 4, 1, and 6 cases respectively. Only 2 cases were treated conservatively. One case required mechanical ventilation for severe respiratory distress. None of the patients died.
PCP manifesting as a form of IRD is not a rare phenomenon. As shown in our previous study, it happens in 7 out of 10 (70%) of HIV-negative immunosuppressed hosts infected with Pj [32]. However, the diagnosis of PCP is usually delayed in this group of patients because of atypical presentation. In this clinical setting, PCP manifesting as IRD often runs an acute and fulminant course, with nonspecific lesions on chest radiographs, and high absolute lymphocyte counts [32]. In our own reported series, despite the administration of steroid therapy to suppress the immunopathological damage, more than 80% of patients developed acute respiratory failure and required mechanical ventilation. Patients who developed PCP during reversal of immunosuppressive therapy in our series tended to be older, and this might partially explain the increased mortality observed in this group [32].
Rapid reduction of immunosuppressive therapy such as steroids has been implicated as a predisposing factor for the development of PCP in HIV-negative patients [11,17,23,24]. In one study, PCP occurred in 79 (70%) of 113 patients during steroid tapering [17]. Another study suggested that 8 (72.2%) out of 11 episodes of PCP developed when steroid therapy was tapered [23]. A subsequent study also demonstrated that 43% of patients had a rapid reduction of steroid dosing before the clinical manifestations of PCP [11]. A similar experience was reported in children, and 17 (89.5%) out of 19 children were diagnosed to have PCP during steroid tapering according to a previous report [21]. Another series revealed that 7 of 11 patients experienced acutely symptomatic PCP when the dose of steroids was decreased or terminated 5 days to 3 weeks before the diagnosis of PCP [22]. However, all these cases were not analyzed from a perspective of IRD. Serial changes of the absolute lymphocyte counts or their subsets were either not noted or reported [11,17,21-24]. Hence we have not included these cases for further analysis in this review.
Among HIV positive patients, PCP manifesting acutely during the initiation anti-retroviral therapy is a well-recognized phenomenon. The underlying immunopathological nature of this condition, which is reminiscent to IRD occurring in non-HIV infected patients, has been confirmed by histological examination of the lungs and transbronchial biopsy specimens, which demonstrated mixed inflammatory infiltrates including macrophages, neutrophils, lymphocytes, and plasma cells. Almost all infiltrating lymphocytes found in the tissues were of the T cell lineage, shown by immunophenotyping to be predominantly CD4 and CD8 cells [28]. In another study [30], the BAL fluid obtained from one of six patients with an IRD-type presentation of PCP was analyzed. Infiltration of predominantly CD4 and CD8 lymphocytes with the proliferative marker (Ki67) and perforin-positive cell were seen in the BAL specimen. Therefore, it is likely that the phagocytosed Pj is presented by alveolar macrophage to T cells, which trigger the inflammatory response [30].
In our own experience, as well as from the review of published literature, it appears that a surge of absolute lymphocyte count, especially the CD4 lymphocyte count in HIV-positive patients, could potentially act as a surrogate marker for immunopathological damage during IRD in both HIV-negative and HIV-positive patients. In our recent publication [32], 7 out of 10 non-HIV immunosuppressed patients demonstrated a consistent rise in the absolute lymphocyte count during tapering of immunosuppression prior to the onset of symptomatic PCP. In this group of patients, the surge in lymphocyte count is likely the result of withdrawal of lymphocytotoxic immunosuppressants such as corticosteroids. Similarly, a rising trend of the CD4 lymphocyte count, consistent with immune reconstitution after HAART, was also observed in 13 HIV-positive cases before and during the development of symptomatic PCP [28-31]. In fact, an upsurge in the absolute lymphocyte count has been shown to be a marker of IRD in our previous publications involving viral and tuberculous infections [33-36]. However, it must be emphasized that the number of circulating lymphocytes may not always correlate with their number in the affected tissues or their in vivo functional activity. This can be exemplified by a case of PCP occurring during steroid withdrawal, in which the lymphocyte counts surged to a very high level and then rapidly dropped to a low level within one day. The migration of lymphocytes from the circulation to tissue might explain this rapid drop in lymphocyte count and the resulting immunopathological damage [27]. In the future, further studies on lymphocyte subsets and cytokine profiles of susceptible hosts during the development of IRD should be performed to elucidate the underlying immunopathological mechanisms behind this interesting phenomenon.
From the result of this review, it appears that HIV-positive patients with PCP are at risk of clinical deterioration due to IRD if HAART therapy is started within 1 to 2 weeks after the initiation of treatment for PCP (table 2). With a better understanding of the pathogenetic mechanisms resulting in IRD, we may be able to prevent the occurrence of IRD by delaying the initiation of HAART in HIV-positive patients with PCP. However, in non-HIV immunosuppressed patients, it is even more important to recognize the atypical presentations of PCP in the context of IRD. Since the clinical and/or radiological features alone may not be sufficient for diagnosis, analysis of serial changes in lymphocyte counts in patients undergoing a reduction of immunosuppression can alert the clinician to the possibility of IRD due to occult pathogens such as Pj. To prevent IRD in non-HIV immunosuppressed patients, the use of prophylactic antibiotics against Pj to reduce the microbial load in selected patients remains an important issue. Recently, a multi-center study showed that the CD4 lymphocyte count may be a useful marker to monitor the risk of development of PCP in non-HIV immunosuppressed hosts [37], and patients with low CD4 lymphocyte counts of less than 300 or 400 may require prophylaxis. In fact, asymptomatic colonization of Pj has been demonstrated in HIV-negative patients when the CD4 lymphocyte count was less than 400 [38]. Nested polymerase chain reaction (PCR) identified a significant percentage of clinically silent Pj colonization in 20% of non-HIV immunosuppressed patients [39]. Therefore, early detection of asymptomatic infection of Pj in blood and respiratory specimens before, and during intense immunosuppression may enable selection of cases for pre-emptive treatment of Pj infection in order to prevent the development of IRD during reversal of immunosuppression [40,41].
Conclusions
PCP occurring in the context of IRD is not a rare phenomenon and is likely to be under-reported in the literature. In this setting, it may be more common for PCP to manifest acutely with a fulminant clinical course. Clinicians caring for immunosuppressed patients should be alert to this unique phenomenon so as to initiate timely and appropriate investigations and treatment for their patients. Serial monitoring of lymphocyte count, or if possible CD4 count, could serve as a useful adjunct to facilitate the diagnosis and management of this condition in a wide range of immunosuppressed hosts, especially in the presence of new pulmonary symptoms and/or radiographic abnormalities compatible with the diagnosis.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RAL and DSH were involved in the clinical evaluation and treatment of patients. BST and IFH helped with literature searching and review. AKW and VCC drafted and refined the manuscript. KYY conceived the study, participated in its design and coordination, and supervised the preparation of the manuscript. All authors have read and approved the final draft of the manuscript before submission.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Table 1 Summary of literature reported cases of HIV-negative immunocompromised patients with PCP manifested as IRD
Case [Ref.] Sex/Age (years) Underlying disease (s) Reduction of IS level before symptoms onset of IRD Symptoms & signs at IRD; change of lymphocyte count before & during IRD (if mentioned) Treatment, clinical progress & outcome
1–7 [22] M/F: 4:3 Median age 12, range 2–25 Acute leukemia in remission (4), acute leukemia in relapse (1), Hodgkin's disease (1), embryonal carcinoma of testes (1) P ↓ from 100 mg to 40 mg over 3 weeks in 1 patients;
In another 6 patients, P stopped in a median of 10.5 days, range (1–21 days) before symptoms onset NM Died (5) & survived (2)
8–15 [23] NM Primary brain tumour (8) Dexa ↓ over a median of 12.5 days, range (1–63 days) Fever (4), nonproductive cough (4), productive cough (2), dyspnoea (7), chest pain (4); CXR: bilateral infiltrates (3), diffuse infiltrates (3), focal infiltrates (1), clear (1) Died (3) & survived (5)
16 [24] M/55 Primary brain tumour (glioblastoma multiforme) Dexa ↓ from 16 mg qd to 2 mg qd over 8 weeks Intermittent fever, nonproductive cough, progressive dyspnoea; CXR: bilateral interstitial infiltrates; PaO2 (RA): 51 mmHg Treated with intravenous cotrimoxazole; survived
17 [24] F/74 Primary brain tumour (meningioma) Dexa ↓ from 12 mg qd to 4 mg qd over 2 weeks Intermittent fever, nonproductive cough; CXR: bilateral interstitial infiltrates; PaO2 (RA): 45 mmHg Treated with intravenous cotrimoxazole; survived
18 [24] M/50 Primary brain tumour (astrocytoma) Dexa ↓ from 16 mg qd to 1 mg qd over 8 weeks Fever, nonproductive cough, dyspnoea; CXR: bilateral interstitial infiltrates; PaO2 (RA): 73 mmHg Treated with intravenous cotrimoxazole; mechanical ventilation; survived
19 [24] M/75 Primary brain tumour (glioblastoma multiforme) Dexa ↓ from 16 mg qd to 4 mg qd over 6 weeks Fever, nonproductive cough, bloody diarrhoea; CXR: clear; PaO2 (RA): 89 mmHg Treated with intravenous cotrimoxazole; survived
20 [25] M/24 ACTH- producing metastatic bronchial carcinoid Serum cortisol ↓ from 138 pg/ml to 18 pg/ml over 54 days Fever, nonproductive cough, weakness, sweats; CXR: bilateral fluffy infiltrates; PaO2 (RA): 40 mmHg Treated with intravenous cotrimoxazole; mechanical ventilation; died of malignancy
21 [26] F/38 Endogenous Cushing's syndrome Metyrapone 750 mg qd added 1 day before symptoms onset Productive cough, dyspnoea; CXR: right lower upper lobe infiltrates; PaO2 (RA):31 mmHg Treated with intravenous cotrimoxazole; mechanical ventilation; died
22–28 [32] M/F 4:3 Mean (SD) age 53.1 (13.6) ITP (2), GN (2), bullous pemphigoid (1), endogenous Cushing's syndrome (1), and renal transplantation (1) Reduction of steroid but details of tailing regimen was not mentioned An upsurge of lymphocyte counts from the reduction of immunosuppression (median 300/μL, range 290 to 740/μL) to the onset of IRD (median 1500/μL, range 600 to 5620/μL) Treated with steroid as anti-PJP therapy in 7 (100%); mechanical ventilation in 6 (85.7%), died in 3 (42.9%)
29 M/33 (Our patient) Systemic lupus erythematosus/dermato-myositis overlapping syndrome P ↓ from 45 mg to 15 mg over 4 days Fever, dyspnoea; CXR: increased perihilar infilitrates; lymphocyte count increased from 600 to 1300/μL Treated with intravenous cotrimoxazole and steroid; survived
Note. Aza, azathioprine; CXR, chest radiograph; Dexa, dexamethasone; IRD, immunorestitution disease; ITP, immune thrombocytopenia purpura; IS, immunosuppression; GN, glomerulonephritis; P, prednisolone; PCP, Pneumocystis jiroveci pneumonia; RA, room air.
Table 2 Summary of literature reported cases of HIV-positive patients with IRD to PCP after HAART
Case [Ref.] Sex/Age CD4 (/μL) & HIVRNA (log10 copies/ml) before HAART Therapy of PJP & HAART regimen Day of HAART after initiation of PCP treatment Symptoms & signs during IRD Day of IRD after initiation of HAART Day of steroid withdrawal before the onset of IRD CD4 (/μL) & HIVRNA (log10 copies/ml) during IRD Therapy of IRD & clinical outcome
1 [28] M/37 7 & 5.1 Cotrimoxazole & MP; zidovudine, lamivudine, & indinavir 16 days High fever, acute respiratory failure; CXR: patchy alveolar opacities in both upper lobes 7 days 7 days 38 & UD Restart cotri-moxazole & stop HAART; survived
2 [28] M/47 28 & 5.0 Cotrimoxazole & MP, then aerosolized pentamidine; viramune, stavudine, & didanosine 1 day High fever, acute respiratory failure requiring intubation; CXR: diffuse alveolar opacities 17 days 2 days 40 & 4.5 Restart MP & stop HAART; survived
3 [28] F/50 230 & 5.8 Cotrimoxazole & MP; zidovudine, lamivudine, & indinavir 16 days High fever, acute respiratory failure; CXR: patchy alveolar opacities in both upper lobes 7 days 7 days 564 & 3.1 Start Atovaquone, aerosolized pentamidine, & steroid; survived
4–6 [29] NM 26 & 5.5 (median) Cotrimoxazole & high dose steroid; NM 15 – 18 days (range) Swinging fever, acute respiratory failure, & radiological deterioration 5 days (median); 3–17 days (range) NM 62 & 2.87 (median) Re-introducing high dose steroids & alternative PJP therapy; all three patients survived
7 [30] M/38 4 & 5.5 Atovaquone; didanosine, efavirenz, nelfinavir, & stavudine 35 days Fever, cough, dyspnoea, & night sweats; CXR: bilateral mid & lower zone airspace shadow 14 days NA 125 & 3.6 Intravenous pentamidine & hydrocortisone; survived
8 [30] NM 70 & NM Cotrimoxazole; zidovudine 182 & NM Cotrimoxazole; survived
9 [30] NM 10 & NM Cotrimoxazole; zidovudine 21 days (median) 17–24 days (range) Fever, dyspnoea, with or without cough 15 days (median) 5–30 days (range) NM 30 & NM Supportive therapy
10 [30] NM 216 & NM Cotrimoxazole & steroid; zidovudine 340 & NM Cotrimoxazole & steroid
11 [30] NM 290 & NM Cotrimoxazole; zidovudine, & didanosine 430 & NM Cotrimoxazole
12 [30] NM 60 & NM Cotrimoxazole; zidovudine 130 & NM Supportive therapy
13 [31] M/34 46 & > 5.9 Cotrimoxazole; zidovudine, lamivudine, lopinavir-ritonavir 18 days Recurrent fever, chest discomfort, cough, & dyspneoa; CXR showed diffuse bilateral reticulonodular infiltrates 14 days NA 435 & 4.5 Cotrimoxazole (pro-phylactic dose) & keeping HAART; survived
Note. CXR, chest radiograph; HAART, highly active antiretroviral therapy; IRD, immunorestitution disease; MP, methylprednisolone; NA, not applicable; NM, not mentioned; PCP, pneumocystis jiroveci pneumonia
==== Refs
Gajdusek DC Pneumocystis carinii; etiologic agent of interstitial plasma cell pneumonia of premature and young infants Pediatrics 1957 19 543 65 13419426
Dutz W Pneumocystis carinii pneumonia Pathol Annu 1970 5 309 41 5000423
Burke BA Good RA Pneumocystis carinii infection Medicine (Baltimore) 1973 52 23 51 4540153
Hughes WT Current issues in the epidemiology, transmission, and reactivation of Pneumocystis carinii Semin Respir Infect 1998 13 283 8 9872624
Masur H Ognibene FP Yarchoan R Shelhamer JH Baird BF Travis W Suffredini AF Deyton L Kovacs JA Falloon J CD4 counts as predictors of opportunistic pneumonias in human immunodeficiency virus (HIV) infection Ann Intern Med 1989 111 223 31 2546472
Chien SM Rawji M Mintz S Rachlis A Chan CK Changes in hospital admissions pattern in patients with human immunodeficiency virus infection in the era of Pneumocystis carinii prophylaxis Chest 1992 102 1035 9 1395739
Palella FJ JrDelaney KM Moorman AC Loveless MO Fuhrer J Satten GA Aschman DJ Holmberg SD Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators N Engl J Med 1998 338 853 60 9516219 10.1056/NEJM199803263381301
Pulvirenti J Herrera P Venkataraman P Ahmed N Pneumocystis carinii pneumonia in HIV-infected patients in the HAART era AIDS Patient Care STDS 2003 17 261 5 12880489 10.1089/108729103322108139
Sepkowitz KA Pneumocystis carinii pneumonia in patients without AIDS Clin Infect Dis 1993 S416 22 8274607
Varthalitis I Aoun M Daneau D Meunier F Pneumocystis carinii pneumonia in patients with cancer. An increasing incidence Cancer 1993 71 481 5 8422642
Arend SM Kroon FP van't Wout JW Pneumocystis carinii pneumonia in patients without AIDS, 1980 through 1993. An analysis of 78 cases Arch Intern Med 1995 155 2436 41 7503602 10.1001/archinte.155.22.2436
Lufft V Kliem V Behrend M Pichlmayr R Koch KM Brunkhorst R Incidence of Pneumocystis carinii pneumonia after renal transplantation. Impact of immunosuppression Transplantation 1996 62 421 3 8779695 10.1097/00007890-199608150-00022
Kulke MH Vance EA Pneumocystis carinii pneumonia in patients receiving chemotherapy for breast cancer Clin Infect Dis 1997 25 215 8 9332513
Bachelez H Schremmer B Cadranel J Mouly F Sarfati C Agbalika F Schlemmer B Mayaud CM Dubertret L Fulminant Pneumocystis carinii pneumonia in 4 patients with dermatomyositis Arch Intern Med 1997 157 1501 3 9224230 10.1001/archinte.157.13.1501
Varela JM Fernandez-Alonso J Wichmann I Calderon EJ Pneumocystis carinii investigation in patients with Wegener's granulomatosis Br J Rheumatol 1998 37 349 50 9566687 10.1093/rheumatology/37.3.349
Sepkowitz KA Brown AE Armstrong D Pneumocystis carinii pneumonia without acquired immunodeficiency syndrome. More patients, same risk Arch Intern Med 1995 155 1125 8 7763117 10.1001/archinte.155.11.1125
Sepkowitz KA Brown AE Telzak EE Gottlieb S Armstrong D Pneumocystis carinii pneumonia among patients without AIDS at a cancer hospital JAMA 1992 267 832 7 1732656 10.1001/jama.267.6.832
Yale SH Limper AH Pneumocystis carinii pneumonia in patients without acquired immunodeficiency syndrome: associated illness and prior corticosteroid therapy Mayo Clin Proc 1996 71 5 13 8538233
Anthony LB Greco FA Pneumocystis carinii pneumonia: a complication of Cushing's syndrome Ann Intern Med 1981 94 488 9 6971068
Graham BS Tucker WS Jr Opportunistic infections in endogenous Cushing's syndrome Ann Intern Med 1984 101 334 8 6331781
Johnson HD Johnson WW Pneumocystic carinii pneumonia in children with cancer. Diagnosis and treatment JAMA 1970 214 1067 73 5312183 10.1001/jama.214.6.1067
Singer C Armstrong D Rosen PP Schottenfeld D Pneumocystis carinii pneumonia: a cluster of eleven cases Ann Intern Med 1975 82 772 7 1094880
Henson JW Jalaj JK Walker RW Stover DE Fels AO Pneumocystis carinii pneumonia in patients with primary brain tumors Arch Neurol 1991 48 406 9 2012515
Slivka A Wen PY Shea WM Loeffler JS Pneumocystis carinii pneumonia during steroid taper in patients with primary brain tumors Am J Med 1993 94 216 9 8381584 10.1016/0002-9343(93)90187-T
Natale RB Yagoda A Brown A Singer C Stover D Bajorunas D Combined Pneumocystis carinii and Nocardia asteroides pneumonitis in a patient with an ACTH-producing carcinoid Cancer 1981 47 2933 5 7020921
Fulkerson WJ Newman JH Endogenous Cushing's syndrome complicated by Pneumocystis carinii pneumonia Am Rev Respir Dis 1984 129 188 9 6608298
Cheng VC Yuen KY Chan WM Wong SS Ma ES Chan RM Immunorestitution disease involving the innate and adaptive response Clin Infect Dis 2000 30 882 92 10880300 10.1086/313809
Wislez M Bergot E Antoine M Parrot A Carette MF Mayaud C Cadranel J Acute respiratory failure following HAART introduction in patients treated for Pneumocystis carinii pneumonia Am J Respir Crit Care Med 2001 164 847 51 11549544
Dean GL Williams DI Churchill DR Fisher MJ Transient clinical deterioration in HIV patients with Pneumocystis carinii pneumonia after starting highly active antiretroviral therapy: another case of immune restoration inflammatory syndrome Am J Respir Crit Care Med 2002 165 1670 12070070
Barry SM Lipman MC Deery AR Johnson MA Janossy G Immune reconstitution pneumonitis following Pneumocystis carinii pneumonia in HIV-infected subjects HIV Med 2002 3 207 11 12139660 10.1046/j.1468-1293.2002.00115.x
Koval CE Gigliotti F Nevins D Demeter LM Immune reconstitution syndrome after successful treatment of Pneumocystis carinii pneumonia in a man with human immunodeficiency virus type 1 infection Clin Infect Dis 2002 35 491 3 12145736 10.1086/341974
Cheng VC Hung IF Wu AK Tang BS Chu CM Yuen KY Lymphocyte surge as a marker for immunorestitution caused by Pneumocystis carinii pneumonia in non-HIV immunosuppressed hosts Eur J Clin Microbiol Infect Dis 2004 23 512 4 15141339 10.1007/s10096-004-1140-6
Cheng VC Yuen KY Wong SS Woo PC Ho PL Lee R Chan RM Immunorestitution diseases in patients not infected with HIV Eur J Clin Microbiol Infect Dis 2001 20 402 6 11476440 10.1007/s100960100507
Cheng VC Woo PC Lau SK Cheung CH Yung RW Yam LY Yuen KY Peripartum tuberculosis as a form of immunorestitution disease Eur J Clin Microbiol Infect Dis 2003 22 313 7 12743831 10.1007/s10096-003-0998-z
Cheng VC Ho PL Lee RA Chan KS Chan KK Woo PC Lau SK Yuen KY Clinical spectrum of paradoxical deterioration during antituberculosis therapy in non-HIV-infected patients Eur J Clin Microbiol Infect Dis 2002 21 803 9 12461590 10.1007/s10096-002-0821-2
Cheng VC Yam WC Woo PC Lau SK Hung IF Wong SP Cheung WC Yuen KY Risk factors for development of paradoxical response during antituberculosis therapy in HIV-negative patients Eur J Clin Microbiol Infect Dis 2003 22 597 602 14508660 10.1007/s10096-003-0998-z
Mansharamani NG Balachandran D Vernovsky I Garland R Koziel H Peripheral blood CD4 + T-lymphocyte counts during Pneumocystis carinii pneumonia in immunocompromised patients without HIV infection Chest 2000 118 712 20 10988193 10.1378/chest.118.3.712
Nevez G Raccurt C Vincent P Jounieaux V Dei-Cas E Pulmonary colonization with Pneumocystis carinii in human immunodeficiency virus-negative patients: assessing risk with blood CD4+ T cell counts Clin Infect Dis 1999 29 1331 2 10524989 10.1086/313478
Sing A Trebesius K Roggenkamp A Russmann H Tybus K Pfaff F Bogner JR Emminger C Heesemann J Evaluation of diagnostic value and epidemiological implications of PCR for Pneumocystis carinii in different immunosuppressed and immunocompetent patient groups J Clin Microbiol 2000 38 1461 7 10747126
Rabodonirina M Cotte L Boibieux A Kaiser K Mayencon M Raffenot D Trepo C Peyramond D Picot S Detection of Pneumocystis carinii DNA in blood specimens from human immunodeficiency virus-infected patients by nested PCR J Clin Microbiol 1999 37 127 31 9854076
Ribes JA Limper AH Espy MJ Smith TF PCR detection of Pneumocystis carinii in bronchoalveolar lavage specimens: analysis of sensitivity and specificity J Clin Microbiol 1997 35 830 5 9157136
| 15581427 | PMC539248 | CC BY | 2021-01-04 16:32:03 | no | BMC Pregnancy Childbirth. 2004 Dec 6; 4:23 | latin-1 | BMC Pregnancy Childbirth | 2,004 | 10.1186/1471-2393-4-23 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-4-781554117210.1186/1471-2407-4-78Research ArticleIdentification of a panel of tumor-associated antigens from breast carcinoma cell lines, solid tumors and testis cDNA libraries displayed on lambda phage Pavoni Emiliano [email protected] Paola [email protected] Andrea [email protected]ù Giorgia [email protected] Elisa [email protected] Stefano [email protected] Maria Luisa [email protected] Pasquale Ceratti Adolfo [email protected] Antonio [email protected] Maurizio [email protected] Enrico [email protected] Franco [email protected] Olga [email protected] Kenton Labs, c/o Sigma Tau, Pomezia (Rome), 00040, Italy2 Laboratorio di Immunologia, Reparto Immunologia dei Tumori, Istituto Superiore di Sanità, Roma, 00100, Italy3 Oncologia Medica, Dipartimento di Medicina Sperimentale e Patologia, Facoltà di Medicina e Chirurgia, Università di Roma "La Sapienza", Rome, 00100, Italy4 Dipartimento di Scienze Microbiologiche, Genetiche e Molecolari, Università di Messina, 98100, Italy2004 12 11 2004 4 78 78 18 6 2004 12 11 2004 Copyright © 2004 Pavoni 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
Tumor-associated antigens recognized by humoral effectors of the immune system are a very attractive target for human cancer diagnostics and therapy. Recent advances in molecular techniques have led to molecular definition of immunogenic tumor proteins based on their reactivity with autologous patient sera (SEREX).
Methods
Several high complexity phage-displayed cDNA libraries from breast carcinomas, human testis and breast carcinoma cell lines MCF-7, MDA-MB-468 were constructed. The cDNAs were expressed in the libraries as fusion to bacteriophage lambda protein D. Lambda-displayed libraries were efficiently screened with sera from patients with breast cancer.
Results
A panel of 21 clones representing 18 different antigens, including eight proteins of unknown function, was identified. Three of these antigens (T7-1, T11-3 and T11-9) were found to be overexpressed in tumors as compared to normal breast. A serological analysis of the 21 different antigens revealed a strong cancer-related profile for at least five clones (T6-2, T6-7, T7-1, T9-21 and T9-27).
Conclusions
Preliminary results indicate that patient serum reactivity against five of the antigens is associated with tumor disease. The novel T7-1 antigen, which is overexpressed in breast tumors and recognized specifically by breast cancer patient sera, is potentially useful in cancer diagnosis.
==== Body
Background
A recent development in tumor immunology is based on the idea that the immune system can distinguish between normal and tumor tissues. Various studies suggest that both the cellular and humoral components of the immune system are able to recognize tumors (see review of Lake et al.) [1]. The presence of natural antibodies against cancer cells in peripheral blood of tumor patients probably plays a protective role against tumor development. The latest advances in molecular techniques further support the existence of natural antibodies against cancer antigens. The SEREX approach, based on the serological screening of cDNA expression libraries generated from tumor tissues of various origin, led to the molecular definition of immunogenic tumor proteins (tumor-associated antigens, TAAs) based on their reactivity with autologous patient sera [2]. This type of screening of a cDNA expression library is quite a laborious procedure requiring the preparation of a large number of membrane filters blotted with bacteriophage plaques, which are then screened with sera from cancer patients, usually available in limited quantity. In contrast to SEREX, phage display strategy is based on the selection and enrichment of antigens displayed on the phage surface. A physical link between a displayed fusion protein and the DNA encoding for it makes this phage target selectable through affinity purification. Phage display technology has been successfully applied to the screening of cDNA libraries from different tumors using the antibody repertoire of cancer patients [3-6]. In these experiments different phage display systems were used. Some of the authors used the C-terminus of a filamentous phage minor protein pVI for expression of cDNA libraries from breast cancer cell lines T47D and MCF-7 [3] and from colorectal cancer cell line HT-29 [5]. However, the filamentous phage display system imposes some biological bias for the expression and display of fusion proteins, since a filamentous phage-based library displays only those recombinant proteins able to pass through the inner bacterial membrane during filamentous phage assembly. To overcome this potential problem the lytic bacteriophages T7 [4] and λ [6] were used. By using these latter systems, the phage capsid is assembled in the cytoplasm of bacteria and mature phage particles are released by cell lysis. For example, Hansen and co-workers in their studies screened a commercially available (Novagen) human breast cancer cDNA library cloned in T7 vector [4], identifying positive clones.
Usually cDNA libraries are generated as C-terminal fusions. When such a library is panned on a serum, the presence of a complex antibody repertoire gives to out-of-frame or antisense-derived cross-reactive short peptide sequences a good chance of being enriched. In our previous work [6] we designed a new-concept lambda vector for the display of cDNA-encoded protein fragments as fusion to the N-terminus of bacteriophage gpD, allowing us to overcome this obstacle. In this vector, phage clones display a given protein fragment on the phage surface only when the insert's correct reading frame matches that of gpD. The size of the cloned DNA fragments in our libraries was adjusted to an average of 200–300 base pairs, which is of a size reasonably sufficient to potentially encode for a protein domain. The vast majority of out-of-frame sequences of the above-mentioned length most probably contains at least one in-frame stop codon. Thus, these inserts are not expressed as D fusion, are consequently not displayed on the phage surface and cannot be selected. In such cases, phage capsid contains only wt gpD encoded by lambda genome D gene. The N-terminal display system greatly reduces the selection of artifactual peptides, in comparison with a C-terminal fusion library displayed on lambda ([7] and our unpublished data).
By employing the SEREX approach numerous tumor antigens from different human neoplasms were identified [8,9]. Analysis of TAA expression in tumor samples and normal tissue led to the identification of a group, called cancer/testis antigens. Members belonging to this family are aberrantly expressed in human cancers and only in normal testis, but not in other normal tissue. For this reason, in addition to tumor samples and tumor cell lines, testicular cDNA libraries are also a convenient source of antigens which can be identified by screening with sera derived from tumor patients [10,11].
In the present work we report the construction of lambda-displayed cDNA libraries from breast cancer cell lines MCF-7, MDA-MB-468, from human breast carcinomas and from human testis, generated according to an improved protocol. These libraries were screened by using sera from breast cancer patients. The list of 21 identified antigens contains eight proteins with still unknown functions. Three of the genes (T7-1, T11-3 and T11-9) were found to be overexpressed in tumors as compared to normal breast. Recognition by human sera of five of the selected antigens (T6-2, T6-7, T7-1, T9-21 and T9-27) was associated with cancer diagnosis.
Methods
Tissue and serum samples
Specimens of breast carcinoma and autologous sera from breast cancer patients (B81-B96) were obtained from M. G. Vannini Hospital, Rome. A panel of human sera from breast cancer patients B1-B20, B36-B80 was provided by the Division of Medical Oncology, Federico II University of Naples. All the human biological samples were obtained through informed consent.
Construction of λKM8, λKM10 vectors
λKM8 was constructed by cloning the oligonucleotide duplex KM46 5’-CTAGTCTCCTCAGCGGCCGCGGTTCCGGTTCTGGTTCCGGTTCTGGTTCCGGTTCTGGT-3’ and KM47 5’-GGCCACCAGAACCGGAACCAGAACCGGAACCAGAACCGGAACCGCGGCCGCTGAGGAGA-3’ into SpeI, NotI sites at the 5'-end of the D gene in λKM4 vector [6]. The resulting vector λKM8 maintains the unique SpeI and NotI sites and encodes for a GS linker between the fusion site and gpD, Figure 1.
The plasmid pKM7 is a derivative of pKM3 [6], which was obtained by cloning of the oligonucleotide duplex K52 5’-GACCGCGTTTGCCGGAACGGCAATCAGCATCGTTACTAGTTTATTAAGCGGCCGCTAAGTGAGTG-3’ K53 5’-AATTCACTCACTTAGCGGCCGCTTAATAAACTAGTAACGATGCTGATTGCCGTTCCGGCAAACGCG-3’ into pKM3 previously digested with RsrII and EcoRI restriction enzymes. pKM7 was digested with SpeI and NotI to obtain pKM9, by direct cloning of the oligonucleotide duplex KM48 5’-CTAGCGGTTCCGGTTCTGGTTCCGGTTCTGGTTCCGGTTCTGGCACTAGTCTCCTCAGC-3’ and KM49 5’-GGCCGCTGAGGAGACTAGTGCCAGAACCGGAACCAGAACCGGAACCAGAACCGGAACCG-3’. λKM10 was constructed by cloning pKM9, which was linearized by digestion with XbaI restriction enzyme, into the XbaI site of λDam15imm21nin5 [12]. The resulting vector λKM10 bears unique SpeI, NotI sites at the 3'-end of the D gene and encodes for a flexible GS linker between gpD and the cloned protein fragment, Figure 1.
RNA extraction
mRNA from breast carcinoma cell lines MCF-7 and MDA-MB-468 was isolated in a single step by QuickPrep Micro mRNA Purification Kit (Amersham Pharmacia Biotech, UK) according to manufacturer's instructions.
Tumor samples from breast carcinoma patients were obtained as surgical specimens and immediately frozen in liquid nitrogen. Total RNA was prepared by Total RNA Isolation System (Promega, Madison, WI) and purified to Poly A+ RNA using PolyATract mRNA Isolation Systems (Promega).
Total RNA from normal testis was purchased from Genpak, UK (# 061023). Total RNA from normal breast (pool of 3) was purchased from Stratagene, La Jolla, CA (# 735044).
cDNA library construction
From 1 to 5 μg of the purified poly(A)+ RNA from cell lines or human tissues were used to synthesize cDNA by random priming, using TimeSaver cDNA Synthesis Kit (Amersham Pharmacia Biotech, Piscataway, NJ, USA). RNasin Ribonuclease Inhibitor (Promega) was added to first-strand synthesis reaction.
A mixture of the following oligonucleotides (130 pmol): K64 (5'-GCGGCCGCTGGNNNNNNNNN-3'), K79 (5'-GCGGCCGCTGGCNNNNNNNNN-3'), and K81 (5'-GCGGCCGCTGGCANNNNNNNNN-3') was used for priming. They all carry a NotI site (underlined) at their 5' end, and a random sequence of nine nucleotides at their 3' end, positioned in the three possible reading frames. The second strand was synthesized by nick translation according to the manufacturer's instructions.
One hundred ng of ds cDNA were randomly primed with 25 pmol of oligonucleotide K56 (5'-GGCCGGCCAACNNNNNNNNN-3'), constituted by a constant sequence at the 5' end, and a random 3'sequence. The reaction mixture was purified by QIAgen QIAquick columns.
Approximately 0,2 ng of the above randomly primed ds cDNA was amplified by PCR with biotinylated primers: K59 (bio-5'-GCACTAGTGGCCGGCCAAC-3'), K60 (bio-5'-GCACTAGTCGGCCGGCCAAC-3'), K61 (bio-5'-GCACTAGTCGGGCCGGCCAAC-3') and K65 (bio-5'-GGAGGCTCGAGCGGCCGCTGG-3'). K59, K60 and K61 carry the same constant sequence of K56 positioned in the three possible frames with respect to a SpeI site (underlined) allowing directional cloning. K65 carries a NotI site (underlined), that anneals to the 5' end of the reverse strand of cDNA.
PCR product was purified with QIAquick PCR purification kit (QIAGEN, Germany), filtered by Microcon-100 columns (Millipore, Bedford, MA) to reduce the number of small fragments and additionally fractionated by 6% PAGE. DNA smear, corresponding to 300–1000 base-pair fragments, was cut and eluted from gel according to standard procedure [13].
After digestion with SpeI and NotI enzymes, in order to remove the biotinylated extremities and uncut fragments, a 20-minute incubation with streptavidin M-280 Dynabeads (DYNAL, Norway) was performed. After additional filtration on Microcon-100 the insert was cloned in λKM8 or λKM10 vectors.
The vector was digested with SpeI, NotI enzymes and dephosphorylated. For each library 5 ligation mixtures, each one containing 0.5 μg of vector and about 3 ng of insert, were performed. After overnight incubation at 4°C the ligation mixtures were packaged in vitro by lambda packaging extract (Stratagene, La Jolla, CA). BB4 cells were infected by lambda and plated in top-agar on 100 (15 cm) NZY plates. After overnight incubation phages were eluted from the plates with SM buffer, purified, PEG/NaCl precipitated [13] and stored at -80°C in SM buffer, 7% DMSO.
Affinity selection
Two μl of human serum were preincubated with 10 μl of BB4 bacterial extract and 10 μl of UV-killed lambda phage in 1 ml of blocking buffer (3% BSA, 1X PBS, 10 mM MgSO4, 1% Triton) for 30 minutes at 37°C under gentle agitation. 1010 pfu of lambda library were then added to the preincubated mixture for a further incubation of 1 hr. Magnetic beads (100 μl), linked to Protein A (Dynabeads Protein-A, Dynal, Norway) were washed twice with the blocking solution. Mixture of library with serum was incubated with the beads for 10 min at RT under agitation. The beads were washed 10 times with 1 ml of washing solution (1X PBS, 1% Triton, 10 mM MgSO4). The bound phages were recovered by infection of 600 μl BB4 cells added directly to the beads. After a 20-minute incubation 10 ml of molten NZY-top agar (48°C) was added to the mixture of beads with infected cells and immediately poured onto NZY plates (15 cm). Next day the phage particles were harvested by incubation of the plates under agitation with 15 ml of SM buffer for 4 hours at 4°C. The phage particles were purified by PEG/NaCl precipitation and stored in 1/10 of initial volume of SM with 0.05% NaN3 at 4°C.
Analysis of gene expression by PCR
Five hundred ng of poly(A)+ RNA from breast carcinomas or normal tissue were used to synthesize full-length cDNA by SMART cDNA library construction kit (Clontech, Palo Alto, CA). For maximum sensitivity specific primers for the different genes were designed to amplify sequences located near the 3' end of gene's transcript. Twenty-five cycles of PCR were performed from 1 μl of each cDNA template, normalized through PCR amplification of the β-actin gene.
Results
Construction of the libraries
Lambda libraries were constructed by directional cloning of randomly primed cDNA from human breast carcinoma cell lines MCF-7 and MDA-MB-468, from human breast carcinomas or from human testis into the phage display vector λKM8 to generate fusions with the N-terminus of gpD (see list of libraries in Table 1). Only library T6 was built like C-terminal fusions with protein D by cloning cDNA into λKM10 vector. λKM8 and λKM10 are derivatives of λKM4 vector [6] obtained by introducing a flexible GS-linker between the displayed protein and gpD (Fig. 1). The insert size in the majority of the clones in the libraries ranged from 100 to 400 bp (Fig. 2). Only a tiny fraction of out-of-frame clones of this length do not contain stop codons, and are therefore displayed in the libraries constructed as N-terminus fusions, thus greatly reducing the probability of the selection of mimotopes.
Selection of tumor-associated antigens
The scheme of TAA identification is shown in Figure 3. Typically, one or two rounds of biopanning, performed according to the selection protocol described in Materials and Methods, were sufficient to obtain 2–50% of positive clones in the following immunoscreening procedure. Then, the identified phage clones were tested with a panel of positive and negative human sera by picking the clones in arrayed order on the bacterial lawn, blotting onto nitrocellulose membrane and probing with a number of different sera as previously described [6]. The nucleotide sequences of 21 clones that exhibited specific or preferential reactivity with sera from breast tumor patients as compared to sera from healthy donors were identified, and their nucleotide sequences were determined (Table 2 [see Additional file 1]).
Serological analysis of tumor antigens
Phage lysates were prepared from all the selected clones as previously described [6] and tested in ELISA first with a collection of negative, and subsequently, with positive sera (Table 2 [see Additional file 1]). All the antigens tested reacted exclusively or preferentially with sera from breast cancer patients. Eight of the antigens reacted only with the patient serum used in the corresponding selection. Five antigens had cancer-related profile of reactivity, P < 0.05 (T6-2, T6-7, T7-1, T9-21 and T9-27). The other antigens either reacted with a low percentage of cancer sera, or the total panel of the tested sera was too small to offer any clear conclusion.
Sequence analysis of selected cDNA clones
Twenty-one positive clones were found to encode fragments from 18 different gene products, such as 4 clones (T5-9, T9-21, T9-27, T11-7) showing homology to different regions of the same reverse transcriptase gene (Figure 4). Most of the clones correspond to known gene products in the correct orientation and reading frame, with the exception of clone T5-18 encoding myc oncogen in an alternative frame. Several of these known gene products, such as reverse transcriptase homolog (clones T5-9, T9-21, T9-27, T11-7), protein kinase C-binding protein (T6-1), trap ankyrin repeat (T11-3), heat shock protein apg-2 (T11-13), have been previously identified by SEREX [9,14-17]. Eight of the sequences listed in Table 2 [see Additional file 1] encode for proteins with unknown functions.
Cancer-specific expression of selected tumor antigens
Expression patterns for several of the selected genes were analyzed by semi-quantitative PCR from SMART cDNA template. It has been previously shown [18], by comparing the expression level of target genes in SMART PCR-amplified cDNAs and their corresponding total RNAs, that SMART cDNA accurately reflects gene expression patterns found in total RNA. We normalized the panel of cDNAs from ten different breast carcinomas, one metastasized lymph node, normal breast, normal testis and peripheral blood lymphocytes from healthy donors, by PCR amplification of a housekeeping gene, β-actin (Figure 5). Three of the identified antigens, fucosyltransferase (T6-7), Zinc finger protein 258 (T11-6), and p53-binding protein (T1-52) [6], were ubiquitously expressed in all the tumor and normal tissue samples tested (Figure 5A). Some of the antigens, T5-15 (KIAA1735), T5-13 (Sos1), T11-5 (hypothetical protein MGC4170) were found to be downregulated in many tumors (Figure 5B). T11-9 (hypothetical protein AF225417) was overexpressed in 50% of the primary tumors and the unique metastasized lymph node tested. T11-3 (trap ankyrin repeat) was overexpressed in most of the tumors tested in comparison with normal breast, although it was also transcribed in testis and normal lymphocytes (Figure 5C). T7-1 (KIAA1288) was found to be overexpressed in 50% of the primary breast carcinomas and in the metastasis specimen tested. In order to obtain an evaluation of the accuracy of the method used for the analysis of gene expression, we performed PCR amplification of neu/HER2, a known tumor marker overexpressed in breast cancer. We observed that neu/HER2 is overexpressed in 2 primary tumors among the 7 tested (≈29%) in accordance with the literature on breast carcinoma [19,20].
Discussion
In the present study we report the construction of MCF-7 and MDA-MB-468 cell lines, breast carcinoma and testis cDNA phage-displayed libraries expressed as fusions to bacteriophage lambda gpD. The new phage vectors bear a flexible GS linker between the cloned protein domain and protein D, so as to facilitate lambda head assembly. Moreover, a new efficient protocol to synthesize cDNA was applied. We primed cDNA synthesis on mRNA template with random oligonucleotides containing a constant 5' end. After complete synthesis of double-stranded cDNA, a second round of random priming was applied to generate oriented fragments of cDNA suitable for library construction. This protocol, in comparison with previous version, increases the presence of authentic protein domains in the library twofold, because of correct cDNA orientation. Moreover, some of the clones isolated from our previous libraries were results of chimerical fusion of two or more different genes, generated through double random priming on ds cDNA template. The new protocol has reduced this problem significantly.
We also confirmed the advantage of N-terminal fusion for domain library construction in phage display vectors for screening with sera, because a significant amount of false positive cross-reactive clones, containing stop codons downstream of the fusion site giving rise to short mimotope sequences, were selected from the C-terminal fusion library (T6). Only 4 clones with specific tumor-related reactivity were isolated from the T6 library. However, C-terminal fusion might allow efficient display and selection for some antigenic C-terminal protein domains. In fact, the C-terminal fragment of fucosyltransferase (clone T6-7) was isolated from the T6 library.
The panel of selected TAAs in Table 2 [see Additional file 1] contains several functionally defined gene products, previously unknown as tumor antigens. AKAP450 and Sos1 proteins, corresponding to clones T5-8 and T5-13, are intracellular components of the signal transduction pathway. Sos1 is a well-known guanine nucleotide exchange factor for Ras oncogene [21]. Transgenic mice expressing a dominant form of Sos in basal keratinocytes develop skin papillomas with 100% penetrance [22]. Moreover, a Sos1 mutant, lacking four functionally important proline-rich (SH3 binding) regions was reported to be responsible for gingival fibromatosis [23]. AKAP450 is a member of the A-kinase anchor proteins family. It is located in the centrosome [24], and acts as a microtubule nucleation site [25] and as a scaffold for proteins involved in mitotic process [26].
Other selected antigens with known or predicted intracellular location are alpha-6-fucosyltransferase (clone T6-7) and zinc finger protein 258 (ZNF258, clone T11-6). Alpha-fucosyltransferase catalyzes the transfer of GDP-fucose to oligosaccharide chains linked to proteins, lipids and sugars [27] and resides in the luminal compartment of trans-Golgi vesicles [28]. The predicted protein product ZNF258 contains zinc-binding motif repeats [29]. If ZNF258, together with structural homology, also shares biological properties with zinc finger proteins, thus recognizing and interacting with DNA, it should have a nuclear localization. The presence on our antigen list of proteins with predicted intracellular residence is in agreement with findings from the other groups [30,5] and is related to possible tissue necrosis and cell lysis associated with tumor growth.
The human myc oncogene is transcribed from four alternative promoters giving rise to mRNAs for Myc1, Myc2, MYCHEX1 and 5'ORF [31]. Clone T5-18 is the result of the translation of an alternative frame to Myc1, Myc2 and does not correspond to any known protein product of myc oncogene transcription. It is not clear whether selection of this clone is an artifact of the experiment, or the result of an aberrant genome rearrangement in the tumor cells used for library construction.
Among isolated antigens, there are 4 clones (T5-9, T9-21, T9-27 and T11-7) having between 55–91% sequence identity with that of a reverse transcriptase homolog (Figure 4). Viral antigens corresponding to human endogenous retrovirus were previously isolated from renal cancers and melanomas by SEREX [9]. It is interesting to note that all these clones, isolated with sera from breast cancer patients, derive from libraries constructed with cDNA from every different origin utilized: i.e. cell lines (T5), solid tumor (T9), testis (T11). We have no explanation for the transcription of reverse transcriptase gene in normal testis tissue.
Eight proteins in the tumor antigen panel are unknown, or hypothetical proteins with unknown functions (T5-2, T5-15, T5-19, T6-2, T6-6, T7-1, T11-5, T11-9). The four underlined gene products from the list in parenthesis were analyzed for mRNA expression in tumors and normal breast tissue. The mRNA expression levels were analyzed by PCR from SMART-cDNA template in 7–10 breast cancer specimens, one metastasized lymph node, normal breast, testis and peripheral lymphocytes from healthy donors. Two of these 4 unknown antigens and T11-3 were found to be frequently overexpressed in breast cancer. In particular clone T7-1, which was classified as encoding for an unknown protein since it has 100% identity only with KIAA1288 from EST database, was found to be overexpressed in breast carcinomas. This finding, together with the good reactivity of T7-1 protein with sera from tumor patients, identifies this antigen among the most promising targets for diagnosis of the disease.
In contrast to the other antigens, which are overexpressed in breast cancer, mRNAs of T5-13, T5-15 and T11-5 appear to be underexpressed in 50–90% of breast cancer specimens, in comparison with normal breast tissue. How the immune system succeeds in responding to such antigens is still not clear. However, this finding is common to several SEREX-defined antigens, such as LU-12 [32], REN-9, REN-10 [33] and BR-41 [15], representing a group of TAAs deleted or downregulated in tumors. Lu-12, REN-9, Ren-10 map within cancer tumor suppressor gene locus at chromosome 3p21.3, a region often deleted in small cell lung cancer as well as in renal cancer. Downregulated antigen BR-41 was identified as SNT-1, a membrane-associated adaptor protein interacting with Sos1 [15]. In the present work we show that Sos1 (T5-13) is also downregulated in 50% of breast cancer samples. The downregulated antigens T11-5, T5-13 (Sos1), and T5-15 do not react with sera from patients B82-B96 analyzed for tumor mRNA expression. Furthermore, tumor biopsies from patients with good response for these antigens were not available for expression analysis. Thus, at present, it is not possible to determine whether T11-5, T5-13 (Sos1), and T5-15 are normally expressed, or downregulated, in patients showing an immune response for the corresponding antigen.
Sequence comparison of T11-5 and T5-15 clones with the EST database revealed identity with the hypothetical proteins MGC4170 and KIAA1735. We have derived the aa sequence of the corresponding ORFs and predicted the whole sequence architecture by computer analysis using the SMART program [34,35]). MGC4170 encodes for two NL domains, while KIAA1335 encodes for a 389 aa protein bearing a DIX domain at the carboxy-terminus. The presence of such structural domains indicates that both of these still unknown proteins (corresponding to clones T11-5 and T5-15), which we found downregulated in several tumor specimens, may be involved in the signal transduction machinery.
In spite of the fact that several promising antigens were identified from cDNA library constructed from testis mRNA, none of the antigens derived from testis or other libraries could be classified as specific testis/cancer (CT) antigen, because of their low expression in testis (T11-9, T7-1) or expression in other tissues as well.
In this work we analyzed the frequency of the immune response to the 21 identified antigens by using a panel of sera from tumor patients and healthy donors. In general, we observed a low frequency of serum reactivity with the antigens, which was expected and is similar to that of the vast majority of SEREX-identified clones [36]. A significant number of sera from tumor patients, in comparison with healthy individuals, efficiently recognized five of the identified antigens (T6-2, T6-7, T7-1, T9-21, T9-27). Clones T9-21 and T9-27, isolated from breast carcinoma library, respectively show 70% homology (55% identity) and 62% homology (68% identity) to reverse transcriptase homolog (PO8547). T7-1 is a protein having an unknown function, which was found to be overexpressed in breast carcinoma.
Taken together, these results lead us to believe that analysis of a complex panel of serologically-defined TAAs, with very large panels of sera from patients classified according to clinical parameters, i.e., age of patient, stage, extent and outcome of disease, etc. could lead to a much clearer understanding of the role, specificity and significance of the immune response versus disease in cancer patients.
Conclusions
We demonstrated that a lambda display-based approach permits the efficient identification of tumor antigens, potential immunological targets in breast cancer. The list of 21 antigens identified in this work contains eight proteins of still unknown function. Three of the genes (T7-1, T11-3, T11-9) were found to be overexpressed in tumors as compared to normal breast. Five of the selected antigens (T6-2, T6-7, T7-1, T9-21, T9-27) were recognized specifically by breast cancer patient sera.
List of abbreviations
aa, amino acids; Ag, antigen; EST, expressed sequence tags; ds cDNA, double-stranded cDNA; SEREX, serological identification of antigens by recombinant expression cloning; PAGE, polyacrylamide gel electrophoresis; PEG, polyethylene glycol; pfu, plaque-forming units; Sos1, son of sevenless homolog 1; TAA, tumor-associated antigen.
Competing interests
EP, PV, AP, GM, EB, FF and OM are salaried by an organization holding two patents relating to the content of the manuscript.
Authors' contributions
EP and PV contributed equally to the work. EP, PV and AP carried out cDNA library experiments, recombinant protein production, immunoassays and database search. GM and EB performed selection experiments. SB, MLD and MC contributed to immunological analysis of tumor antigens. ADPC and AL performed clinical studies. EC coordinated all medical aspects of the work. OM planned and performed molecular biology experiments, and teamed with FF in design and coordination of the entire project. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Table 2. List of identified TAAs. The File is given in Microsoft Word format. The table contains information about selected antigens.
Click here for file
Acknowledgements
We wish to thank Dr. Nicola Gargano and Dr. Manlio Di Cristina for very insightful critical discussion during this work, and Mr. Luca Bruno for excellent technical assistance. We also thank Ms. M. Deutsch for the linguistic revision of the text.
Figures and Tables
Figure 1 Cloning sites of λKM8 and λKM10 vectors.
Figure 2 Insert length distribution. Forty-eight random clones from T5 library were amplified by PCR using a couple of primers on the sides of insert. Size of inserts was calculated according to their electrophoretic mobility in 3% agarose gel.
Figure 3 Scheme of selection strategy leading to TAA identification. A phage-displayed tumor cDNA library is preincubated with patient serum. TAA-specific antibodies bind to antigens exposed on the phage surface. Abs-phage complex is captured by protein A-coated solid support (ELISA plate or dynebeads). Non-bound phage are washed away. Bound phage are eluted by infection of added bacteria and amplified. Positive clones are isolated by immunoscreening procedure and then picked in ordered array on a bacterial lawn, transferred to nitrocellulose membrane and probed with different positive and negative sera.
Figure 4 Four identified antigen sequences with partial homology to reverse transcriptase homolog. Peptide sequence is reported in single-letter code. Identical amino acids in the selected clones and reverse transcriptase homolog are represented by a dash. These clones were isolated from libraries of different origin. Clones T9-21 and T9-27, isolated from solid tumor library, had significantly high frequency of reactivity with sera from breast cancer patients.
Figure 5 cDNA-PCR analysis of gene expression was done using specific sequence primers. We used SMART cDNAs from 7–10 different tumor samples (patients B84, B85, B87, B89, B90, B91, B92, B93, B95, B96) as template, from single metastasized lymph node indicated as LMB82 (patient B82) and from normal breast, normal testis, lymphocytes from healthy donors. cDNAs were normalized by amplification of β-actin gene. There are agarose gels with ubiquitously-expressed genes in Figure 5A, underexpressed genes in Figure 5B, overexpressed ones in Figure 5C.
Table 1 Lambda display libraries list.
Library name Point of fusion with gpD Source of cDNA Library complexity
T5 N-terminus MCF-7 + MDA-MB-468 1.7 × 107
T6 C-terminus MCF-7 + MDA-MB-469 3.4 × 107
T7 N-terminus Human breast carcinoma (patient B81) 1.5 × 106
T9 N-terminus Human breast carcinoma (patient B84) 2.3 × 107
T11 N-terminus Human testis 1.3 × 107
==== Refs
Lake DF Huynh WC Hersh EM Natural and induced human antibody response to cancer Cancer investigation 2000 18 480 489 10834032
Sahin U Tureci O Schmitt H Cochlovich B Johannes T Schmits R Stenner F Luo G Schobert I Pfreundschuh M Human neoplasms elicit multiple specific immune responses in the autologous host Proc Natl Acad Sci USA 1995 92 11810 11813 8524854
Sioud M Hansen MH Profiling the immune response in patients with breast cancer by phage-displayed cDNA libraries Eur J Immunol 2001 31 716 725 11241275 10.1002/1521-4141(200103)31:3<716::AID-IMMU716>3.0.CO;2-9
Hansen MH Ostenstad B Sioud M Identification of immunogenic antigens using a phage-displayed cDNA library from an invasive ductal breast carcinoma tumor Int J Oncol 2001 19 1303 1309 11713604
Somers VA Brandwijk RJ Joosten B Moerkerk PT Arends J-W Menheere P Pieterse WO Claessen A Scheper RJ Hoogenboom HR Hufton SE A panel of candidate tumor antigens in colorectal cancer revealed by the serological selection of a phage displayed cDNA expression library J Immunol 2002 169 2772 2780 12193752
Minenkova O Pucci A Pavoni E De Tomassi A Fortugno P Gargano N Cianfriglia M Barca S De Placido S Martignetti A Felici F Cortese R Monaci P Identification of tumor-associated antigens by screening phage-displayed human cDNA libraries with sera from tumor patients Int J Cancer 2003 106 534 544 12845649 10.1002/ijc.11269
Santi E Capone S Mennuni C Lahm A Tramontano A Luzzago A Nicosia A Bacteriophage lambda display of complex cDNA libraries: a new approach to functional genomics J Mol Biol 2000 296 497 508 10669604 10.1006/jmbi.1999.3471
Pfreundschuh M Exploitation of the B cell repertoire for identification of human tumor antigens Cancer Chemother Pharmacol 2000 46 S3 S7 10950139
Chen Y-T Cancer vaccine: identification of human tumor antigens by SEREX Cancer J Sci Am 2000 3 S208 217
Tureci O Sahin U Zwick C Koslowski M Seitz G Pfreundschuh M Identification of a meiosis-specific protein as a member of the class of cancer/testis antigens Proc Natl Acad Sci USA 1998 95 5211 5216 9560255 10.1073/pnas.95.9.5211
Jager D Unkelbach M Frei C Bert F Scanlan MJ Jager E Old LJ Chen Y-T Knuth A Identification of tumor-restricted antigens NY-BR-1, SCP-1, and a new cancer/testis-like antigen NW-BR-3 by serological screening of a testicular library with breast cancer Cancer Immunity 2002 2 5 12747750
Sternberg N Hoess RH Display of peptides and proteins on the surface of bacteriophage λ Proc Natl Acad Sci USA 1995 92 1609 1613 7878027
Sambrook J Fritsch EF Maniatis T Molecular cloning: A laboratory manual 1989 Cold Spring Harbor Laboratory Press, Cold Spring Harbor
Eichmuller S Usener D Dummer R Stein A Thiel D Schadendorf D Serological detection of cutaneous T-cell lymphoma-associated antigens Proc Natl Acad Sci USA 2001 98 629 634 11149944 10.1073/pnas.021386498
Scanlan MJ Gout I Gordon CM Williamson B Stockert E Gure AO Jager D Chen YT Mackay A O'Hare MJ Old LJ Humoral immunity to human breast cancer: antigen definition and quantitative analysis of mRNA expression Cancer Immun 2001 1 4 12747765
Nakatsura T Senju S Yamada K Jotsuka T Ogawa M Nishimura Y Gene cloning of immunogenic antigens overexpressed in pancreatic cancer Biochem Biophys Res Commun 2001 281 936 944 11237751 10.1006/bbrc.2001.4377
Scanlan MJ Chen YT Williamson B Gure AO Stockert E Gordan JD Tureci O Sahin U Pfreundschuh M Old LJ Characterization of human colon cancer antigens recognized by autologous antibodies Int J Cancer 1998 76 652 658 9610721
Zhumabayeva B Diatchenko L Chenchik A Siebert PD Use of SMART-generated cDNA for gene expression studies in multiple human tumors Biotechnique 2001 30 158 163
Thomas E Berner G Prognostic and predictive implications of HER2 status for breast cancer patients Eur J Oncol Nurs 2000 4 10 17 12849612 10.1054/ejon.2000.0073
Latta EK Tjan S Parkes RK O'Malley FP The role of HER2/neu overexpression/ amplification in the progression of ductal carcinoma in situ to invasive carcinoma of the breast Mod Pathol 2002 15 1318 1325 12481013 10.1097/01.MP.0000038462.62634.B1
Chardin P Camonis JH Gale NW van Aelst L Schlessinger J Wigler MH Bar-Sagi D Human Sos1: a guanine nucleotide exchange factor for Ras that binds to GRB2 Science 1993 260 1338 1343 8493579
Sibilia M Fleischmann A Behrens A Stingl L Carroll J Watt FM Schlessinger J Wagner EF The EGF receptor provides an essential survival signal for SOS-dependent skin tumor development Cell 2000 102 211 220 10943841 10.1016/S0092-8674(00)00026-X
Hart TC Zhang Y Gorry MC Hart PS Cooper M Marazita ML Marks JM Cortelli JR Pallos D A mutation in the SOS1 gene causes hereditary gingival fibromatosis type 1 Am J Hum Genet 2002 70 943 954 11868160 10.1086/339689
Witczak O Skalhegg BS Keryer G Bornens M Tasken K Jahnsen T Orstavik S Cloning and characterization of a cDNA encoding an A-kinase anchoring protein located in the centrosome, AKAP450 EMBO J 1999 18 1858 1868 10202149 10.1093/emboj/18.7.1858
Takahashi M Yamagiwa A Nishimura T Mukai H Ono Y Centrosomal proteins CG-NAP and kendrin provide microtubule nucleation sites by anchoring gamma-tubulin ring complex Mol Biol Cell 2002 13 3235 3245 12221128 10.1091/mbc.E02-02-0112
Sillibourne JE Milne DM Takahashi M Ono Y Meek DW Centrosomal anchoring of the protein kinase CK1delta mediated by attachment to the large, coiled-coil scaffolding protein CG-NAP/AKAP450 J Mol Biol 2002 322 785 797 12270714 10.1016/S0022-2836(02)00857-4
Costache M Cailleau A Fernandez-Mateos P Oriol R Mollicone R Advances in molecular genetics of alpha-2- and alpha-3/4-fucosyltransferases Transfus Clin Biol 1997 4 367 382 9269717
Paulson JC Colley KJ Glycosyltransferases: Structure, localization, and control of cell type-specific glycosylation J Biol Chem 1989 264 17615 17618 2681181
Smedley D Hamoudi R Lu YJ Cooper C Shipley J Cloning and mapping of members of the MYM family Genomics 1999 60 244 247 10486218 10.1006/geno.1999.5918
Sahin U Tureci O Pfreundschuh M Serological identification of human tumor antigens Curr Opin Immunol 1997 9 709 716 9368781 10.1016/S0952-7915(97)80053-2
Nanbru C Prats AC Droogmans L Defrance P Huez G Kruys V Translation of the human c-myc P0 tricistronic mRNA involves two independent internal ribosome entry sites Oncogene 2001 20 4270 4280 11464293 10.1038/sj.onc.1204548
Gure AO Altorki NK Stockert E Scanlan MJ Old LJ Chen YT Human lung cancer antigens recognized by autologous antibodies: definition of a novel cDNA derived from the tumor suppressor gene locus on chromosome 3p21.3 Cancer Res 1998 58 1034 1041 9500467
Scanlan MJ Gordan JD Williamson B Stockert E Bander NH Jongeneel V Gure AO Jager D Jager E Knuth A Chen YT Old LJ Antigens recognized by autologous antibody in patients with renal-cell carcinoma Int J Cancer 1999 83 456 464 10508479
Schultz J Milpetz F Bork P Ponting CP SMART, a simple modular architecture research tool: identification of signaling domains Proc Natl Acad Sci USA 1998 95 5857 5864 9600884 10.1073/pnas.95.11.5857
Letunic I Goodstadt L Dickens NJ Doerks T Schultz J Mott R Ciccarelli F Copley RR Ponting CP Bork P Recent improvements to the SMART domain-based sequence annotation resource Nucleic Acids Res 2002 30 242 244 11752305 10.1093/nar/30.1.242
Stockert E Jager E Chen YT Scanlan MJ Gout I Karbach J Arand M Knuth A Old LJ A survey of the humoral immune response of cancer patients to a panel of human tumor antigens J Exp Med 1998 187 1349 1354 9547346 10.1084/jem.187.8.1349
| 15541172 | PMC539249 | CC BY | 2021-01-04 16:03:02 | no | BMC Cancer. 2004 Nov 12; 4:78 | utf-8 | BMC Cancer | 2,004 | 10.1186/1471-2407-4-78 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-4-851556085010.1186/1471-2407-4-85Research ArticleBAG-1 haplo-insufficiency impairs lung tumorigenesis Götz Rudolf [email protected] Boris W [email protected] Guadalupe [email protected] Ulf R [email protected] Institut für Medizinische Strahlenkunde und Zellforschung (MSZ), Universität Würzburg, Versbacher Straße 5, D-97078 Würzburg, Germany2 Universitäts-Kinderklinik Würzburg, Josef-Schneider-Str. 2, D-97080 Würzburg, Germany2004 24 11 2004 4 85 85 4 8 2004 24 11 2004 Copyright © 2004 Götz 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
BAG-1 is a multifunctional co-chaperone of heat shock proteins (Hsc70/Hsp70) that is expressed in most cells. It interacts with Bcl-2 and Raf indicating that it might connect protein folding with other signaling pathways. Evidence that BAG-1 expression is frequently altered in human cancers, in particular in breast cancer, relative to normal cells has been put forward but the notion that overexpression of BAG-1 contributes to poor prognosis in tumorigenesis remains controversial.
Methods
We have evaluated the effect of BAG-1 heterozygosity in mice in a model of non-small-cell lung tumorigenesis with histological and molecular methods. We have generated mice heterozygous for BAG-1, carrying a BAG-1 null allele, that in addition express oncogenic, constitutively active C-Raf kinase (SP-C C-Raf BxB) in type II pneumocytes. SP-C C-Raf BxB mice develop multifocal adenomas early in adulthood.
Results
We show that BAG-1 heterozygosity in mice impairs C-Raf oncogene-induced lung adenoma growth. Lung tumor initiation was reduced by half in BAG-1 heterozygous SP-C C-Raf BxB mice compared to their littermates. Tumor area was reduced by 75% in 4 month lungs of BAG-1 haploinsufficient mice compared to mice with two BAG-1 copies. Whereas BAG-1 heterozygosity did not affect the rate of cell proliferation or signaling through the mitogenic cascade in adenoma cells, it increased the rate of apoptosis.
Conclusion
Reduced BAG-1 expression specifically targets tumor cells to apoptosis and impairs tumorigenesis. Our data implicate BAG-1 as a key player in oncogenic transformation by Raf and identify it as a potential molecular target for cancer treatment.
==== Body
Background
BAG-1 is a multifunctional protein that is expressed in most cells. Originally identified as a Bcl-2 binding protein [1], other interaction partners of BAG-1 were described, including the serine threonine kinase C-Raf [2]. The C-terminal "BAG domain" of BAG-1 mediates the interaction with the Hsc70 and Hsp70 heat shock proteins [3], molecular chaperones that bind proteins in non-native states assisting them to reach a functional active conformation [4]. BAG-1 acts as a nucleotide exchange factor in this activation cycle [3]. The above findings indicated that BAG-1 might connect protein folding with other signaling pathways. Signaling networks promoting cell growth and proliferation are frequently deregulated in cancer [5]. The classical mitogenic cascade transmits stimuli from growth factor receptors via Ras, Raf, MEK and ERK to the cell nucleus [6]. C-Raf, like A- and B-Raf kinases also act at the outer membrane of mitochondria to augment cell survival [7,8]. Previously we had observed the stimulation of C-Raf kinase activity by BAG-1 in vitro [2]. Ras and B-Raf mutations have been found in various human cancers [9,10]. Evidence that BAG-1 expression is frequently altered in human cancers, in particular in breast cancer, relative to normal cells has been put forward but the notion that overexpression of BAG-1 contributes to poor prognosis in tumorigenesis remains controversial [11].
Methods
Animals
Mice used in these studies were generated and maintained according to protocols approved by the animal care and use committee at University of Würzburg. To inactivate the BAG-1 gene, we constructed a vector where exons 1 and 2 are replaced with a neomycin resistance gene. A phage clone with a 15-kb genomic insert from mouse strain 129/Sv spanning all seven exons of BAG-1 was identified and characterised using standard methods. The targeting construct contained 1,1-kb from the BAG-1 locus upstream of the neomycin resistance gene of plasmid pPNT [12] and 6-kb downstream. The upstream arm of 1,1 kb is located 5' to the start codon in the first exon of BAG-1 and the 3' arm of 6 kb is located downstream of exon 2. The mutation was introduced into embryonic stem cells by homologous recombination. Positive clones were identified by Southern blot analysis. Germline transmitting chimeras were obtained and bred to C57BL/6 mice. Further details will be described elsewhere. Heterozygous BAG-1 mice were genotyped by a PCR assay. The targeted BAG-1 allele was detected with primers P1 (5'-GAG TCT CCC GAT CCC TTT TCC), located upstream of exon 1, and P2 (5'-GAT TCG CAG CGC ATC GCC TT), located in the neomycin resistance gene, yielding a product of 600 base pairs. BAG-1 heterozygous mice were backcrossed at least three times onto C57BL/6 background before crossing with SP-C C-Raf BxB mice. Lung tumour mice expressing oncogenic C-Raf BxB were backcrossed at least six times onto C57BL/6 background.
Western blot
For the analysis of BAG-1 expression, lung lysates of the indicated genotypes were separated on 12,5% polyacrylamide-SDS (sodium dodecyl sulphate) gel, transferred to nitrocellulose Protran BA83 membrane (Schleicher&Schüll) and probed with rabbit anti-BAG-1 (FL-274) antibody (1:250, Santa Cruz Biotechnology). Amounts of protein were determined by Bradford protein assay to ensure equal protein loading for the analysis. Blots were developed using the appropriate horseradish peroxidase coupled secondary antibody and the ECL system (Amersham Pharmacia Biotech). Subsequently, the membrane was stripped and reprobed with rabbit antibody to glyceraldehyde 3 phosphate dehydrogense (1:2000, ab9485, Abcam Ltd.).
Histopathology and immunohistochemistry
Animals were sacrificed and lungs were fixed under 25 cm water pressure with 4% paraformaldehyde and embedded in paraffin. 5 μm sections were stained with hematoxylin and eosin and analysed. Pictures were taken using a Leica DMLA microscope and a Hitachi HV-C20A colour camera. Immunohistochemical staining to detect activated caspase-3, phospho-ERK (extracellular signal-regulated kinase), PCNA (proliferating cell nuclear antigen) have been described elsewhere [13]. Apoptotic, PCNA and p-ERK indices were determined by evaluating randomly chosen adenomas or fields of normal lung in 3–4 sections and determining the percentage of positive cells per 2000 cells at ×400.
Results and discussion
BAG-1 heterozygosity impairs C-Raf driven tumorigenesis
In order to assess the functional role of BAG-1 on tumorigenesis, we have generated a null allele of BAG-1. To inactivate the BAG-1 gene, exons 1 and 2 were replaced with a neomycin resistance gene. This strategy was chosen to disrupt the expression of all known isoforms of BAG-1 which are generated by alternate translation initiation of a single mRNA; the start codons are present in exons 1 and 2. Western blot analysis of liver protein extracts of BAG-1 deficient embryos showed the complete loss of all BAG-1 protein isoforms. Embryos homozygous for this allele died at midgestation at around E13,5, but the heterozygous animals (BAG-1+/-) are normal. A comprehensive description of the BAG-1-/- phenotype is subject of another manuscript.
Previously, we had generated a lung cancer mouse model by targeting constitutively active C-Raf kinase (SP-C C-Raf BxB) to the lung [14]. These mice develop multifocal adenomas early in adulthood. Based on the observation, that BAG-1 can activate C-Raf [2], we asked whether heterozygosity for BAG-1 would affect C-Raf BxB driven adenoma growth. We observed that lung tumour initiation was reduced by half in 1, 2 and 4 months old BAG-1+/- mice transgenic for SP-C C-Raf BxB compared to their BAG-1+/+ littermates. Tumour area was reduced by 75% in 4 month lungs of BAG-1 haploinsufficient mice compared to mice with two BAG-1 copies, see Figure 1. The histological picture emphasises the difference in adenoma formation between a representative SP-C C-Raf BxB/BAG-1+/+ and SP-C C-Raf BxB/BAG-1+/- lung. The difference in the staining intensity of the two lung sections derives mainly from the observation that the adenoma cells have a tendency to bind more intensively hematoxylin and eosin compared to normal lung cells. Thus, reduction of the BAG-1 gene dosage impairs the oncogenic activity of C-Raf in vivo.
Reduced BAG-1 expression in BAG-1 heterozygous lungs
Quantitative immunoblots demonstrated that the specific BAG-1 protein concentration in the lungs of BAG-1+/- mice was half the amount of BAG-1+/+ littermates, see Figure 2a. Moreover, immunohistochemical staining showed that BAG-1 was expressed in adenoma cells, see Figure 2b. There was no obvious difference in the BAG-1 immunohistochemistry of SP-C C-RafBxB/BAG-1+/+ and SP-C C-RafBxB/BAG-1+/- lungs.
Tumour cells of BAG-1 heterozygous mice show increased apoptosis
Concerning the molecular mechanism how a reduction of the BAG-1 protein expression in the heterozygous mice would impair tumorigenesis, we determined the fraction of apoptotic cells. Staining for activated caspase-3 revealed indistinguishable apoptosis in healthy regions of the lung of 1 month old SP-C C-Raf BxB mice with either one or two BAG-1 alleles, in line with the unaltered, normal lung structure of BAG-1+/- mice. In the adenomas, however, we observed a significant increase of apoptotic cells in BAG-1+/- SP-C C-Raf BxB mice compared with their SP-C C-Raf BxB/BAG-1+/+ littermates, see Figure 3a. This mechanism of action of BAG-1 on the regulation of cell survival is compatible with the phenotype of embryonic day 12,5 BAG-1 null embryos. Immunohistochemical staining for activated caspase-3 and trypan blue staining of dissociated cells showed hypocellularity and elevated levels of apoptosis in the livers of BAG-1-/- embryos (unpublished observations).
Proliferation and p-ERK signalling are unaffected in BAG-1 heterozygous mice
To exclude the alternative mechanism that the decreased level of BAG-1 expression in heterozygous animals would cause reduced cell proliferation in the adenomas, we performed proliferating cellular antigen (PCNA) staining. No significant differences were observed in the fraction of proliferating adenoma cells between SP-C C-Raf BxB animals heterozygous or wild type for BAG-1, see Figure 3b. Also, the percentages of adenoma cells positive for Ki-67, another proliferation marker and Bmi-1, a chromatin-associated protein expressed in stem cells, were not affected by the BAG-1 heterozygosity (not shown). Furthermore, staining of lung sections for phosphorylated ERK revealed no quantitative differences in the adenomas of SP-C C-Raf BxB animals heterozygous or wild type for BAG-1, see Figure 3c. Thus, signalling through the mitogenic cascade was not affected by the BAG-1 heterozygosity in the adenoma cells.
Conclusions
Tumours often are highly dependent on signalling pathways promoting cell growth or survival and may become hypersensitive to downregulation of key components within these signalling cascades. This study identifies BAG-1 as a protein specifically required at wild type expression levels for the survival of tumour cells and reveals it as potential anticancer target. Since many key components of survival pathways are regulated by interaction with (co-)chaperones [15], our finding is not without precedent but novel insofar as we have uncovered that reduced BAG-1 expression specifically targets tumour cells to apoptosis and impairs tumorigenesis. Whether this effect on adenoma cell survival requires that BAG-1 interacts with C-Raf or Hsc70/Hsp70 or with both partners requires additional studies. Questions concerning specific roles of the different BAG-1 isoforms were not addressed with this BAG-1 deficient mouse as both isoforms of BAG-1, p50 and p32 are absent in protein extracts of knock-out embryos. Another setting where BAG-1 has a physiological role is the heart, where up-regulation of BAG-1 after ischemia rescues cells from apoptosis [16].
A possible model combining the findings of this report and other data indicates that BAG-1 functions as an activator of C-Raf at the outer mitochondrial membrane where enzymatically activated C-Raf finds apoptosis-related targets such as BAD [17], see Figure 4. We can purify overexpressed C-Raf either in an enzymatically inactive form in a complex with Hsp70 or in an enzymatically active form in a complex with Hsp90/50 (unpublished observations), and BAG-1 is proposed to regulate this activation with ATP generated in the mitochondria. Experiments dealing with this questions are currently ongoing.
Therefore, the therapeutic efficacy of a standard chemotherapeutic agent [13] should be increased dramatically by co-application with a BAG-1 inhibitor, since it would target the adaptability of cancer cells to environmental stress and overcome their genetic plasticity. One way to reduce BAG-1 expression is through use of RNA interference-based gene silencing, in particular as BAG-1 overexpression has been observed in human tumours [11]. Drugs that bind to the ATP binding site of Hsc70/Hsp70 might also be expected to be effective as they would inhibit the interaction of BAG-1 with the ATPase domain of heat shock proteins. Such new specific BAG-1 inhibitors may be identified, aided by the known three-dimensional structure of the BAG domain [18,19].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RG caried out the molecular and histological studies and participated in the design and co-ordination of the study. BWK carried out the histological and immunohisto-chemical studies. GC participated in the histological and immunohistochemical experiments. URR participated in the design and co-ordination of the study. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank L. Fedorov, N. Gribanow, D. Heim, S. Hilz and T. Potapenko and Y. Yang for support. This work was supported by Deutsche Krebshilfe – Mildred Scheel Foundation (grants 10-1793-Ra7; 10-1935-Ra8) and by the DFG (grants TR17-TPB7; -TPZ2). G. Camarero was a postdoctoral fellow supported by Spanish Government (Ministerio de Educación y Cultura).
Figures and Tables
Figure 1 BAG-1 haplo-insufficiency delays C-Raf driven adenoma growth. (a) Adenoma initiation in SP-C C-Raf BxB mice (BAG-1+/+) and their Bag-1 haplo-insufficient littermates (BAG-1+/-) at 1, 2 and 4 months of age. Adenoma foci values represent mean ± s.e. from at least 4 mice of each genotype analyzed in a blinded fashion by two independent readers. (b) Adenoma area in the lungs of 4 months old mice. Each value represents mean ± s.e. from at least 4 mice of each genotype analyzed in a blinded fashion. (c-d) Examples of hematoxylin-eosin stained sections of lungs from SP-C C-Raf BxB transgenic mice wildtype for BAG-1 in comparison to a BAG-1 heterozygous littermate. Scale bar, 200 μm.
Figure 2 BAG-1 expression in the lung of SP-C C-Raf BxB transgenic mice (a) Lanes 1–8 show immunoblotting data for expression of BAG-1 in the lungs of 8 month old SP-C C-Raf BxB transgenic mice heterozygous (+/-) or homozygous (+/+) for BAG-1 as indicated below the lanes. Lane 9 shows absence of BAG-1 expression in a BAG-1 null (-/-) embryonic day 12,5 liver extract; lane 10 control liver. The markers along the left indicate relative molecular mass. The same blots were subsequently reacted with an antibody against GAPDH to demonstrate protein equal loading and are shown below. (b) BAG-1 immunostaining in SP-C C-Raf BxB transgenic mouse lung cancer tissue.
Figure 3 Increased apoptosis but no change in PCNA and p-ERK in tumor cells of SP-C C-RafBxB transgenic BAG-1 heterozygous mice (a-c) Quantification of immunohistochemical staining for apoptosis using an antibody that detects activated caspase-3 (a), PCNA (b) and phosphorylated ERK (p-ERK, c) of adenoma cells from 1-month-old SP-C C-Raf BxB transgenic mice of the indicated BAG-1 genotype. Each value represents mean ± s.e. from at least 4 mice of each genotype analyzed in three different experiments.
Figure 4 Model for cooperative action of BAG-1 and C-Raf in tumorigenesis A possible model combining the findings of this report and other data is shown. It indicates that BAG-1 functions as an activator of C-Raf at the outer mitochondrial membrane where enzymatically activated C-Raf finds apoptosis-related targets (for details see text).
==== Refs
Takayama S Sato T Krajewski S Kochel K Irie S Millan JA Reed JC Cloning and functional analysis of BAG-1: A novel Bcl-2- binding protein with anti-cell death activity Cell 1995 80 279 284 7834747 10.1016/0092-8674(95)90410-7
Wang HG Takayama S Rapp UR Reed JC Bcl-2 interacting protein, BAG-1, binds to and activates the kinase Raf-1 Proc Natl Acad Sci USA 1996 93 7063 7068 8692945 10.1073/pnas.93.14.7063
Townsend PA Cutress RI Sharp A Brimmell M Packham G BAG-1: a multifunctional regulator of cell growth and survival Biochim Biophys Acta 2003 1603 83 98 12618309
Mayer MP Brehmer D Gassler CS Bukau B Hsp70 chaperone machines Adv Protein Chem 2001 59 1 44 11868269
Weinstein IB Cancer. Addiction to oncogenes--the Achilles heal of cancer Science 2002 297 63 64 12098689 10.1126/science.1073096
Avruch J Khokhlatchev A Kyriakis JM Luo Z Tzivion G Vavvas D Zhang XF Ras activation of the Raf kinase: tyrosine kinase recruitment of the MAP kinase cascade Recent Prog Horm Res 2001 56 127 155 11237210 10.1210/rp.56.1.127
Wang HG Rapp UR Reed JC Bcl-2 targets the protein kinase Raf-1 to mitochondria Cell 1996 87 629 638 8929532 10.1016/S0092-8674(00)81383-5
Alavi A Hood JD Frausto R Stupack DG Cheresh DA Role of Raf in vascular protection from distinct apoptotic stimuli Science 2003 301 94 96 12843393 10.1126/science.1082015
Jr JT McCubrey JA Targeting the Raf kinase cascade in cancer therapy - novel molecular targets and therapeutic strategies Expert Opin Ther Targets 2002 6 659 678 12472379
Davies H Bignell GR Cox C Stephens P Edkins S Clegg S Teague J Woffendin H Garnett MJ Bottomley W Davis N Dicks E Ewing R Floyd Y Gray K Hall S Hawes R Hughes J Kosmidou V Menzies A Mould C Parker A Stevens C Watt S Hooper S Wilson R Jayatilake H Gusterson BA Cooper C Shipley J Hargrave D Pritchard-Jones K Maitland N Chenevix-Trench G Riggins GJ Bigner DD Palmieri G Cossu A Flanagan A Nicholson A Ho JW Leung SY Yuen ST Weber BL Seigler HF Darrow TL Paterson H Marais R Marshall CJ Wooster R Stratton MR Futreal PA Mutations of the BRAF gene in human cancer Nature (London) 2002 417 949 954 12068308 10.1038/nature00766
Cutress RI Townsend PA Brimmell M Bateman AC Hague A Packham G BAG-1 expression and function in human cancer Br J Cancer 2002 87 834 839 12373595 10.1038/sj.bjc.6600538
Tybulewicz VLJ Crawford CE Jackson PK Bronson RT Mulligan RC Neonatal lethality and lymphopenia in mice with a homozygous disruption of the c-abl proto-oncogene Cell 1991 65 1153 1163 2065352 10.1016/0092-8674(91)90011-M
Kramer BW Götz R Rapp UR Use of mitogenic cascade blockers for treatment of C-Raf induced lung adenoma in vivo: CI-1040 strongly reduces growth and improves lung structure BMC Cancer 2004 4 24 15171791 10.1186/1471-2407-4-24
Kerkhoff E Fedorov LM Siefken R Walter AO Papadopoulus T Rapp UR Lung-targeted expression of the c-Raf-1 kinase in transgenic mice exposes a novel oncogenic character of the wild-type protein Cell Growth Diff 2000 11 185 190 10775035
Takayama S Reed JC Homma S Heat-shock proteins as regulators of apoptosis Oncogene 2003 22 9041 9047 14663482 10.1038/sj.onc.1207114
Townsend PA Cutress RI Carroll CJ Lawrence KM Scarabelli TM Packham G Stephanou A Latchman DS BAG-1 proteins protect cardiac myocytes from simulated ischemia/reperfusion-induced apoptosis via an alternate mechanism of cell survival independent of the proteasome J Biol Chem 2004 279 20723 20728 14978028 10.1074/jbc.M400399200
Rapp UR Rennefahrt U Troppmair J Bcl-2 proteins: master switches at the intersection of death signaling and the survival control by Raf kinases Biochim Biophys Acta 2004 1644 149 158 14996499 10.1016/j.bbamcr.2003.10.015
Briknarova K Takayama S Brive L Havert ML Knee DA Velasco J Homma S Cabezas E Stuart J Hoyt DW Satterthwait AC Llinas M Reed JC Ely KR Structural analysis of BAG1 cochaperone and its interactions with Hsc70 heat shock protein Nat Struct Biol 2001 8 349 352 11276257 10.1038/86236
Sondermann H Scheufler C Schneider C Höhfeld J Hartl FU Moarefi I Structure of a Bag/Hsc70 complex: Convergent functional evolution of Hsp70 nucleotide exchange factors Science 2001 291 1553 1557 11222862 10.1126/science.1057268
| 15560850 | PMC539250 | CC BY | 2021-01-04 16:03:02 | no | BMC Cancer. 2004 Nov 24; 4:85 | utf-8 | BMC Cancer | 2,004 | 10.1186/1471-2407-4-85 | oa_comm |
==== Front
BMC PsychiatryBMC Psychiatry1471-244XBioMed Central London 1471-244X-4-401556656610.1186/1471-244X-4-40Research ArticleNegative and positive childhood experiences across developmental periods in psychiatric patients with different diagnoses – an explorative study Saleptsi Evangelia [email protected] Dana [email protected] Brigitte [email protected] Frank [email protected] Margarete [email protected] Karl [email protected] Klaus [email protected] Thomas [email protected] Department of Psychology, University of Konstanz, Fach D-25, 78457 Konstanz, Germany2 Department of Psychology, University of Jassy, Jassy, Romania3 Center for Psychiatry Reichenau, Konstanz, Germany4 Psychiatric Hospital Münsterlingen, Münsterlingen, Switzerland2004 26 11 2004 4 40 40 10 6 2004 26 11 2004 Copyright © 2004 Saleptsi et al; licensee BioMed Central Ltd.2004Saleptsi 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 high frequency of childhood abuse has often been reported in adult psychiatric patients. The present survey explores the relationship between psychiatric diagnoses and positive and negative life events during childhood and adulthood in psychiatric samples.
Methods
A total of 192 patients with diagnoses of alcohol-related disorders (n = 45), schizophrenic disorders (n = 52), affective disorders (n = 54), and personality disorders (n = 41) completed a 42-item self-rating scale (Traumatic Antecedents Questionnaire, TAQ). The TAQ assesses personal positive experiences (competence and safety) and negative experiences (neglect, separation, secrets, emotional, physical and sexual abuse, trauma witnessing, other traumas, and alcohol and drugs abuse) during four developmental periods, beginning from early childhood to adulthood. Patients were recruited from four Psychiatric hospitals in Germany, Switzerland, and Romania; 63 subjects without any history of mental illness served as controls.
Results
The amount of positive experiences did not differ significantly among groups, except for safety scores that were lower in patients with personality disorders as compared to the other groups. On the other side, negative experiences appeared more frequently in patients than in controls. Emotional neglect and abuse were reported in patients more frequently than physical and sexual abuse, with negative experiences encountered more often in late childhood and adolescence than in early childhood. The patients with alcohol-related and personality disorders reported more negative events than the ones with schizophrenic and affective disorders.
Conclusions
The present findings add evidence to the relationship between retrospectively reported childhood experiences and psychiatric diagnoses, and emphasize the fact that a) emotional neglect and abuse are the most prominent negative experiences, b) adolescence is a more 'sensitive' period for negative experiences as compared to early childhood, and c) a high amount of reported emotional and physical abuse occurs in patients with alcohol-related and personality disorders respectively.
==== Body
Background
It is difficult to assess the impact of childhood traumatic events on the psychiatric disorders in adulthood, as neither prospective research studies, nor experimental approaches are possible. Nevertheless, an increasing number of retrospective reports suggest that psychiatric disorders may be related to childhood psychological traumas such as neglect, physical or emotional abuse [1-6]. In particular, significant correlations between the severity of psychiatric symptoms and that of stressful and traumatic experiences during childhood were found [7-12]. Reports of physical and sexual abuse in childhood are more frequent in psychiatric patients than in the healthy population [13-16]; among these are patients diagnosed with affective disorders [17-19], somatization disorders [20-22], borderline personality disorders [3,7,23-25], substance-related disorders [26-28], and schizophrenic disorders [29-31]. Specifically, several studies have documented high rates of trauma in individuals with severe mental illness [32]. For a sample of schizophrenic women, Friedmann and Harisson (1984) reported that 60% of them had suffered childhood sexual abuse [33]. Abused patients displayed more pronounced symptoms such as hallucinations [34,35] and delusions [36].
Any conclusion to such reports, however, must be drawn by taking into consideration that the validity of childhood memories, particularly in psychiatric patients, may be questioned, as the range of childhood traumas indexed in these studies is generally limited, and often only childhood sexual abuse is targeted. Moreover, the observed relationships are correlational in nature, and do not justify the conclusion that childhood trauma favors the development of psychiatric disorders. Antecedents of developing psychopathology may also provoke certain parental behavior. Also, a third variable, such as social conditions, may have caused both childhood abuse and later pathological development. Another notable finding is that the prevalence rates of antecedent traumatic events vary considerably across studies. This may be due to different definitions of abuse which include more detailed [13] or more global [23] descriptions. Furthermore, the amount of psychosocial elements such as neglect, family disturbance, the nature of preexisting and subsequent attachment patterns, special competencies, etc., is difficult to be assessed or taken into account. Only a limited number of studies [37,38] have so far included control groups, allowing one to compare self-reports of abusive sexual experiences during childhood in psychiatric patients to those in the healthy population. There is also a lack of research studies that assess these issues within different cultural backgrounds.
The present study sets out to evaluate reported positive and negative life events from early childhood to adulthood in psychiatric patients. We addressed some of the above-mentioned problems by examining abuse histories across a range of several psychiatric diagnoses within a controlled cross-national design. We sought to examine whether (a) negative life experiences are positively associated with psychiatric diagnoses in adulthood, and (b) early childhood and adolescence were 'sensitive periods', that is, whether psychiatric diagnoses were more closely related to negative experience in these developmental periods.
The present study includes a German/Swiss and a Romanian psychiatric group, in order to determine whether reports vary between cultural backgrounds.
Methods
Subjects
Patients were recruited from four Psychiatric Hospitals within two different cultural settings, Switzerland/Southern Germany versus the Moldavia region in Romania: the Center for Psychiatry Reichenau and the Center for Psychiatry Weissenau in Germany, the Psychiatric Hospital Münsterlingen in Switzerland, and the Psychiatric Hospital "Socola", Jassy in Romania. A total of 192 psychiatric inpatients (98 German and 94 Romanian psychiatric patients, range 18–78 years) filled in the questionnaire. Sixty-three control subjects without any history of psychiatric diagnosis were recruited from the clinical staff and the university employees (Konstanz in Germany, Jassy in Romania) as controls (38 Germans and 25 Romanians). The control subjects have been simply inquired whether they had any stationary hospitalization in the psychiatry; no further assessments have been done. After a full explanation of the study, written informed consent was obtained from all subjects.
By considering the clinician-made diagnoses which were written down from the medical files available in the psychiatric clinics the patients were recruited from, the patients were distributed in four diagnostic groups: alcohol-related disorders (n = 45), schizophrenic disorders (n = 52), affective disorders (n = 54), and personality disorders (n = 41). At all psychiatric clinics in Germany/Switzerland and Romania the diagnoses were made according to the ICD-10 criteria. Within our patient groups, the following lifetime mental disorders were assessed by using the ICD-10: alcohol-related disorders (dependence syndrome, psychotic and unspecified mental disorders due to the use of alcohol), schizophrenic disorders (paranoid schizophrenia, schizoaffective disorder, and undifferentiated schizophrenia), affective disorders (bipolar depressive disorders, recurrent depressive disorder, cyclothymia, and dysthymia), and personality disorders (borderline, schizoid, paranoid, histrionic, dissocial, and dependent personality disorder respectively). A few patients within our sample were diagnosed with comorbid symptoms: 4 patients with affective disorders had symptoms of substance abuse and 7 of them had anxiety symptoms; also, within the schizophrenic disorders group, 2 patients had symptoms of alcohol abuse and 6 had depressive symptoms. There were also patients who received two diagnoses: one of which was a personality disorder (i.e., 5 patients with affective disorders, 7 with alcohol-related disorders, and 2 with schizophrenic disorders). In these cases, we considered the other diagnosis for the distribution into the diagnostic groups.
Table 1 summarizes the demographical characteristics of all subjects. The patient groups were similar with respect to the psychiatric history. There were differences among groups concerning gender distribution, age, and education. The gender-distribution differences among groups were due to the high number of women within the control and the affective disorders groups. With regard to the noted age differences, the patients with affective and alcohol-related disorders respectively had higher mean age as compared to all the other groups. The education differences among groups are only due to the lower educational level in patients with alcohol-related disorders. Romanian patients with affective disorders had a longer psychiatric history than the German/Swiss ones [t(51) = 2.3, p < 0.05]. Regarding gender distribution and the average duration of education, the German/Swiss and Romanian diagnostic groups were similar. The German/Swiss controls were significantly older than the Romanian ones [t(43) = 4.4, p < 0.001] and the German/Swiss patients with alcohol-related disorders were significantly younger than the Romanian ones [t(61) = 3.5, p < 0.001].
Table 1 Demographic characteristics of the control and of the patient groups1, 2
Alcohol Related Disorders Schizophrenic Disorders Affective Disorders Personality Disorders Controls
G/S R G/S R G/S R G/S R G/S R Analysis
N N N N N χ2 p
Gender 12 <.05
Female 6 10 10 11 14 19 11 6 23 15
Male 14 15 18 13 10 11 15 9 15 10
Mean ± SD Mean ± SD Mean ± SD Mean ± SD Mean ± SD F p
Age 31 ± 9 45 ± 12 34 ± 8 36 ± 10 40 ± 12 44 ± 8 34 ± 8 32 ± 11 38 ± 13 28 ± 7 6 <.001
Education 2 ± 1 2 ± 1 3 ± 1 3 ± 1 3 ± 1 3 ± 1 2 ± 1 3 ± 1 3 ± 1 3 ± 1 4 <.01
Psychiatric history (yrs) 6 ± 1 7 ± 10 8 ± 9 11 ± 10 4 ± 6 9 ± 9 4 ± 5 6 ± 7 - - 2 n.s.
1 Abbreviations: G/S: German/Swiss; R: Romanian; Education: 0 = no education, 1 = school for the mentally challenged, 3 = middle school, 4 = high school;
2 Areas where the values are written with bold characters indicate significant differences between German/Swiss and Romanian subjects within the diagnostic groups (i.e., alcohol-related disorders, schizophrenic disorders, affective disorders, personality disorders and control group respectively).
Material
Life experiences were assessed with the Traumatic Antecedents Questionnaire (TAQ) [40]. The TAQ is a 42-item self-rating questionnaire, which covers 11 subscales enquiring into the severity of positive (i.e., competence and safety) and negative experiences (i.e., neglect, separation, secrets, emotional abuse, physical abuse, sexual abuse, witnessing, other traumas, and alcohol and drugs) during four developmental periods (ages 0–6, 7–12 13–18, and ≥ 19). Each subscale includes 2–6 items. Each item requires the occurrence of a certain type of experience for each of the different age periods. The subjects were asked to score on a frequency/intensity scale the degree to which it describes their experience: 0 ("never or not at all"), 1 ("rarely or a little bit"), 2 ("occasionally or moderately"), 3 ("often or very much"), and DK ("don't know"). In a subsequent step, the average scores were calculated within each developmental period for each of the 11 subscales. The procedure we used was the following: first, the "don't know" responses were noted in a non-numerical manner, by using asterisks (*) to indicate missing values and these values were counted as 0; secondly, the response scores were added up and the sum was divided by the total number of items within the subscale in that age period for which there were numerical scores. By using this procedure, we excluded "don't know" responses from the total scores calculation.
Data analysis
Comparisons of demographic data were made with analysis of variance (ANOVA) and with two-tailed unpaired t-tests for continuous variables. Chi-square analysis was used to compare nominal data. The differences between groups were evaluated individually for each TAQ scale by repeated-measures ANOVA with the cultural background (German/Swiss versus Romanian), psychiatric status (alcohol-related disorders, schizophrenic disorders, affective disorders, personality disorders or controls), and gender (female versus male) as between-subjects factors, and developmental period (4 periods) as within-subjects factor. The probability level for rejecting the null hypothesis was set at P < 0.05. Post-hoc comparisons were carried out to evaluate main effects and interactions using Bonferroni/Dunn tests. A principal components analysis was also applied to the entire sample in order to identify those factors, which could account for individual variability across the eleven scales of the TAQ. The principal components were derived by using varimax rotation to orthogonalize solutions.
Results
Positive experiences
Table 2 lists group mean scores on each of the two positive experiences scales. The patients generally exhibited lower mean scores on reported positive experiences as compared to the controls. The reported level of competence did not differ between diagnostic groups [F(4,198) = 0.8, n.s] or cultural samples [F(1,198) = 0.1, n.s]. A main effect of developmental period [F(3,594) = 25.7, P < 0.001] was explained by the increase of competence from early childhood to adolescence (P < 0.05), and by the decrease of competence in adulthood as compared to adolescence (P < 0.05).
Table 2 Mean scores of positive experiences across developmental periods among all groups
Positive Experiences and Age at Onset Alcohol Related Disorders Schizophrenic Disorders Affective Disorders Personality Disorders Control Group Analysis
Mean ± SD Mean ± SD Mean ± SD Mean ± SD Mean ± SD F p
Early Childhood (0–6)
Competence 1.5 ± 0.9 1.9 ± 0.9 1.7 ± 1.0 1.7 ± 0.9 1.7 ± 0.1 0.7 n.s
Safety 1.4 ± 0.8 1.7 ± 0.8 1.6 ± 0.8 1.2 ± 0.8 1.6 ± 0.7 3.2 <.05
Latency (7–12)
Competence 2.1 ± 0.8 2.0 ± 0.8 2.1 ± 0.7 1.9 ± 0.8 2.1 ± 0.9 0.2 n.s
Safety 1.8 ± 0.8 1.8 ± 0.8 1.9 ± 0.7 1.3 ± 0.7 1.9 ± 0.7 5.1 <.001
Adolescence (13–18)
Competence 2.1 ± 0.9 2.1 ± 0.7 2.2 ± 0.6 2.1 ± 0.8 2.3 ± 0.8 0.6 n.s
Safety 1.9 ± 0.8 1.7 ± 0.8 1.8 ± 0.8 1.3 ± 0.8 2.0 ± 0.8 4.0 <.01
Adulthood (≥19)
Competence 1.9 ± 1.0 1.8 ± 0.8 2.1 ± 0.8 2.0 ± 0.8 2.2 ± 0.8 1.3 n.s
Safety 1.7 ± 0.8 1.7 ± 0.8 1.7 ± 0.8 1.5 ± 0.7 2.1 ± 0.7 4.9 <.001
For both cultural samples, patients with personality disorders reported lower values on the safety subscale than any of the other groups [F(4,215) = 4.5, P < 0.01]. Post hoc tests showed that patients with personality disorders (P < 0.001) and those with affective disorders (P < 0.01) reported less such experiences as compared to the controls. The reported level of safety increased through adolescence [F(3,645) = 11.1, P < 0.001], but the interaction of the developmental period with the psychiatric status revealed a decrease in safety accounts from the age of 13–18 years towards adulthood in all patient groups [F(12,645) = 2.8, P < 0.001].
Negative experiences
Table 3 shows the mean scores of traumatic experiences for all patient groups and for the control group across developmental periods. Negative experiences were more frequent in patients than in controls as indicated by significant main effects of the psychiatric status on each of the nine subscales. In addition, there was an important increase of amount of reported negative experiences across developmental periods.
Table 3 Mean scores of negative experiences among all groups
Negative Experiences and Age at Onset Alcohol Related Disorders Schizophrenic Disorders Affective Disorders Personality Disorders Control Group Analysis
Mean ± SD Mean ± SD Mean ± SD Mean ± SD Mean ± SD F p
Early Childhood (0–6)
Neglect 0.6 ± 0.5 0.7 ± 0.7 0.5 ± 0.5 0.8 ± 0.6 0.2 ± 0.3 8.1 <.001
Separation 0.5 ± 0.6 0.4 ± 0.5 0.4 ± 0.6 0.6 ± 0.8 0.2 ± 0.4 3.9 <.01
Secrets 0.9 ± 0.9 1.2 ± 1.1 1.0 ± 0.9 1.4 ± 1.1 0.6 ± 0.7 4.8 <.01
Emotional Abuse 0.7 ± 0.6 1.0 ± 1.0 0.7 ± 0.8 1.2 ± 0.9 0.3 ± 0.5 8.2 <.001
Physical Abuse 0.4 ± 0.6 0.4 ± 0.5 0.3 ± 0.5 0.7 ± 0.7 0.1 ± 0.3 5.7 <.001
Sexual Abuse 0.0 ± 0.3 0.2 ± 0.6 0.1 ± 0.3 0.3 ± 0.6 0.1 ± 0.4 2.0 n.s.
Trauma Witnessing 0.4 ± 0.5 0.4 ± 0.5 0.5 ± 0.6 0.7 ± 0.9 0.1 ± 0.2 7.6 <.001
Other Traumas 0.4 ± 0.5 0.3 ± 0.4 0.2 ± 0.4 0.4 ± 0.6 0.2 ± 0.3 3.1 <.05
Alcohol/Drug Abuse 0.4 ± 0.8 0.3 ± 0.6 0.5 ± 0.7 0.5 ± 0.7 0.1 ± 0.2 4.0 <.01
Latency (7–12)
Neglect 0.7 ± 0.6 0.8 ± 0.7 0.7 ± 0.5 1.1 ± 0.8 0.5 ± 0.5 5.0 <.001
Separation 0.6 ± 0.7 0.5 ± 0.6 0.6 ± 0.7 0.9 ± 0.8 0.5 ± 0.6 2.8 <.05
Secrets 1.0 ± 0.9 1.2 ± 0.1 1.1 ± 0.9 1.6 ± 1.1 0.8 ± 0.8 4.2 <.01
Emotional Abuse 0.8 ± 0.6 1.1 ± 0.9 1.0 ± 0.8 1.4 ± 0.9 0.7 ± 0.8 4.6 <.01
Physical Abuse 0.5 ± 0.7 0.6 ± 0.7 0.6 ± 0.6 0.9 ± 0.8 0.5 ± 0.6 2.5 <.05
Sexual Abuse 0.1 ± 0.3 0.2 ± 0.5 0.1 ± 0.2 0.5 ± 0.8 0.1 ± 0.4 5.3 <.001
Trauma Witnessing 0.5 ± 0.6 0.5 ± 0.5 0.7 ± 0.6 0.9 ± 0.8 0.4 ± 0.5 5.5 <.001
Other Traumas 0.4 ± 0.5 0.4 ± 0.4 0.4 ± 0.5 0.5 ± 0.6 0.3 ± 0.4 1.4 n.s.
Alcohol/Drug Abuse 0.5 ± 0.7 0.3 ± 0.6 0.6 ± 0.7 0.6 ± 0.8 0.2 ± 0.4 4.1 <.01
Adolescence (13–18)
Neglect 1.0 ± 0.6 0.9 ± 0.7 1.0 ± 0.5 1.2 ± 0.7 0.8 ± 0.6 2.5 <.05
Separation 0.9 ± 0.8 0.8 ± 0.8 0.9 ± 0.7 0.9 ± 0.7 0.6 ± 0.7 2.4 <.05
Secrets 1.1 ± 0.9 1.3 ± 1.0 1.1 ± 0.9 1.5 ± 1.0 0.7 ± 0.8 4.7 <.01
Emotional Abuse 0.8 ± 0.6 1.3 ± 0.9 1.1 ± 0.7 1.4 ± 0.9 0.8 ± 0.8 4.6 <.01
Physical Abuse 0.8 ± 0.7 0.5 ± 0.6 0.5 ± 0.6 1.0 ± 0.9 0.5 ± 0.6 5.0 <.001
Sexual Abuse 0.2 ± 0.4 0.2 ± 0.4 0.2 ± 0.4 0.5 ± 0.8 0.1 ± 0.3 3.2 <.05
Trauma Witnessing 0.6 ± 0.6 0.5 ± 0.5 0.7 ± 0.6 1.0 ± 0.8 0.4 ± 0.5 5.9 <.001
Other Traumas 0.6 ± 0.6 0.5 ± 0.5 0.4 ± 0.5 0.6 ± 0.5 0.3 ± 0.3 3.4 <.05
Alcohol/Drug Abuse 1.1 ± 1.0 0.6 ± 0.8 0.7 ± 0.9 0.9 ± 0.8 0.4 ± 0.7 5.3 <.001
Adulthood (19≥)
Neglect 1.4 ± 0.8 1.3 ± 0.9 1.2 ± 0.7 1.2 ± 0.8 0.9 ± 0.7 3.0 <.01
Separation 1.6 ± 0.7 1.1 ± 0.9 1.5 ± 0.8 1.2 ± 0.9 1.0 ± 0.7 4.3 <.01
Secrets 1.1 ± 1.0 1.3 ± 1.1 1.2 ± 0.9 1.6 ± 0.9 0.5 ± 0.7 8.1 <.001
Emotional Abuse 0.9 ± 0.7 1.2 ± 0.9 1.1 ± 0.9 1.4 ± 0.9 0.6 ± 0.6 6.8 <.001
Physical Abuse 1.0 ± 0.9 1.0 ± 0.8 0.9 ± 0.9 1.0 ± 0.9 0.4 ± 0.6 5.1 <.001
Sexual Abuse 0.3 ± 0.6 0.6 ± 0.8 0.4 ± 0.7 0.5 ± 0.7 0.2 ± 0.4 2.8 <.05
Trauma Witnessing 0.8 ± 0.6 0.7 ± 0.7 1.0 ± 0.8 1.0 ± 0.9 0.5 ± 0.4 4.3 <.01
Other Traumas 1.2 ± 0.6 1.0 ± 0.7 1.1 ± 0.8 1.0 ± 0.6 0.4 ± 0.4 14.8 <.001
Alcohol/Drug Abuse 2.1 ± 0.8 0.8 ± 0.9 1.1 ± 0.9 1.0 ± 0.9 0.3 ± 0.6 29.3 <.001
With respect to the experiences of neglect, the psychiatric patients reported higher rates than the controls [F(4,214) = 5.7, P < 0.001], the post hoc tests revealing that all patient groups reported more such experiences as compared to the controls: patients with personality disorders (P < 0.001), alcohol-related disorders (P < 0.05), schizophrenic disorders (P < 0.05), and affective disorders (P < 0.05). There was an increase of the amount of reported neglect experiences across developmental periods [F(3,642) = 91.5, P < 0.001]. Across developmental periods there were significant effects of the psychiatric status [F(12,642) = 3.2, P < 0.001]: the post hoc tests showed that patients with personality disorders (P < 0.001) and with alcohol-related disorders (P < 0.01) reported a highly significant increase of the amount of neglect experiences across developmental periods as compared to the controls (Figure 1).
Figure 1 Mean neglect score across developmental periods among all groups. The psychiatric patients reported higher rates than the controls [F(4,214) = 5.7, P < 0.001]. There was an increase of the amount of reported neglect experiences across developmental periods [F(3,642) = 91.5, P < 0.001]. Error bars stand for standard error of the mean. Asterisks indicate significant main effects of the psychiatric status.
Irrespective of the psychiatric status and developmental periods, Romanian subjects generally reported a higher amount of neglect experiences, as shown by the main effect of the cultural background [F(1,214) = 6.4, P < 0.05]. Romanian patients, particularly those with schizophrenic disorders, reported a higher incidence of neglect experiences than their German counterparts (P < 0.01), as revealed by the interaction between the psychiatric status and cultural background [F(4,214) = 5.6, P < 0.001]. As indicated by the interaction between the developmental period and the cultural background, the mean scores of neglect experiences were higher in the Romanian sample as compared to the German/Swiss one for the earliest (0–6 years) period [F(3,642) = 5.3, P < 0.001].
Separation
Patients, particularly those with alcohol-related disorders, personality disorders, and affective disorders reported more often separation experiences than controls [F(4,227) = 3.3, P < 0.01, P < 0.01 for post-hocs]. Mean scores on separation increased with age, and were highest in adulthood [F(3,681) = 103.0, P < 0.001].
Secrets
Higher patient mean scores were confirmed by the main effect of the psychiatric status [F(4,182) = 6.8, P < 0.001], especially for those with personality disorders (P < 0.001) and with schizophrenic disorders (P < 0.001), as revealed by post-hoc tests. There was also an indication of sensitivity to the cultural background [F(1,182) = 7.5, P < 0.01], as there was an increase in these scores in the Romanian sample, irrespective of the diagnosis and developmental period.
Emotional abuse was more frequently reported by patients than by controls [F(4,194) = 7.0, P < 0.001] and more frequently by patients with personality disorders (P < 0.01) and by schizophrenic patients (P < 0.05) than the ones with a history of alcohol-related disorders (Figure 2). A main effect of the developmental period [F(3,582) = 24.0, P < 0.001] was explained by an increase of the reported emotional abuse from early childhood to adolescence (P <0.05), and a decrease in adulthood (P < 0.05) were noted. Similar to the case of the neglect experiences, the Romanian sample scored also higher than the German/Swiss sample, mainly for the earliest (0–6 yr.) period, as revealed by the interaction between the development period and the cultural background [F(3,582) = 5.4, P < 0.01].
Figure 2 Mean emotional abuse score across developmental periods among all groups. Emotional abuse was more frequently reported by patients than by controls [F(4,194) = 7.0, p < 0.001]. A main effect of the developmental period [F(3,582) = 24.0, P < 0.001] was explained by an increase of the reported emotional abuse from early childhood to adolescence, and a decrease in adulthood were noted. Error bars stand for standard error of the mean. Asterisks indicate significant main effects of the psychiatric status.
Irrespective of the developmental period, physical abuse was more often reported by patients with personality disorders [F(4,202) = 5.7, P < 0.001] (Figure 3). Post-hoc comparisons also revealed higher rates of physical abuse reports among patients with alcohol-related disorders (P < 0.01) and with schizophrenic disorders (P < 0.05) than among controls. The reports of physical abuse generally increased across developmental periods, with adulthood as the most susceptible period of such reports [F(3, 606) = 35.1, P < 0.001]. The interaction of the developmental period with the psychiatric status showed that this increase in physical abuse reports across developmental periods was mainly to be remarked in patients [F(12,606) = 3.0, P < 0.001].
Figure 3 Mean physical abuse score across developmental periods among all groups. Physical abuse was more often reported by patients with personality disorders [F(4,202) = 5.7, P < 0.001]. The reports of physical abuse generally increased across developmental periods, with adulthood as the most susceptible period of such reports [F(3, 606) = 35.1, P < 0.001]. Error bars stand for standard error of the mean. Asterisks indicate significant main effects of the psychiatric status.
Sexual abuse (Figure 4) was primarily reported by patients, and not by controls [F(4,205) = 5.2, P < 0.001], and particularly by patients with personality disorders (P < 0.001). Higher rates of sexual abuse were reported among patients with alcohol-related disorders (P < 0.01), with schizophrenic disorders (P < 0.05), and with affective disorders (P < 0.05) than among controls as shown by post-hoc tests. If sexual abuse was experienced, it occurred particularly in later developmental periods [F(3,615) = 20.4, P < 0.001]. Sexual abuse was more often experienced by female patients [F(1,205) = 10.0, P < 0.001] after puberty [F(3, 615) = 10.0, P < 0.001], as revealed by the interaction between the developmental period and gender. We also found a 3-way interaction between the developmental period, psychiatric status and cultural background [F(12,615) = 2.6, P < 0.01). The interaction between the developmental period and cultural background revealed that Romanian but not German/Swiss schizophrenics reported more frequently sexual abuse particularly in adulthood [F(3,615) = 5.0, P < 0.01].
Figure 4 Mean sexual abuse score across developmental periods among all groups. Sexual abuse was primarily reported by patients, and not by controls, and particularly by patients with personality disorders [F(4,205) = 5.2, P < 0.001]. If sexual abuse was experienced, it occurred particularly in later developmental periods [F(3,615) = 20.4, P < 0.001]. Error bars stand for standard error of the mean. Asterisks indicate significant main effects of the psychiatric status.
Trauma witnessing was reported most often by patients with personality disorders as compared to all other groups [F(4,209) = 8.0, P < 0.001]. Post-hoc tests showed that patients with affective disorders (P < 0.01) and with alcohol-related disorders (P < 0.05) also reported more experiences of trauma witnessing than the controls. Irrespective of the diagnosis, Romanian patients, but not controls, reported higher mean scores on this variable and more often than their German/Swiss counterparts [F(1,209) = 17.0, P < 0.001]. The interaction between the cultural background and developmental period indicated in the Romanian sample an increase of trauma witnessing in adulthood [F(3,627) = 8.0, P < 0.001].
Other traumas
Similar to the pattern of trauma witnessing, all patients reported a greater number of traumatic events than the control group [F(4,211) = 8.0, P < 0.001]: alcohol-related disorders (P < 0.001), personality disorders (P < 0.001), schizophrenic disorders (P < 0.01), and affective disorders (P < 0.01), as explained by post-hocs. An increase in the amount of other traumas reports across the developmental periods with highest values in adulthood for all patient groups [F(12, 633) = 7.0, P < 0.001] was also revealed.
Alcohol and drug abuse
As previously expected, patients treated for alcohol-related disorders reported more alcohol and drug abuse than all the other groups [F(4,213) = 12.3, P < 0.001]. The post-hoc tests showed that the other patient groups also reported more alcohol and drug abuse when compared to the control group: affective disorders (P < 0.001), personality disorders, (P < 0.001) and schizophrenic disorders (P < 0.05). As also anticipated, abuse increased across developmental periods until adulthood [F(3,639) = 110.0, P < 0.001], particularly in patients with alcohol-related disorders, as revealed by the interaction between the developmental period and psychiatric status [F(12,639) = 14.3, P < 0.001]. Irrespective of the diagnosis, Romanian patients, but not controls, showed higher mean scores on reporting alcohol and drug abuse than their German/Swiss counterparts [F(4,213) = 3.4, P < 0.01], as shown by the interaction between the psychiatric status and cultural background. Compared to the German group, alcohol and drug abuse in the Romanian sample was higher, particularly in adulthood [F(3,639) = 9.8, P < 0.001].
Interrelations
A principal components factor analysis was performed to explore interrelationships among TAQ subscales. The results of this analysis indicated that the most appropriate solution involved five factors that jointly accounted for 56.2% of the total variance in the dataset. Table 4 summarizes the results of the varimax rotation for the five-factor solution. The first factor showed high positive loadings on physical abuse, sexual abuse, trauma witnessing, and other traumas, obviously explains the traumatic experiences. The second factor showed high positive loadings on competence and safety, apparently accounting for variance attributed to positive experiences. The third factor showed high positive loadings on the first year of illness with alcohol and drug abuse. The fourth factor, consisting of separation, evidently explains disruptions of attachment. The fifth factor, which included secrets and emotional abuse, appeared to account for family chaos. Thus, the structure of the study instrument was well reproduced for the present sample, which included different psychiatric diagnoses and different cultural backgrounds.
Table 4 Varimax solution with five factors for negative and positive childhood experiences across developmental periods in psychiatric patients with different diagnoses1
Factor Loading2
Variables FACTOR 1: Traumatic Experiences3 FACTOR 2: Positive Experiences4 FACTOR 3: Vulnerability to Alcohol Abuse5 FACTOR 4: Disruptions of Attachment6 FACTOR 5: Family Chaos7
Competence -0.0 0.8 0.2 0.1 0.1
Safety 0.2 0.8 -0.0 0.0 -0.2
Neglect 0.2 -0.3 0.1 -0.0 0.3
Separation 0.4 0.2 -0.0 0.8 0.1
Secrets -0.0 -0.1 -0.0 0.1 0.7
Emotional Abuse 0.2 0.0 0.6 -0.0 0.6
Physical Abuse 0.7 -0.0 -0.0 0.0 0.0
Sexual Abuse 0.4 0.0 0.0 -0.6 0.1
Witnessing 0.6 0.0 -0.0 0.0 0.2
Other Traumas 0.6 0.2 0.1 0.2 0.1
Alcohol & Drug Abuse 0.4 -0.2 0.5 0.0 -0.3
First Year of Illness -0.1 0.1 0.8 -0.1 0.2
1 Total percent of variance = 56.2%
2 Shaded areas indicate specific domains of the TAQ contributing to each factor
3 Eigenvalue = 4.88; percent of variance = 40.7%
4 Eigenvalue = 1.476; percent of variance = 12.3%
5 Eigenvalue = 1.013; percent of variance = 8.4%
6 Eigenvalue = 0.835; percent of variance = 7.0%
7 Eigenvalue = 0.815; percent of variance = 6.8%
Discussion
The study aimed at exploring whether psychiatric diagnoses, e.g. alcohol-related disorders, schizophrenic disorders, affective disorders, and personality disorders are related to retrospectively reported positive and negative life events across developmental periods, and if so, whether special developmental periods are characterized by more negative experiences than others.
Our findings demonstrate a strong association between reports of traumatic events and certain psychiatric disorders. In other studies, negative experiences were reported by individuals with diagnoses such as affective disorders [18,41] and schizophrenic disorders [42,43], but these experiences were less common and cumulatively less severe. Negative experiences were particularly prominent in patients with personality disorders [24,25,44] and in patients with substance-related disorders [26,45,46]. Negative experiences were reported more often in late childhood and adolescence than in early childhood and adulthood. Previous studies indicated that the earlier onset of abuse was associated with greater severity and longer duration of mental problems [2,10,45,47].
If the present findings are consistent with some prior studies [5,9,16] in that they indicate a relationship between physical and sexual abuse and psychiatric disorders, they do not support the view expressed by Van der Kolk et al. about early abuse at an early stage of development [48]. The current investigation showed that many psychiatric patients had terrible histories of childhood physical and/or sexual abuse. This finding was marginally significant for the childhood sexual abuse histories and must therefore be interpreted with caution. However, one should keep in mind that self-report questionnaires depend heavily upon conscious retrieval capacity for autobiographic events. It is conceivable that in the current group of patients, early abuse events were less remembered as compared to abuse events experienced later in childhood. An advantage of the TAQ used in the present study is the assessment of negative experiences during both childhood and adulthood, while most of the other studies have so far focused primarily on the impact of childhood abuse, except Cascardi et al. [49] and Goodman et al. [32]. Another advantage of the TAQ is that it addresses the issue of neglect [50]. Given the sample of patients with different psychiatric diagnoses, this replicates Van der Kolk's et al. notion that patients who experience neglect early in their lives develop serious problems with affect regulation [51]. The present data add to the evidence, suggesting that neglect, emotional and physical abuse are experienced by many psychiatric patients [52,53]. This implies that although childhood traumas may contribute to a mental disorder in adulthood, the lack of secure attachments maintains it. Although emotional neglect has received less attention, perceived emotional rejection by parents has been associated with alcohol abuse [54] and delinquency [55] during adolescence and adulthood. Early emotional injuries could possibly trigger vulnerability to noxious experiences.
Furthermore, experiences of parental loss or separation were prominent in adulthood especially for the patients with alcohol-related disorders and with affective disorders. The high incidence of such negative experiences during this period in the patients with alcohol-related disorders could be, at the same time, a direct consequence of the behavioral deviance of these individuals and contribute to the maintenance of alcohol abuse.
Limitations of the study
The present data has to be considered in the light of several possible limitations. First, the information obtained by self-report and without external evidence could be less reliable and valid, especially if we take into account the sensitive nature of this research. Herman and Schatzow, however, provide empirical support for the validity of abused patients' self-reports as well [56]. They found that when corroborating evidence is sought, the majority of women are able to obtain confirmation of abuse. No independent corroborating evidence was sought for any self-reported case of childhood negative experiences. Therefore, the validity of abuse reports cannot be assured. Recall may be biased, but there is no evidence that psychiatric patients are more likely to lie about or imagine child abuse [57,58]. There is some evidence, however, that "patients are biased to underreport abuse histories rather than to over report them" [59]. There were some "don't know" subject answers regarding abuse/neglect experiences, most of them in the early childhood. Most probably, the patients had difficulties recalling experiences that occurred at a very young age rather than trying to evade giving a positive answer. Furthermore, another methodological limitation in this study is that measuring neglect/emotional abuse in early childhood is particularly difficult as the awareness of it necessitates the development of a degree of differentiation and autonomy, which is seldom the case with psychiatric patients.
Both individual interviews and self-report questionnaire methods present higher figures than chart reviews do, indicating that patients usually do not spontaneously offer such information to their therapists. When offered, the information is not reliably documented [57]. However, the data from our ongoing study in patients with personality disorders suggest that reports on events in general and physical abuse events in particular are highly stable across two measurement periods of time separated by 24 months. We also note that our sample consisted of psychiatric inpatients, and thus may not be representative of the broader population of patients with these disorders. The clinical validity of the TAQ has also been criticized [60]. The questionnaire is meant to be an applied clinically oriented measure, which has not yet been proved to be a psychometrically sound research instrument. This issue should be addressed in future studies using both convergent and divergent instruments.
Romanian patients diagnosed with schizophrenic disorders differed significantly, with respect to the number of negative events, as compared to their German counterparts. One factor accounting for this difference might be the stressful environment during the former Ceausescu regime in Romania. During this 25-year period violations of human rights, terror, and corruption prevailed [61,62]. This result may also be due to the different diagnostic procedures used by Romanian and German/Swiss clinicians. Reports of higher rates of psychotic-like or specifically schizophrenic symptoms do not necessarily imply a diagnosis of schizophrenia. Once abuse is identified, a change of diagnosis, from schizophrenia to PTSD, is often made, with significant advantages for the individuals [30].
Conclusions
The present study demonstrates an association between negative life events in childhood and psychiatric diagnoses in adult life, which is in line with a number of other studies [6,63]. Unlike previous reports [3,64], we found that psychiatric patients were more likely to report more negative life events during late childhood and adolescence rather than during early childhood and adulthood. These conclusions corroborate with one of the central hypothesis of life-span psychotraumatology, that is, adolescence is an extremely critical phase in the development of later psychopathology [65,66]. However, in line with findings offered by earlier controlled studies [37,38], psychiatric patients were more likely to report higher rates of negative life events during childhood than controls did.
Although one cannot assume a direct causal relationship between childhood abuse and adult psychopathology from the present data, the present study provides further preliminary and explorative evidence for the high load of negative life events in psychiatric patients. An advantage of this study is the examination of the abuse histories across a range of four psychiatric diagnoses within a controlled comparison design. Our findings are important and clinically highly relevant for further etiological research of causal and maintenance factors of psychiatric symptomatic, as well as for the research on the treatment of these conditions.
The special value of the study lies in its cross-national comparison from a clinical psychological point of view including a highly underresearched country like Romania. More attention should be paid to the sad situation of the patients in Romania who are often under inadequate pharmacological and insufficient psychotherapeutic treatment, as well as under inappropriate hospitalization conditions. Further research should concentrate on the epidemiology and developmental psychopathology of psychiatric populations in other countries than the usually researched ones. Generally, reports of traumatic experiences during the whole lifespan should be more carefully considered in the clinical diagnosis process and in the development of treatment programs for the psychiatric patients.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ES carried out the study in Germany, performed the statistical analysis and drafted the manuscript. DB carried out the study in Romania and drafted the manuscript. BR conceived of the study and drafted the manuscript. FN participated in the design of the study. MS participated in the design of the study. KS participated in the coordination of the study. KH participated in the coordination of the study. TE conceived of the study and drafted the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Our research was supported by the cooperation Kanton Thurgau-University of Konstanz. We thank Dr. Sabine Heim and Dr. Victor Candia for their helpful comments during the preparation of the manuscript.
==== Refs
Beck JC van der Kolk B Reports of childhood incest and current behavior of chronically hospitalized psychotic women Am J Psychiatry 1987 144 1474 6 3674230
Briere J Zaidi LY Sexual abuse histories and sequelae in female psychiatric emergency room patients Am J Psychiatry 1989 146 1602 6 2589554
Herman JL Perry JC van der Kolk BA Childhood trauma in borderline personality disorder Am J Psychiatry 1989 146 490 5 2929750
Chu JA Dill DL Dissociative symptoms in relation to childhood physical and sexual abuse Am J Psychiatry 1990 147 887 92 2104510
Mullen PE Martin JL Anderson JC Romans SE Herbison GP Childhood sexual abuse and mental health in adult life Br J Psychiatry 1993 163 721 32 8306113
Ellason JW Ross CA Childhood trauma and psychiatric symptoms Psychol Rep 1997 80 447 50 9129365
Bryer JB Nelson BA Miller JB Krol PA Childhood sexual and physical abuse as factors in adult psychiatric illness Am J Psychiatry 1987 144 1426 30 3674223
Powell RA Childhood abuse and psychosis Br J Psychiatry 1994 165 838 9 7881794
Wurr CJ Partridge IM The prevalence of a history of childhood sexual abuse in an acute adult inpatient population Child Abuse Negl 1996 20 867 72 8886470 10.1016/0145-2134(96)00074-9
Read J Child abuse and severity of disturbance among adult psychiatric inpatients Child Abuse Negl 1998 22 359 68 9631248 10.1016/S0145-2134(98)00009-X
Lange A de Beurs E Dolan C Lachnit T Sjollema S Hanewald G Long-term effects of childhood sexual abuse: objective and subjective characteristics of the abuse and psychopathology in later life J Nerv Ment Dis 1999 187 150 8 10086471 10.1097/00005053-199903000-00004
Grilo CM Sanislow C Fehon DC Martino S McGlashan TH Psychological and behavioral functioning in adolescent psychiatric inpatients who report histories of childhood abuse Am J Psychiatry 1999 156 538 43 10200731
Carmen EH Rieker PP Mills T Victims of violence and psychiatric illness Am J Psychiatry 1984 141 378 83 6703102
Jacobson A Richardson B Assault experiences of 100 psychiatric inpatients: evidence of the need for routine inquiry Am J Psychiatry 1987 144 908 13 3605402
Swett C JrSurrey J Cohen C Sexual and physical abuse histories and psychiatric symptoms among male psychiatric outpatients Am J Psychiatry 1990 147 632 6 2327492
Palmer RL Effects of childhood sexual abuse on adult mental health Br J Hosp Med 1992 48 9 10 1504693
Levitan RD Parikh SV Lesage AD Hegadoren KM Adams M Kennedy SH Goering PN Major depression in individuals with a history of childhood physical or sexual abuse: relationship to neurovegetative features, mania, and gender Am J Psychiatry 1998 155 1746 52 9842786
Gladstone G Parker G Wilhelm K Mitchell P Austin MP Characteristics of depressed patients who report childhood sexual abuse Am J Psychiatry 1999 156 431 7 10080560
Hyun M Friedman SD Dunner DL Relationship of childhood physical and sexual abuse to adult bipolar disorder Bipolar Disord 2000 2 131 5 11252652 10.1034/j.1399-5618.2000.020206.x
Morrison J Childhood sexual histories of women with somatization disorder Am J Psychiatry 1989 146 239 41 2912266
Pribor EF Yutzy SH Dean JT Wetzel RD Briquet's syndrome, dissociation, and abuse Am J Psychiatry 1993 150 1507 11 8379555
Saxe GN Chinman G Berkowitz R Hall K Lieberg G Schwartz J van der Kolk BA Somatization in patients with dissociative disorders Am J Psychiatry 1994 151 1329 34 8067489
Craine LS Henson CE Colliver JA MacLean DG Prevalence of a history of sexual abuse among female psychiatric patients in a state hospital system Hosp Community Psychiatry 1988 39 300 4 3356438
Shearer SL Peters CP Quaytman MS Ogden RL Frequency and correlates of childhood sexual and physical abuse histories in adult female borderline inpatients Am J Psychiatry 1990 147 214 6 2301663
Zanarini MC Yong L Frankenburg FR Hennen J Reich DB Marino MF Vujanovic AA Severity of reported childhood sexual abuse and its relationship to severity of borderline psychopathology and psychosocial impairment among borderline inpatients J Nerv Ment Dis 2002 190 381 7 12080208 10.1097/00005053-200206000-00006
Windle M Windle RC Scheidt DM Miller GB Physical and sexual abuse and associated mental disorders among alcoholic inpatients Am J Psychiatry 1995 152 1322 8 7653688
Moncrieff J Drummond DC Candy B Checinski K Farmer R Sexual abuse in people with alcohol problems. A study of the prevalence of sexual abuse and its relationship to drinking behaviour Br J Psychiatry 1996 169 355 60 8879723
Langeland W Hartgers C Child sexual and physical abuse and alcoholism: a review J Stud Alcohol 1998 59 336 48 9598715
Norman RM Malla AK Stressful life events and schizophrenia. I: A review of the research Br J Psychiatry 1993 162 161 6 8435685
Read J Child abuse and psychosis: A literature review and implications for professional practice Professional Psychology: Research and Practice 1997 28 448 456 10.1037//0735-7028.28.5.448
Resnick SG Bond GR Mueser KT Trauma and posttraumatic stress disorder in people with schizophrenia J Abnorm Psychol 2003 112 415 23 12943020 10.1037/0021-843X.112.3.415
Goodman LA Rosenberg SD Mueser KT Drake RE Physical and sexual assault history in women with serious mental illness: prevalence, correlates, treatment, and future research directions Schizophr Bull 1997 23 685 96 9366004
Friedman S Harrison G Sexual histories, attitudes, and behavior of schizophrenic and "normal" women Arch Sex Behav 1984 13 555 67 6517688
Heins T Gray A Tennant M Persisting hallucinations following childhood sexual abuse Aust N Z J Psychiatry 1990 24 561 5 2073233
Read J Argyle N Hallucinations, delusions, and thought disorder among adult psychiatric inpatients with a history of child abuse Psychiatr Serv 1999 50 1467 72 10543857
Goff DC Brotman AW Kindlon D Waites M Amico E The delusion of possession in chronically psychotic patients J Nerv Ment Dis 1991 179 567 71 1919560
Palmer RL Coleman L Chaloner D Oppenheimer R Smith J Childhood sexual experiences with adults. A comparison of reports by women psychiatric patients and general-practice attenders Br J Psychiatry 1993 163 499 504 8252289
Palmer RL Bramble D Metcalfe M Oppenheimer R Smith J Childhood sexual experiences with adults: adult male psychiatric patients and general practice attenders Br J Psychiatry 1994 165 675 9 7866685
WHO The ICD-10 Classification of Mental and Behavioural Disorders. Clinical Descriptions and Diagnostic Guidelines 1992 Geneva: World Health Organization
Herman JL van der Kolk B van der Kolk B Traumatic antecedents of borderline personality disorder In Psychological Trauma 1987 Washington, DC: American Psychiatric Press 111 126
Bernet CZ Stein MB Relationship of childhood maltreatment to the onset and course of major depression in adulthood Depress Anxiety 1999 9 169 74 10431682
Cullberg J Stressful life events preceding the first onset of psychosis. An explorative study Nord J Psychiatry 2003 57 209 14 12775296 10.1080/08039480310001373
Read J Ross CA Psychological trauma and psychosis: another reason why people diagnosed schizophrenic must be offered psychological therapies J Am Acad Psychoanal Dyn Psychiatry 2003 31 247 68 12722898 10.1521/jaap.31.1.247.21938
Paris J Childhood trauma as an etiological factor in the personality disorders J Personal Disord 1997 11 34 49 9113821
Brown GR Anderson B Psychiatric morbidity in adult inpatients with childhood histories of sexual and physical abuse Am J Psychiatry 1991 148 55 61 1984707
Ellason JW Ross CA Sainton K Mayran LW Axis I and II comorbidity and childhood trauma history in chemical dependency Bull Menninger Clin 1996 60 39 51 8742671
Fergusson DM Horwood LJ Lynskey MT Childhood sexual abuse and psychiatric disorder in young adulthood: II. Psychiatric outcomes of childhood sexual abuse J Am Acad Child Adolesc Psychiatry 1996 35 1365 74 8885591 10.1097/00004583-199610000-00023
van der Kolk BA Hostetler A Herron N Fisler RE Trauma and the development of borderline personality disorder Psychiatr Clin North Am 1994 17 715 30 7533284
Cascardi M Mueser KT DeGiralomo J Murrin M Physical aggression against psychiatric inpatients by family members and partners Psychiatr Serv 1996 47 531 3 8740498
Cohen P Brown J Smaile E Child abuse and neglect and the development of mental disorders in the general population Dev Psychopathol 2001 13 981 99 11771917
van der Kolk BA Fisler RE Childhood abuse and neglect and loss of self-regulation Bull Menninger Clin 1994 58 145 68 7519094
Muenzenmaier K Meyer I Struening E Ferber J Childhood abuse and neglect among women outpatients with chronic mental illness Hosp Community Psychiatry 1993 44 666 70 8192738
Bierer LM Yehuda R Schmeidler J Mitropoulou V New AS Silverman JM Siever LJ Abuse and neglect in childhood: relationship to personality disorder diagnoses CNS Spectr 2003 8 737 54 14712172
Campo AT Rohner RP Relationships between perceived parental acceptance-rejection, psychological adjustment, and substance abuse among young adults Child Abuse Negl 1992 16 429 40 1617477 10.1016/0145-2134(92)90052-S
Haapasalo J Hamalainen T Childhood family problems and current psychiatric problems among young violent and property offenders J Am Acad Child Adolesc Psychiatry 1996 35 1394 401 8885594 10.1097/00004583-199610000-00026
Herman JL Schatzow E Recovery and verification of memories of childhood sexual trauma Psychoanalytic Psychology 1987 4 1 14 10.1037//0736-9735.4.1.1
Margo GM McLees EM Further evidence for the significance of a childhood abuse history in psychiatric inpatients Compr Psychiatry 1991 32 362 6 1935027 10.1016/0010-440X(91)90086-R
Goodman LA Thompson KM Weinfurt K Corl S Acker P Mueser KT Rosenberg SD Reliability of reports of violent victimization and posttraumatic stress disorder among men and women with serious mental illness J Trauma Stress 1999 12 587 99 10646178 10.1023/A:1024708916143
Dill DL Chu JA Grob MC Eisen SV The reliability of abuse history reports: a comparison of two inquiry formats Compr Psychiatry 1991 32 166 9 2022116 10.1016/0010-440X(91)90009-2
Bessel van der Kolk's Trauma assessment Package
Stefanis CN Reisby N The involvement of the World Psychiatric Association in the issue of the political abuse of psychiatry Acta Psychiatr Scand Suppl 1993 370 14 8 8452050
Loue S The involuntary civil commitment of mentally ill persons in the United States and Romania: a comparative analysis J Leg Med 2002 23 211 50 12079591 10.1080/01947640252987303
Mulder RT Beautrais AL Joyce PR Fergusson DM Relationship between dissociation, childhood sexual abuse, childhood physical abuse, and mental illness in a general population sample Am J Psychiatry 1998 155 806 11 9619154
Coverdale JH Turbott SH Sexual and physical abuse of chronically ill psychiatric outpatients compared with a matched sample of medical outpatients J Nerv Ment Dis 2000 188 440 5 10919703 10.1097/00005053-200007000-00008
Schutzwohl M Maercker A Effects of varying diagnostic criteria for posttraumatic stress disorder are endorsing the concept of partial PTSD J Trauma Stress 1999 12 155 65 10027149 10.1023/A:1024706702133
Maercker A Michael T Fehm L Becker ES Margraf J Age of traumatisation as a predictor of post-traumatic stress disorder or major depression in young women Br J Psychiatry 2004 184 482 7 15172941 10.1192/bjp.184.6.482
| 15566566 | PMC539251 | CC BY | 2021-01-04 16:33:01 | no | BMC Psychiatry. 2004 Nov 26; 4:40 | utf-8 | BMC Psychiatry | 2,004 | 10.1186/1471-244X-4-40 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-4-551556337710.1186/1471-2458-4-55Research ArticleThe lack of public health research output from India Dandona Lalit [email protected] Yegnanarayana S [email protected] Mukkamala N [email protected] VS Udaya [email protected] Rakhi [email protected] Centre for Public Health Research, Administrative Staff College of India, Raj Bhavan Road, Hyderabad – 500 082, India2004 25 11 2004 4 55 55 26 8 2004 25 11 2004 Copyright © 2004 Dandona et al; licensee BioMed Central Ltd.2004Dandona 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
Systematic assessment of recent health research output from India, and its relation with the estimated disease burden, is not available. This information would help understand the areas in health research that need improvement in India to enhance the health of India's population.
Methods
The health research output from India during 2002, which was accessible in the public domain, was assessed by searching PubMed and other internet health literature databases, and was related to the disease burden suggested by the Global Burden of Disease Study. The main outcome measures were number of health papers with abstracts in basic, clinical and public health sciences; quality-adjusted research output based on the impact factors of journals in which the papers were published; classification of papers in disease/condition categories and comparison of research output with the estimated disease burden in each category. Comparison of the health papers from India during 2002 included in PubMed was done with those from Australia during one quarter of 2002.
Results
Of the 4876 health papers from India in 2002 in PubMed, 48.4%, 47.1% and 4.4% were in basic, clinical and public health sciences, respectively. Of the 4495 papers based on original research, only 3.3% were in public health. Quality-adjusted original research output was highest for non-communicable diseases (62% of total). Of the total quality-adjusted original research output, the proportions in injuries (0.7%), cardiovascular diseases (3.6%), respiratory infections (0.2%), diarrhoeal diseases (1.9%), perinatal conditions (0.4%), childhood cluster diseases (0.5%), unipolar major depression (0%), and HIV/AIDS (1.5%) were substantially lower than their proportional contribution to the disease burden in India. Human resources, health policy, health economics, and impact assessment of interventions were particularly poorly represented in public health research. The Australia-India ratio for quality-adjusted health research output per unit gross domestic product was 20 and for public health research output was 31.
Conclusions
Good-quality public health research output from India is grossly inadequate, and strategic planning to improve it is necessary if substantial enhancement of population health were to be made possible. There is inordinately low relative research output in several diseases/conditions that cause major disease burden in India.
==== Body
Background
India suffers a large proportion of the disease burden of the world, which has been estimated to be more than its 16.8% share of the world's population [1,2]. One of the vital elements in improving this situation is the need for a comprehensive and relevant evidence base that would equip India to take informed actions. A systematic assessment of recent health research output from India is not available. Without objective information about the current deficiencies and strengths in the health research output from India, it is difficult to plan substantial improvements in health research output that could enhance India's health status. We analysed the health research output from India in 2002 and related it with the estimated disease burden to identify areas that require particular attention to facilitate effective action to reduce disease burden in this world's second most populous country.
Methods
Health research output was defined as tangible research information related to human health that was readily accessible in the public domain. PubMed [3,4] (which includes MEDLINE) of the US National Library of Medicine, the most widely used online health literature search database in the world, and websites of major academic institutions in India, international agencies, and publishing houses, were searched to ascertain the health research output from India in the year 2002.
PubMed was searched for papers published from India in 2002 using "India" in the author affiliation option in PubMed for all journals, and also by searching the Indian journals in PubMed as several papers in these journals mention only city and state but not India in the author affiliation. Only papers with abstracts were included, as the aim was to review all abstracts and classify the papers in various categories, including type of research, type of paper, disease/condition covered, allopathic or traditional system of medicine, and type and location of first author's institution. PubMed gives institutional affiliation and its location only for the first author. Papers that showed the first author affiliation with an Indian institution were considered as research output from India.
Definitions were used to classify the Indian papers located in PubMed. Health research was defined as research that could be related to health. Basic research was considered either pure or applied, pure being experimental or theoretical work to advance health knowledge without a defined specific application and applied having such an application. Clinical research was categorised as patient series/management if the paper was about clinical cases or issues in management of patients, laboratory if it dealt mainly with laboratory analysis of patient specimens, clinical trial if it was a trial in the clinical setting, and clinical epidemiology if it was about distribution and determinants of disease assessed in the clinical setting. Public health research was categorised into epidemiology, environment/ social, and health systems/policy. Epidemiology included population epidemiology that dealt with study of distribution and determinants of disease and health in the population, and biostatistics/methods that dealt with methodological issues in epidemiology. Environment/social included environmental sciences that dealt with environmental influences on health, and social aspects that dealt with social dimensions of health. Health system/policy included health services that dealt with aspects of health service provision, and health policy that dealt with concepts and frameworks related to the health system. A paper was classified as original research if it had original data collection and its analysis, and review/viewpoint if it was not based on original data. An attempt was made to classify each paper under the disease/condition that it covered, according to the listing used in the Global Burden of Disease Study [2]. If a paper covered generic issue(s) which could not be classified under a particular disease/condition, it was considered unclassifiable for disease/condition.
The 2002 impact factor of the journal, in which each paper was published, was used as a measure of the quality of each paper [5,6]. The proportion of papers and the quality-adjusted output for the diseases/conditions were related to the proportion of burden caused by each disease/condition in India as estimated for 2000 by the Global Burden of Disease Study [2]. The publications of 2002 were related to the disease burden of 2000, as research initiation to publication may take on an average a couple of years.
Percent quality-adjusted research output was calculated for papers in the categories of several classifications as follows:
IndMED [7], an online database of the Indian Medlars Centre, which covers several Indian biomedical journals was also searched. However, this database could not be included in the study, as the abstracts/papers for all the months of 2002 were not included in this database with substantial portions missing.
As reports on commissioned research in public health may be available on the websites of agencies/organisations, the websites of several international agencies (DFID, European Commission, UNAIDS, USAID, WHO, World Bank), twelve academic institutions of India involved with public health, and sixteen publishing houses, were searched to locate research on public health reported in the public domain from India in 2002.
For comparison of the Indian health research output with a developed country, a PubMed search was also done for papers published from Australia during the April–June 2002 quarter, using "Australia" OR "names/abbreviations of the states and territories of Australia [8]" in the author affiliation option in PubMed for all journals as several papers in Australian journals mentioned only city and state but not Australia in the author affiliation.
Data were entered in an MS Access database and analysed using SPSS software.
Results
Of the 5718 papers with abstracts located on PubMed that were published from India in 2002, 842 (14.7%) papers were considered as non-health papers as they were on pure botany, chemistry, physics or zoology that could not be related to human health, and the other 4876 were health papers. The distribution of the types of research and the types of papers for the health papers is shown in Table 1. The basic and clinical science papers predominated, with public health papers comprising a very small fraction (4.4% of the total). The proportion of papers based on original research was substantially lower for public health (68.5%) than for basic sciences (94.4%) and clinical sciences (92.3%); of the total 4495 original research papers, public health made up only 3.3%. 4700 (96.4%) of the total health papers were on the allopathic system of medicine and 176 (3.6%) on the traditional systems of medicine in which the majority were on ayurveda (144 [81.8%]).
Table 1 Distribution of the types of health research and papers from India in 2002 included in PubMed
Type of research No. (%)* of papers Type of paper
Original research No. [%]† (%)‡ Review / Viewpoint No. [%]† (%)‡
Basic science 2358 (48.4) 2227 [49.6] (94.4) 131 [34.3] (5.6)
Pure 525 (10.8) 518 [11.5] (98.7) 7 [1.8] (1.3)
Applied 1833 (37.6) 1709 [38.0] (93.2) 124 [32.5] (6.8)
Clinical science 2296 (47.1) 2119 [47.2] (92.3) 177 [46.3] (7.7)
Patient series / management 1805 (37.0) 1639 [36.5] (90.8) 166 [43.5] (9.2)
Laboratory 283 (5.8) 277 [6.2] (97.9) 6 [1.6] (2.1)
Clinical trials 155 (3.2) 153 [3.4] (98.7) 2 [0.5] (1.3)
Clinical epidemiology 53 (1.1) 50 [1.1] (94.3) 3 [0.8] (5.7)
Public health 216 (4.4) 148 [3.3] (68.5) 68 [17.8] (31.5)
Epidemiology 85 (1.7) 72 [1.6] (84.7) 13 [3.4] (15.3)
Social / environmental 38 (0.8) 31 [0.7] (81.6) 7 [1.8] (18.4)
Health systems / policy 93 (1.9) 45 [1.0] (48.4) 48 [12.6] (51.6)
Other§ 6 (0.1) 0 [0.0] (0.0) 6 [1.6] (100.0)
Total 4876 (100) 4494 [100.0] (92.2) 382 [100.0] (7.8)
*Percent of the total 4876 papers
†Percent of total in each type of paper
‡Percent of total in each type of research
§Papers that could not be classified in the above categories of type of research; these mostly consisted of biographies of persons or organizations
Table 2 shows the distribution of the diseases/conditions covered by the original research papers from India as compared with the estimated disease burden. A large proportion of the basic science papers (49%) were not classifiable into specific disease/condition categories, as they were generic in nature, as compared with 2.9% papers in clinical science and 13% in public health. Overall, the relative proportion of quality-adjusted original research output for non-communicable diseases was higher than their relative contribution to the disease burden, and this was most marked for clinical sciences. However, some major categories/sub-categories within non-communicable diseases were not covered adequately, as a fairly large proportion of research output was on conditions or issues that were not contributing as much to the disease burden. For example, cardiovascular diseases with a disease burden of 11.4% of the total in 2000 had a relatively low quality-adjusted research output of 3.6% of the total. The estimated disease burden due to neuro-psychiatric conditions was 9.6% of the total and the quality adjusted original research output in this category was relatively fair at 8.8%, but the two major sub-categories of unipolar major depression and biopolar disorder that made up 5.2% of the total disease burden had only 0.2% of the total quality-adjusted original research output. A similar mismatch was seen for infectious & parasitic diseases and respiratory infections that had 33.3% of the total quality-adjusted original research output for 33.9% of the total disease burden, but the six major sub-categories under this group contributing 30.1% of the total disease burden had only 11.8% of the total quality-adjusted original research output (Table 2).
Table 2 Distribution of original research health papers from India as compared with the estimated disease burden
Disease / Condition* % DALY loss in 2000* % DALY loss in 2010* No. (%) of original research health papers† % quality-adjusted output for original research health papers‡ No. (%) of original research basic science papers§ % quality-adjusted output for original research basic science papers¶ No. (%) of original research clinical science papers# % quality-adjusted output for original research clinical science papers** No. (%) of original research public health papers†† % quality-adjusted output for original research public health papers‡‡
Communicable, Maternal, Perinatal and Nutritional Conditions 44.2 34.1 950 (28.6) 37.4 397 (34.9) 42.9 484 (23.5) 29.1 69 (53.9) 59.4
Infectious & parasitic diseases 25.9 22.7 762 (22.9) 33.1 358 (31.5) 40.2 362 (17.6) 23.6 42 (32.8) 48.6
Tuberculosis 6.8 7.0 143 (4.3) 7.4 49 (4.3) 7.2 87 (4.2) 5.6 7 (5.5) 29.2
STDs excluding HIV 1.5 1.1 13 (0.4) 0.3 1 (0.1) 0.1 12 (0.6) 0.6 0 (0.0) 0.0
HIV 3.3 6.0 48 (1.4) 1.6 14 (1.2) 1.5 29 (1.4) 1.8 5 (3.9) 1.2
Diarrhoeal diseases 6.7 4.2 34 (1.0) 1.9 17 (1.5) 2.2 16 (0.8) 1.8 1 (0.8) 0.3
Childhood cluster diseases 4.1 2.5 12 (0.4) 0.5 4 (0.4) 0.4 5 (0.2) 0.6 3 (2.3) 0.0
Respiratory infections 8.0 5.0 18 (0.5) 0.2 2 (0.2) 0.1 15 (0.7) 0.4 1 (0.8) 0.0
Lower respiratory infections 7.7 4.9 8 (0.2) 0.1 2 (0.2) 0.1 6 (0.3) 0.2 0 (0.0) 0.0
Maternal conditions 1.4 0.6 84 (2.5) 1.8 17 (1.5) 1.1 60 (2.9) 2.4 7 (5.5) 2.4
Perinatal conditions 6.1 3.9 25 (0.8) 0.4 1 (0.1) 0.1 23 (1.1) 0.8 1 (0.8) 1.0
Nutritional deficiencies 2.9 1.8 45 (1.4) 1.4 8 (0.7) 0.5 22 (1.1) 1.8 15 (11.7) 7.1
Protein energy malnutrition 1.2 0.7 6 (0.2) 0.2 1 (0.1) 0.0 1 (0.0) 0.0 4 (3.1) 3.4
Iron deficiency anaemia 1.5 1.0 10 (0.3) 0.2 1 (0.1) 0.3 6 (0.3) 0.2 3 (2.4) 0.0
Noncommunicable diseases 38.7 47.5 2344 (70.6) 62.0 732 (64.4) 56.5 1555 (75.6) 70.1 57 (44.5) 40.2
Malignant neoplasms 3.8 5.4 370 (11.1) 11.2 118 (10.4) 9.1 251 (12.2) 14.4 1 (0.8) 2.9
Diabetes mellitus 0.8 0.8 129 (3.9) 3.2 64 (5.6) 3.6 57 (2.8) 2.3 8 (6.3) 8.5
Neuro-psychiatric conditions 9.6 11.5 248 (7.5) 8.8 112 (9.9) 10.5 124 (6.0) 6.6 12 (9.4) 11.6
Unipolar major depression 4.0 5.0 0 (0.0) 0.0 0 (0.0) 0.0 0 (0.0) 0.0 0 (0.0) 0.0
Bipolar disorder 1.2 1.4 3 (0.1) 0.2 0 (0.0) 0.0 1 (0.0) 0.1 2 (1.6) 3.5
Sense organ diseases 1.5 2.1 185 (5.6) 4.8 25 (2.2) 2.1 148 (7.2) 7.5 12 (9.4) 8.8
Cataract 1.2 1.7 25 (0.8) 0.9 6 (0.5) 0.3 16 (0.8) 1.3 3 (2.3) 2.5
Cardiovascular diseases 11.4 14.6 203 (6.1) 3.6 38 (3.3) 2.3 159 (7.7) 5.2 6 (4.7) 0.7
Ischaemic heart disease 5.3 7.1 56 (1.7) 0.9 11 (1.0) 0.7 41 (2.0) 1.2 4 (3.1) 0.2
Cerebrovascular disease 2.1 2.7 20 (0.6) 0.3 3 (0.3) 0.2 17 (0.8) 0.4 0 (0.0) 0.0
Respiratory diseases 3.7 4.8 68 (2.0) 1.5 15 (1.3) 1.1 45 (2.2) 1.6 8 (6.3) 3.2
Chronic obstructive pulmonary disease 1.4 2.0 2 (0.1) 0.0 0 (0.0) 0.0 2 (0.1) 0.0 0 (0.0) 0.0
Digestive tract diseases 2.3 2.4 198 (6.0) 5.3 52 (4.6) 3.7 143 (6.9) 7.5 3 (2.3) 1.5
Cirrhosis of liver 1.1 1.2 12 (0.4) 0.6 1 (0.1) 0.6 11 (0.5) 0.6 0 (0.0) 0.0
Congenital anomalies 3.4 3.5 105 (3.2) 1.6 2 (0.2) 0.2 103 (5.0) 3.3 0 (0.0) 0.0
Injuries 17.2 18.4 28 (0.8) 0.7 7 (0.6) 0.6 19 (0.9) 0.8 2 (1.6) 0.4
Unintentional injuries 15.0 15.9 24 (0.7) 0.6 7 (0.6) 0.6 15 (0.7) 0.6 2 (1.6) 0.4
Road traffic injuries 3.5 5.1 2 (0.1) 0.0 0 (0.0) 0.0 2 (0.1) 0.0 0 (0.0) 0.0
Falls 3.5 3.0 0 (0.0) 0.0 0 (0.0) 0.0 0 (0.0) 0.0 0 (0.0) 0.0
Fires 2.1 2.0 4 (0.1) 0.1 2 (0.2) 0.1 2 (0.1) 0.2 0 (0.0) 0.0
Intentional injuries 2.1 2.5 1 (0.0) 0.1 0 (0.0) 0.0 1 (0.0) 0.1 0 (0.0) 0.0
Self-inflicted injuries 1.4 1.7 1 (0.0) 0.1 0 (0.0) 0.0 1 (0.0) 0.1 0 (0.0) 0.0
Total 100 100 3322 (100) 100 1136 (100) 100 2058 (100) 100 128 (100) 100
*According to the Global Burden of Disease Study [2]; only diseases/conditions with disease burden of >1% of the total are listed, plus diabetes mellitus; since all diseases/conditions are not listed, the sum of sub-categories shown may not add up to the total for their categories; DALY is disability-adjusted life year
†Denominator for this percent calculation is 3322, which excludes 1172 papers that were not classifiable into specific disease/condition categories
‡ Based on the total impact factor of 3456.262 for the 3322 original health research papers included in this table
§Denominator for this percent calculation is 1136, which excludes 1091 papers that were not classifiable into specific disease/condition categories
¶Based on the total impact factor of 1757.491 for the 1136 original basic health research papers included in this table
#Denominator for this percent calculation is 2058, which excludes 61 papers that were not classifiable into specific disease/condition categories
**Based on the total impact factor of 1557.811 for the 2058 original clinical health research papers included in this table
††Denominator for this percent calculation is 128, which excludes 20 papers that were not classifiable into specific disease/condition categories
‡‡Based on the total impact factor of 140.96 for the 128 original public health research papers included in this table
Overall, the diseases/conditions that were substantially underrepresented in the relative proportion of quality-adjusted original research output as compared with their contribution to the disease burden were injuries, cardiovascular disease, respiratory infections, diarrhoeal diseases, perinatal conditions, childhood cluster diseases (including measles and tetanus), unipolar major depression, and HIV/AIDS (Table 2).
As the research output was least in public health, a brief description follows to understand this deficiency better. Figure 1 shows the diseases/conditions that were estimated to contribute more than 4% of the total disease burden in 2000 or 2010, and for which the original research output in public health was less than one-third of their proportional contribution to the disease burden estimated for 2010, suggesting that these diseases/conditions needed particular attention. Table 3 shows the distribution of original research in the various areas of public health, which suggests that original research in human resources, health policy, and health economics is relatively more deficient within the already low public health research output. Only six of the original public health research papers were on assessing interventions across the various areas, suggesting that the existing public health research in India has not yet evolved to the stage of methodically assessing the impact of public health interventions, which is a necessary step in the evolution of effective public health action.
Figure 1 Diseases/conditions poorly represented in original public health research relative to their contribution to the disease burden in India.
Table 3 Distribution of the types of original public health research from India
Type of public health research No. (%) of original research public health papers Total impact factor of original research public health papers in each type % quality-adjusted output for all original research public health papers†
Epidemiology 72 (48.6) 59.6 38.9
Population epidemiology 69 (46.6) 56.5 36.9
Biostatistics / Methods 3 (2.0) 3.0 2.0
Environment / Social 31 (20.9) 29.3 19.1
Environmental sciences 14 (9.5) 16.7 10.9
Social aspects 17 (11.5) 12.7 8.3
Health Systems / Policy 45 (30.4) 64.4 42.0
Health services 42 (28.4) 63.3 41.3
Health economics* 8 (5.4) 5.7 3.7
Training / human resources* 5 (3.4) 0.6 0.4
Health policy 3 (2.0) 1.1 0.7
Total 148 (100) 153.3 100
*Health economics and training / human resources are sub-categories of health services
†Based on the denominator of 153.3
Of the total 4876 health papers from India in PubMed for 2002, 1300 (26.7%) were published in Indian journals, but these papers accounted for only 1.5% of the total impact factor of all health papers from India due to the very low impact factors of Indian journals. Among the public health papers 44.4% were published in Indian journals, for clinical sciences papers this was 39.7%, whereas this proportion was much smaller for basic sciences (12.4%).
The highest proportion of quality-adjusted basic research output was by university departments, institutions affiliated with the Council of Scientific and Industrial Research, and technical institutions; the predominant proportion of clinical research was by medical colleges / hospitals; and public health research by medical colleges / hospitals, government departments (due to one paper in a very high impact factor journal), and institutions affiliated with the Indian Council of Medical Research (Table 4). The National Capital Territory of Delhi accounted for the highest health research output among all states / union territories or cities (Table 5). The top ten research producing cities, with 6% of the population of India, produced 75.6% of the quality-adjusted research output, suggesting a concentration of quality research activity in parts of the country.
Table 4 Distribution of health research output from various types of institutions in India
Type of institution All health Basic science Clinical science Public health
No. (%) of health papers % quality-adjusted output for health papers No. (%) of basic science papers % quality-adjusted output for basic science papers No. (%) of clinical science papers % quality-adjusted output for clinical science papers No. (%) of public health papers % quality-adjusted output for public health papers
Medical college / Hospital 2571 (52.7) 33.4 407 (17.3) 11.9 2044 (89.0) 82.1 119 (55.1) 45.9
Indian Council of Medical Research* 159 (3.3) 4.0 52 (2.2) 2.3 78 (3.4) 6.7 28 (13.0) 13.1
Council of Scientific and Industrial Research† 387 (7.9) 13.3 361 (15.3) 18.9 21 (0.9) 1.7 5 (2.3) 0.6
Technical institutions‡ 281 (5.8) 12.2 268 (11.4) 17.7 7 (0.3) 0.6 5 (2.3) 2.3
Paramedical college/institution 164 (3.4) 2.9 154 (6.5) 4.1 8 (0.3) 0.4 2 (0.9) 0.9
University department 813 (16.7) 16.4 743 (31.5) 23.0 53 (2.3) 2.5 16 (7.4) 4.7
NGO / Foundation / Society 69 (1.4) 1.2 12 (0.5) 0.6 33 (1.4) 1.6 22 (10.2) 8.0
Government department 4 (0.1) 0.6 0 (0.0) 0.0 0 (0.0) 0.0 4 (1.9) 17.9§
Industry 31 (0.6) 0.6 28 (1.2) 0.9 3 (0.1) 0.1 0 (0.0) 0.0
Other 397 (8.1) 15.4 333 (14.1) 20.6 49 (2.1) 4.3 15 (5.9) 6.7
Total 4876 (100) 100 2358 (100) 100 2296 (100) 100 216 (100) 100
*Institutions affiliated with the Indian Council of Medical Research [15]
†Institutions affiliated with the Council of Scientific and Industrial Research [16]
‡Indian Institutes of Technology, Indian Institute of Science, and other technical institutions
§Percentage high due to one paper in a very high impact factor journal
The total of basic, clinical and public health papers does not add up to the "all health" papers in all rows, as 6 "other" papers that could not be classified as basic, clinical or public health (Table 1) are not included in this table
Table 5 Distribution of health research output from states and cities in India
State / Union Territory* Population in millions† No. (%)‡ of health papers No. of health papers per million population Total impact factor of health papers % quality-adjusted health research output§ Total impact factor per million population
National Capital Territory of Delhi 13.8 1014 (20.8) 73.5 1216.0 20.8 88.1
Karnataka 52.7 491 (10.1) 9.3 786.8 13.5 14.9
Maharashtra 96.7 573 (11.8) 5.9 710.1 12.2 7.3
Uttar Pradesh 166.0 484 (9.9) 2.9 591.0 10.1 3.6
West Bengal 80.0 362 (7.4) 4.5 510.6 8.7 6.4
Tamil Nadu 62.1 476 (9.8) 7.7 476.7 8.2 7.7
Andhra Pradesh 75.7 299 (6.1) 3.9 461.1 7.9 6.1
Union Territory of Chandigarh 0.9 364 (7.5) 404.4¶ 336.2 5.8 373.6¶
Kerala 31.8 183 (3.8) 5.8 177.5 3.0 5.6
Punjab 24.3 105 (2.2) 4.3 123.4 2.1 5.1
Gujarat 50.6 74 (1.5) 1.5 68.4 1.2 1.4
Madhya Pradesh 60.4 71 (1.5) 1.2 61.9 1.1 1.0
Union Territory of Pondicherry 1.0 66 (1.4) 66.0 53.7 0.9 53.7
Haryana 21.1 95 (1.9) 4.5 44.6 0.8 2.1
Orissa 36.7 41 (0.8) 1.1 43.2 0.7 1.2
Rajasthan 56.5 62 (1.3) 1.1 40.8 0.7 0.7
Jammu And Kashmir 10.1 20 (0.4) 2.0 28.6 0.5 2.8
Assam 26.6 18 (0.4) 0.7 26.0 0.4 1.0
Uttaranchal 8.5 21 (0.4) 2.5 25.7 0.4 3.0
Meghalaya 2.3 9 (0.2) 3.9 13.7 0.2 6.0
Himachal Pradesh 6.1 12 (0.2) 2.0 13.1 0.2 2.1
Andaman & Nicobar Islands 0.4 6 (0.1) 15.0 9.9 0.2 24.8
Goa 1.3 6 (0.1) 4.6 8.6 0.1 6.6
Bihar 82.9 5 (0.1) 0.1 5.5 0.1 0.1
Jharkhand 26.9 5 (0.1) 0.2 3.4 0.1 0.1
Sikkim 0.5 3 (0.1) 6.0 1.6 0.0 3.2
Manipur 2.4 4 (0.1) 1.7 1.5 0.0 0.6
Chhattisgarh 20.8 3 (0.1) 0.1 1.3 0.0 0.1
Arunachal Pradesh 1.1 2 (0.0) 1.8 0.9 0.0 0.8
Tripura 3.2 2 (0.0) 0.6 0.3 0.0 0.1
Top fifteen cities (State / Union Territory)*
Delhi (National Capital Territory of Delhi) 13.8 1014 (20.8) 73.5 1216.0 20.8 88.1
Bangalore (Karnataka) 8.4 258 (5.3) 30.7 598.2 10.2 71.2
Mumbai (Maharashtra) 11.9 393 (8.1) 33.0 499.4 8.5 42.0
Kolkata (West Bengal) 4.6 299 (6.1) 65.0 463.6 7.9 100.8
Hyderabad (Andhra Pradesh) 3.7 233 (4.8) 63.0 404.2 6.9 109.2
Chandigarh (Union Territory of Chandigarh) 0.9 364 (7.5) 404.4¶ 336.2 5.8 373.6¶
Lucknow (Uttar Pradesh) 3.7 272 (5.6) 73.5 332.1 5.7 89.8
Chennai (Tamil Nadu) 4.2 246 (5.0) 58.6 267.6 4.6 63.7
Pune (Maharashtra) 7.2 108 (2.2) 15.0 163.0 2.8 22.6
Varanasi (Uttar Pradesh) 3.1 87 (1.8) 28.1 138.8 2.4 44.8
Thiruvananthapuram (Kerala) 3.2 121 (2.5) 37.8 135.7 2.3 42.4
Mysore (Karnataka) 2.6 74 (1.5) 28.5 96.2 1.6 37.0
Vellore (Tamil Nadu) 3.5 84 (1.7) 24.0 95.6 1.6 27.3
Pondicherry (Union Territory of Pondicherry) 0.7 66 (1.4) 94.3 53.7 0.9 76.7
Visakhapatnam (Andhra Pradesh) 2.2 34 (0.7) 15.5 38.5 0.7 17.5
*Listed in descending order of total impact factor of health papers; the states / union territories of Mizoram, Nagaland, Dadra & Nagar Haveli, Daman & Diu, and Lakshadweep had no publications in PubMed in 2002
†Population for 2001 from the Census of India [14]
‡Percent of the total 4876 health research papers from India in 2002
§Percent of the total impact factor of 5842.055 for all 4876 health research papers from India
¶This high per capita output is likely related to the small population of Chandigarh and the high concentration of academic institutions
Search of websites of major academic institutions in India, international agencies, and publishing houses revealed that substantial original public health research output that was accessible in the public domain was not readily available from these sources. Among the major academic institutions in India involved with public health research, only one was found to have a few reports on health research accessible on its website [9] and another had some health research abstracts on its website [10]. The international agencies had some reports on their websites on India-related health research that were mostly authored by non-Indian authors.
In the April-June quarter of 2002, 1905 health papers published from Australia were located on PubMed, of which 722 (37.9%) were in basic sciences, 954 (50.1%) in clinical sciences, and 229 (12%) in public health. Taking into account the population and total gross domestic product (GDP) adjusted for purchasing power parity (PPP) of Australia and India [1], the quality-adjusted health research output and public health research output were 19.6 and 31 times higher from Australia than India, respectively, per unit GDP adjusted for PPP (Table 6).
Table 6 Comparison of health research output from India and Australia in 2002
India Australia Australia-India ratio
Total Per million population* Per billion GDP-PPP† Total‡ Per million population* Per billion GDP-PPP† Per million population Per billion GDP-PPP
No. of health papers 4876 4.72 1.66 7620 392.78 15.49 83.2 9.3
Impact factor for health papers 5842 5.65 1.99 19231 991.27 39.10 175.3 19.6
No. of basic science papers 2358 2.28 0.80 2888 148.87 5.87 65.2 7.3
Impact factor for basic science papers 3944 3.82 1.35 10598 546.31 21.55 143.1 16.0
No. of clinical science papers 2296 2.22 0.78 3816 196.70 7.76 88.5 9.9
Impact factor for clinical papers 1698 1.64 0.58 7624 393.01 15.50 239.1 26.7
No. of public health papers 216 0.21 0.07 916 47.22 1.86 225.9 25.3
Impact factor for public health papers 193 0.19 0.07 1008 51.95 2.05 277.6 31.0
*Based on 1033.4 million population for India and 19.4 million for Australia in 2001 [1]
†Based on the gross domestic product adjusted for purchasing power parity (GDP-PPP) of US$ 2930 billion for India and US$ 491.8 billion for Australia in 2001 [1]
‡Based on multiplying the number of papers and their total impact factor for the April-June 2002 quarter by four to obtain the estimate for the year 2002
The total of the number and impact factor for basic, clinical and public health papers does not add up to that for the health papers, as 6 "other" papers (with total impact factor 6.012) that could not be classified as basic, clinical or public health (Table 1) are not included in this table
Discussion
The data presented in this paper suggest that the health research output from India is not commensurate with the magnitude and distribution of disease burden. The research output in public health is particularly meagre, which is a major concern as public health sciences are a necessary tool to facilitate improvement in population health. Within this low research output, several diseases/conditions contributing substantially to the disease burden and several major areas of public health importance have relatively less representation. Without dynamic, relevant, good quality and adequate original research in the various aspects of public health it is difficult to imagine how the sub-optimal health status of the Indian population would improve on rhetoric or theoretical concepts alone [14,15].
In this paper we used impact factors for journals as a measure of the quality of papers published in those journals. Although impact factors are not without their limitations, they still offer a tangible, and perhaps the best available, option to compare the quality of publications in journals [6].
We explored several sources where information about health research output from India could be available in the public domain, as the utilisation of research findings is facilitated most if they are readily accessible in the public domain. However, we did not find any source that would add substantially to the information available in the PubMed database. Indeed, there are more Indian health journals than are included in PubMed, but their quality in general is not as high as those included in PubMed with none of them having an impact factor above zero. Non-inclusion in our analysis of the papers published in these journals, therefore, did not bias our assessment of quality-adjusted research output based on impact factors. The relative low quality and impact factor of a large proportion of Indian journals has been discussed previously [16,17]. PubMed lists affiliation of the first author only, and therefore, the analysis presented in this paper includes only those publications in which the first author had Indian affiliation. There would be other publications with non-Indians as first author and Indians as co-author(s), which we estimate to be a very small fraction of those with Indians as first author. In the general context, the PubMed/MEDLINE database has been used previously to assess the health research output from several countries [18-25].
We used the disease burden in India as estimated by the Global Burden of Disease Study [2]. Although the limitations of this Study have been debated previously in the literature, we could not find a better alternative for use for our study, as these were the most comprehensive estimates available for India. In any case, these estimates can be taken only as indicative, and therefore, we highlight only gross deviations of health research output from these trends.
There has been a previous attempt to assess the health research output from India using the Science Citation Index of 1981–85 and relating the number of papers published in journals of various medical/health specialities with the perceived areas of major disease burden [26]. However, review of all published abstracts to classify each paper in various categories, the approach used by us, has not been used previously to assess health research output from India to our knowledge. Systematic tracking of health research output, and its relation to the estimated trends in disease burden, are necessary for guiding further appropriate development of health research in India. In addition to the overview of research needs identified in this paper, more in-depth assessment of research needs for major diseases/conditions would also be necessary, as was reported recently for the evidence base needed to control HIV/AIDS in India [27].
Since public health sciences seem to be the weakest link in improving health in India currently, it is imperative that a strategic framework for developing original public health research in India be evolved. To do so, the demand, supply and environment issues would have to be addressed:
• Demand. Among the multitude of factors that influence the demand for relevant public health research, the role of policy makers and senior health academics is of particular importance. This is seriously sub-optimal in India at present. Political compulsions push many policy makers into short-term gains instead of investments in comprehensive research for long-term benefits. Although there has recently been an increasing trend in India towards commissioned research by government and international agencies in some aspects of public health, this by itself is not enough to boost comprehensive public health research in India, and the reports of such studies are many times not available in the public domain which reduces the chance of their widespread utilisation. Many senior health academics in India continue to disregard public health research as a less-respectful cousin of basic and clinical research. Systematic efforts are needed to demonstrate to these groups the linkages between all aspects of health research (basic, clinical and public health), and the linkages between public health research and improvements in population health, in order to boost the demand for relevant and good-quality public health research in India.
• Supply. Enhancing the output of public health research will require effort on various fronts. Establishing schools of public health and other institutions to train quality scientists in public health is a priority, as India has a surprisingly few number of institutions that can provide proper training in public health research. Another area that needs quick attention is to make public health exposure in medical and paramedical colleges more practical to encourage hands-on investigative thinking, as currently it is so theoretical that it rarely inspires enthusiasm in young professionals towards public health research. Setting higher standards for the research dissertations currently required for post-graduate degrees in preventive and social medicine would also encourage better quality and practically relevant public health research. It is also necessary to systematically develop performance-based opportunities to public health research scholars for career enhancement. Another element that would help develop public health research capacity in India is evolving mechanisms to encourage contribution to this effort by the many Indian public health researchers living abroad.
• Environment. A conducive environment is necessary for the demand and supply of public health research to function optimally. Efforts are needed to develop this by attempting to develop broad-based coalitions, that include health care providers, civil society and non-governmental sector, for-profit private sector and industry, and national and international agencies providing financial support, which would understand and support the need for vibrant public health research as a vital element of societal development. This is a necessary element that has so far received scant attention, which must be addressed if sustainable development of public health research to improve population health is to become possible in India. An environment of good-quality and comprehensive public health research in India would also infuse the much-needed originality in teaching public health sciences and their practical application to the local context.
Evolving such frameworks would require building up a critical momentum for this effort through perseverance and wisdom. One such opportunity is provided by the recent initiative of the Indian Ministry of Health and Family Welfare to develop more effective institutes of public health in India, with relevant public health research and its utilisation an important key to improving population health [28].
The recent attention towards revitalising the academic aspects of health care / medicine through evidence [29] and evidence-based global health [30] is particularly relevant for developing nations. Evolving a strong, dynamic and locally-relevant evidence base is even more important for developing nations as this is likely to yield relatively higher returns by contributing to improvements in the health, lives and economy of a larger proportion of the world's population. For this to happen, theoretical concepts alone would obviously be not enough. The practical solutions for this effort would have to be developed wisely. The data and its interpretation presented in this paper are, we hope, an example of how the deficiencies in the evidence-base needed for adequate health care in developing nations can be understood objectively in order to plan its strengthening.
Conclusions
• Publications from India in PubMed were 11 times less in public health than those in basic sciences and in clinical sciences in 2002.
• Injuries, cardiovascular diseases, respiratory infections, diarrhoeal diseases, perinatal conditions, childhood cluster diseases, unipolar major depression, and HIV/AIDS had substantially less proportion of quality-adjusted original research output in India as compared with their contribution to the disease burden.
• India produced 20 times less quality-adjusted health research output than Australia per unit gross domestic product adjusted for purchasing power parity, and this ratio for public health research output was even higher at 31 times.
• Good-quality public health research output from India is grossly inadequate, and strategic planning to improve it is necessary if substantial enhancement of population health were to be made possible.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LD conceived the idea of this study, guided the design, data collection and analysis, and wrote the initial draft of this paper. YSS contributed to the design, data collection and analysis. MNJ contributed to data collection and analysis. VSUB contributed to data management and analysis. RD contributed to the idea of this study, design and data analysis. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
United Nations Development Programme Human Development Report 2003 New York 2003
Murray CJL Lopez AD The Global Burden of Disease 1996 Boston: Harvard University Press
National Library of Medicine, National Institutes of Health, USA PubMed
Canese K Jentsch J Myers C PubMed: The Bibliographic Database In The NCBI Handbook 2003 Bethesda: National Center for Biotechnology Information, USA
Institute for Scientific Information Journal Citation Reports: Impact Factor Philadelphia 2003
Garfield E Journal impact factor: a brief overview Can Med Assoc J 1999 161 979 980 10551195
Indian Medlars Centre, National Informatics Centre, Department of Information Technology IndMED
Australian Government Australian States & Territories Map
National Council for Applied Economic Research, New Delhi Publications
Sree Chitra Tirunal Institute for Medical Sciences & Technology, Thiruvananthapuram Public Health – Publications
Indian Council of Medical Research (ICMR) ICMR's Permanent Institutes/Centres
Council of Scientific & Industrial Research (CSIR) CSIR Network of R&D Laboratories
Census of India Census 2001 – Results
Dandona L Conceptualizing health policy Natl Med J India 2002 15 226 231 12296481
Dandona L Public health sciences need strengthening in developing countries BMJ 2004 328 47 14703555 10.1136/bmj.328.7430.47-a
Sahni P Reddy PP Kiran R Reddy KS Pande GK Nundy S Indian medical journals Lancet 1992 339 1589 1591 1351561 10.1016/0140-6736(92)91844-X
Vohora SB Vohora D Why are Indian journals' impact factors so low? Nature 2001 412 583 11493892 10.1038/35088214
Favaloro EJ Medical research in New South Wales 1993–1996 assessed by Medline publication capture Med J Aust 1998 169 617 622 9887911
Rosselli D Latin American biomedical publications: the case of Colombia in Medline Med Educ 1998 32 274 277 9743781 10.1046/j.1365-2923.1998.00214.x
Bunout D Reyes H Biomedical papers written by Chilean authors published in international journals in 1997. A review from MEDLINE Rev Med Chil 1998 126 677 688 9778877
Hefler L Tempfer C Kainz C Geography of biomedical publications in the European Union, 1990–98 Lancet 1999 353 1856 10359422 10.1016/S0140-6736(99)01278-7
Thompson DF Geography of U.S. biomedical publications, 1990 to 1997 N Engl J Med 1999 340 817 818 10075537 10.1056/NEJM199903113401020
Rahman M Laz TH Fukui T Health related research in Bangladesh: MEDLINE based analysis J Epidemiol 1999 9 235 239 10510580
Han MC Lee CS Scientific publication productivity of Korean medical colleges: an analysis of 1988–1999 MEDLINE papers J Korean Med Sci 2000 15 3 12 10719802
Tadmouri GO Bissar-Tadmouri N Biomedical publications in an unstable world: the Arab world, 1988–2002 Lancet 2003 362 1766 14643139 10.1016/S0140-6736(03)14868-4
Arunachalam S Does India perform medical research in areas where it is most needed? Natl Med J India 1998 11 27 34 9557519
Dandona L Enhancing the evidence base for HIV/AIDS control in India Natl Med J India 2004 17 160 166 15253404
Ministry of Health and Family Welfare, Government of India National Consultation on Institutes of Public Health in India – From Concept to Reality: 16–17 September 2004 2004 New Delhi: Ministry of Health and Family Welfare
International Working Party to Promote and Revitalis Academic Medicine Academic medicine: the evidence base BMJ 2004 329 789 792 15459055 10.1136/bmj.329.7469.789
Buekens P Keusch G Belizan J Bhutta ZA Evidence-based global health JAMA 2004 291 2639 2641 15173158 10.1001/jama.291.21.2639
| 15563377 | PMC539252 | CC BY | 2021-01-04 16:28:48 | no | BMC Public Health. 2004 Nov 25; 4:55 | utf-8 | BMC Public Health | 2,004 | 10.1186/1471-2458-4-55 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-4-571557162210.1186/1471-2458-4-57Research ArticleThe family as a determinant of stunting in children living in conditions of extreme poverty: a case-control study Reyes Hortensia [email protected]érez-Cuevas Ricardo [email protected] Araceli [email protected] Raúl [email protected] José Ignacio [email protected] Svetlana V [email protected]érrez Gonzalo [email protected] Unidad de Investigación Epidemiológica y en Servicios de Salud, Centro Médico Nacional Siglo XXI, Instituto Mexicano del Seguro Social, México City, México2 Coordinación de Políticas de Salud, Dirección de Prestaciones Medicas, Instituto Mexicano del Seguro Social, México City, México3 Departamento de Epidemiología, Secretaría de Salud Estado de Guerrero, Chilpancingo, México4 Dirección General, Hospital Infantil de México "Federico Gómez", México City, México5 Unidad de Salud Pública. Dirección de Prestaciones Medicas, Instituto Mexicano del Seguro Social, México City, México2004 30 11 2004 4 57 57 5 3 2004 30 11 2004 Copyright © 2004 Reyes et al; licensee BioMed Central Ltd.2004Reyes 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
Malnutrition in children can be a consequence of unfavourable socioeconomic conditions. However, some families maintain adequate nutritional status in their children despite living in poverty. The aim of this study was to ascertain whether family-related factors are determinants of stunting in young Mexican children living in extreme poverty, and whether these factors differ between rural or urban contexts.
Methods
A case-control study was conducted in one rural and one urban extreme poverty level areas in Mexico. Cases comprised stunted children aged between 6 and 23 months. Controls were well-nourished children. Independent variables were defined in five dimensions: family characteristics; family income; household allocation of resources and family organisation; social networks; and child health care. Information was collected from 108 cases and 139 controls in the rural area and from 198 cases and 211 controls in the urban area. Statistical analysis was carried out separately for each area; unconditional multiple logistic regression analyses were performed to obtain the best explanatory model for stunting.
Results
In the rural area, a greater risk of stunting was associated with father's occupation as farmer and the presence of family networks for child care. The greatest protective effect was found in children cared for exclusively by their mothers. In the urban area, risk factors for stunting were father with unstable job, presence of small social networks, low rate of attendance to the Well Child Program activities, breast-feeding longer than six months, and two variables within the family characteristics dimension (longer duration of parents' union and migration from rural to urban area).
Conclusions
This study suggests the influence of the family on the nutritional status of children under two years of age living in extreme poverty areas. Factors associated with stunting were different in rural and urban communities.
Therefore, developing and implementing health programs to tackle malnutrition should take into account such differences that are consequence of the social, economic, and cultural contexts in which the family lives.
==== Body
Background
Despite improvements in the health of children under five years of age, malnutrition remains as an important public health problem in Mexico [1] and in other developing countries [2-7]. Malnutrition in young children affects linear and brain growth and intelligence quotient, and is synergistically associated with child morbidity and mortality [8-13].
Previous reports have stated that malnutrition occurs mainly in rural areas and worsens under conditions of extreme poverty [4,14,15]. In Mexico, malnutrition is more frequent in the southern states, which are underdeveloped and have a predominantly indigenous population living in poor housing with unsanitary conditions [16]. The most recent nutritional survey at national level showed that the prevalence of stunting decreased during the last decade from 23% to less than 18%. Stunting is defined as the proportion of children under five years old whose height-for-age is less than -2 standard deviations of the reference population median. Wasting (the proportion of children under five years old whose weight-for-height is less than -2 standard deviations of the reference population median) has declined from 6% to 2%. However, analysis shows that prevalence of stunting is higher in rural (31.7%) than in urban strata (11.6%) [17]. The extent of malnutrition in extremely impoverished rural areas has fuelled implementation of public health programs aimed at improving children's nutritional status [18], but the impact of these programs has not been completely evaluated.
Poverty is not confined to rural areas, nor is malnutrition; they are also present in urban environments [19]. Uncontrolled and unplanned growth of cities has fostered the emergence of urban slums that lack basic sanitary services. In such settings, underprivileged families live overcrowded and malnutrition is undoubtedly present. Unfortunately, specific information about nutritional status of children living in urban slums cannot be obtained from existing epidemiological data. The extent of child malnutrition in these areas is probably underestimated because of a flawed health information system.
Malnutrition is an outcome of various factors resulting from unfavourable socioeconomic circumstances such as difficulties in obtaining food, unemployment -which determines an irregular income for the family's breadwinner-, limited access to education and health services, or illness caused by unsanitary conditions [2,5,14,20-24]. These circumstances are worsened by unequal access to, and distribution of resources among members of the family. However, some families are able to cope with such adverse environments and to maintain their children in an adequate nutritional status. The purpose of this study was to ascertain whether family-related variables constitute risk factors for stunting among young Mexican children living in extreme poverty, and whether these factors differ between rural and urban settings.
Methods
Study design
A case-control study was conducted from August through December 1998. Two extreme poverty level areas (one urban and one rural) were selected. Each area was analysed separately.
Setting
The study took place in the south-western State of Guerrero, one of the poorest in Mexico, with a population of nearly three million.
The rural area, named Alto Balsas, is located next to the Balsas River; the urban area, named El Sinai, is a poor neighbourhood on the outskirts of the port of Acapulco, which has half a million inhabitants. For the rural area, four villages with a combined population of 4,638 were included in the study. The closest city is approximately 45 miles away. The principal language spoken in these villages is Náhuatl, and agriculture is the main economic activity. Because of seasonal conditions there is a high rate of cyclic migration each year. This migration occurs during harvest season, where farmers migrate from the State of Guerrero to another states. There is one medical doctor and one primary care centre per approximately 2300 people.
El Sinai is located in the outskirts of Acapulco and has 6,860 inhabitants, most people speak Spanish and they work in the local industry as unskilled manual workers. In this area, there is one medical doctor and one clinic per 3400 people. In both study areas, sanitation and public services are deficient. Table 1 shows the characteristics of both study areas.
Table 1 Characteristics of study areas
Alto Balsasa El Sinai
Rural area Urban area
Population 4,638 6,860
Distance to nearest urban centre 45–60 km -----
Number of schools
Kindergarten 4 2
Elementary school 4 2
Secondary school 0 1
Health Facilities (belonging to the Ministry of Health) Rural clinic (2)
Health post (2) Urban clinic (1)
Health Personnel (employed by the Ministry of Health)
Physicians 2 2
Nurses 3 2
Primary care technicians 2 2
Indigenous language 40–70% 4%b
Economic activities
Agriculture 75–90% 0
Industry 5–20% 66%
Services 5–25% 12%
Annual migrationc 30–50% 10%
a Four communities
b Estimations based on data from this study
cProportion of families who migrate seasonally every year
Study population
The units of study were children between 6 and 23 months of age. Children older than 6 months were included on the assumption that this is the minimum amount of time the child is exposed to family-related factors; also, this is the average age in which children are weaning (this is defined as the time when mothers begin to introduce food other than milk into the child's diet). Thus infants become more exposed to environmental causes of under-nutrition. The family was defined as next of kin, or persons sharing the same household and food expenses [25]. Only one child from each family was included in the study; if there was more than one eligible child, the oldest was selected. Infants with congenital diseases or those with low birth weights (less than 2,500 g) were excluded.
Case definition
Cases were stunted children. This was ascertained by using the height-for-age indicator [26]. The criterion for stunting was: Z-value less than -2.00 standard deviations (SDs) below the median height-for-age [27].
Control definition
Controls were children without stunting: equal to or above -2.00 SDs below the median in families that had no stunted children.
The sample size was calculated by using the case-control study formula in accordance with the following assumptions [28]: α = 0.05, β = 0.20, minimum risk to be detected = 2.5, proportion of controls exposed to the least frequent variable (migration) = 0.15, and case-control ratio 1:1. The required sample size was 112 children per group in each area.
Study variables
A five-dimension framework to identify family-related factors that might have influence on the child's nutritional status was constructed. Each dimension comprises several variables selected to build up a comprehensive scenario.
The dimensions are the following:
1. Family characteristics: parents' age and literacy, type and duration of parents' union; family structure (nuclear or extended); presence of both parents (complete or incomplete family); number of members of the family and mother's use of contraceptives.
2. Family income: parents' employment (type of job, time on the same job, per capita family income), unemployment and migration during the past two years.
3. Household allocation of resources and family organisation: time spend by the mother to care for the child and to do domestic activities, and the way the family distributed and shared its income (percentage of income spent in food, clothes, transportation, rent, etc);
4. Social networks: type of networks (within or outside the family), size of networks, frequency of interactions and type of support (economic, child care, etc.).
5. Child health care: patterns of breast-feeding and health-care-seeking for preventive care (immunisations and Well Child Program visits) or curative care.
In addition, the following variables were included: child characteristics (age, sex, birth order and birth weight); and housing characteristics (type of construction, crowding conditions, indoor plumbing, sewerage system, and whether the kitchen was in a separate room).
Data collection
In each study area, a local health worker able to communicate in Náhuatl and Spanish was trained to carry out a census to identify children fulfilling inclusion criteria, to interview the mother and to obtain the anthropometric measurements. Data were collected by using a pro-forma. The mother or caregiver was personally interviewed. During the visit, the interviewer measured and recorded the height and weight of each child aged between 6 and 23 months. The weight was measured using a digital scale with a precision error of ± 1 oz. The mother helped to measure the recumbent length of her child and this was done by using a portable calibrated board. To assure accuracy and reliability, one of the researchers (HR or RC) visited 10% of the households within the following week to confirm the data. There were no inconsistencies in data or anthropometric measures that could affect the results.
Mexican Institute of Social Security IRB and Ministry of Health authorities of the State of Guerrero approved the study. Local community leaders accepted and collaborated in the study and each family head as well as child's mother gave their informed consent.
Statistical analysis
Analysis was carried out separately for each area. Firstly, a bivariate analysis was run; crude odds ratios (OR) and 95% confidence intervals (95%CI) were calculated for each variable in every dimension. The analysis included estimates of correlation and interaction among variables. Secondly, to obtain the best predictive model for stunting, all statistically or conceptually significant variables within each dimension were included in an unconditional multiple logistic regression analysis; the method selected for modelling started from a saturated model until finding the best explanatory model, after assessing the significance of each covariate and adjusting for major potential confounders such as age and sex of the child, literacy of the mother and household income. Once the best explanatory model was found, goodness of fit assessment was performed. The statistical analysis was carried out using SPSS (SPSS Professional Statistics 7.5 SPSS Inc. 1997) and STATA (STATA Statistical Software: Release 5.0 Stata Corporation, 1997).
Results
Through the census, 326 eligible children aged between 6 and 23 months living in the rural area and 448 in the urban area were identified. Two hundred and forty-seven children (75.8%) from the rural area and 409 (91.3%) from the urban area who fulfilled the inclusion criteria were located. Because of the harvest season, some families living in the rural area migrated temporarily, so the remaining 24.2% of children could not be located.
Regarding the children living in the rural area 43.7% were stunted. Therefore, the group was divided into 108 cases and 139 controls; as to the urban area, 48.4% of children were stunted, resulting into 198 cases and 211 controls.
Table 2 shows children and housing characteristics of both study groups.
Table 2 Children and housing characteristics of study groups
Variable Rural Urban
Cases n = 108 % Controls n = 139 % Cases n = 198 % Controls n = 211 %
Child characteristics
Age in months (median, min-max) 16 (6–23) 13 (6–23)** 15 (6–23)| 12 (6–23)**
Sex (male) 53.7 53.2 58.1 48.4*
First birth order 11.1 20.9* 26.8 35.1*
Birth weight (g) mean (SD+) 2947(444) 2984(355) 3178(485) 3370(484)
Housing characteristics
Dirt floor 87.0 85.6 40.9 31.1*
No indoor plumbing 78.7 71.9 60.6 49.3*
No sewerage system 100.0 100.0 39.4 28.0
No separate kitchen 87.0 79.9 57.1 44.1**
Overcrowding 79.6 64.7 48.0 33.2**
In the rural area, the proportion of first birth order children in the group of cases was lower than in the control group. Housing conditions were similar in both groups. Regarding urban children, cases had poorer housing conditions than controls; also, cases were older than controls in both study areas.
Table 3 shows the distribution of variables within each dimension.
Table 3 Distribution of variables by dimensions
Variable Rural Urban
Cases n = 108 % Controls n = 139 % Cases n = 198 % Controls n = 211 %
Dimension 1. Family characteristics
Mother's age (years)
Median (min-max) 30 (18–43) 27 (15–44) 25 (15–43) 24 (16–47)
Father's age (years)
Median (min-max) 32 (19–45) 30 (15–60) 28 (18–56) 28 (18–60)
Illiteracy of mother 50.0 37.4* 62.9 44.8**
Illiteracy of father 65.7 53.2* 50.3 46.2
Parents' civil status (married) 85.2 72.7** 63.1 68.2
Duration of parents' union longer than two years 95.4 84.2** 15.6 27.5**
Type of family (nuclear) 61.1 57.6 68.2 71.6
Completeness of the family 94.4 93.5 90.9 88.6
Size of the family (number of members) median (min-max) 8 (3–16) 7 (2–16) 5 (3–14) 5 (2–18)
Mother's use of contraceptive method 13.9 26.6* 58.7 69.6*
Dimension 2. Family income
Mother's occupation
Housewife 80.6 76.3 86.4 84.8
Other 19.4 23.7 13.6 15.2
Father's occupation
Worker 13.0 20.9 27.8 20.4*
Farmer 65.721.3 53.2* 17.7 22.3
Other 25.9 55.5 57.3
Time during which father has worked in the same place (months) Median (min-max) 7 (1–30) 8 (0–36) 4 (1–60) 4 (1–30)
Per capita family income per month (USD)
Mean (SD) 22.8 (11.7) 25.2 (6.0) 39.5 (18.6) 45.2 (24.1)*
migration of parents
From rural to urban settings 71.2 55.8*
From rural to rural 2.8 8.6
Dimension 3. Household allocation of resources and family organisation
Child care provided exclusively by the mother 80.6 89.9* 89.9 87.2
Time spend by the mother to care for the child (hours/day) Median (min-max) 5 (3–17) 6 (0–21) 5 (0–21) 5 (1–14)
% of family income spent in:
Food. Mean (SD+) 39.0 (13.5) 37.5 (14.9) 52.0 (15.6) 50.7 (14.8)
Transportation. Mean(SD+) 8.5 (5.9) 8.1 (4.1) 12.3 (7.3) 12.7 (7.0)
Dimension 4. Social networks
Lack of social networks 24.1 24.5 14.1 14.7
Size of network (Small) 63.0 59.0 75.3 63.0**
Type of support
Child care 64.8 50.7* 62.1 64.5
Economic 62.0 63.3 22.2 14.2**
Dimension 5. Child health care
Breast feeding (months)
Median (min-max) 7 (0–12) 6 (0–17) 4 (2–15) 3 (1–12)
Age at weaning (months)
Median (min-max) 7 (1–14) 6 (1–13)** 4 (2–15) 4 (1–12)
Complete immunisation scheme 83.3 87.8 77.3 82.5
Attendance to the Well Child Program activities (number of visits)
Median (min-max) 2 (0–4) 2 (0–9) 2 (0–6) 3 (0–8)**
* p < 0.05, **p < 0.01 (between cases and controls within the same area)
+ Standard deviation
In the rural area, the following variables showed statistically significant differences when comparing cases and controls: parents' illiteracy, parents' civil status, duration of parental union, mother's use of contraceptive method, father engaged in farming activities, exclusive provision of care by the mother, social networks for child care, and age at weaning. At the urban area, mother's illiteracy, duration of parental union longer than two years, father's occupation, per capita family income, parents' migration from rural areas, size of social networks and frequency of attendance to the Well Child Program activities, were statistically significant variables.
Table 4 shows the results of the crude analysis of each dimension. In the rural area there was at least one significantly associated variable, while in the urban area there were two or more, except in dimension 3 (household allocation of resources and family organisation), in which no statistically significant variables were found.
Table 4 Variables associated with stunting, bivariate analysis
Variable Odds Ratio Confidence interval (95%) p value
RURAL AREA
Dimension 1. Family characteristics
Duration of the parents' union:
longer than two years 3.81 1.38 – 10.53 .01
two years or less 1.00
Mother not using a contraceptive method 2.22 1.13 – 4.34 .02
Mother using contraceptive method 1.00
Dimension 2. Family income
Father occupation: Farmer 1.68 1.00 – 2.83 .04
Another type of job 1.00
Dimension 3. Household allocation of resources and family organisation
Child care: provided exclusively by mother 0.46 0.22 – 0.96 .03
shared with other caretakers 1.00
Dimension 4. Social networks
Family networks for child care 1.79 1.06 – 3.00 .02
Without family networks 1.00
Dimension 5. Child health care
Weaning: after six months of age 2.22 1.33–3.70 .002
at/before six months of age 1.00
URBAN AREA
Dimension 1. Family characteristics
Illiteracy of mother 2.09 1.40 – 2.14 .001
Non-illiteracy of mother 1.00
Duration of parents' union:
longer than two years 2.04 1.24 – 3.35 .005
two years or less 1.00
Family with more than 4 members 1.52 1.03 – 2.26 .03
Family with 4 members or less 1.00
Mother not using a contraceptive method 1.61 1.06 – 2.42 .02
Mother using a contraceptive method 1.00
Dimension 2. Family income
Per capita family income:
below $25 USD per month 1.65 1.03 – 2.64 .03
above $25 USD per month 1.00
Father engaged at the same work place:
equal or less than 2 years 1.84 1.19 – 2.82 .005
more than 2 years 1.00
Parents migration:
migrant from rural to urban area 1.96 1.28 – 2.99 .002
non migrant 1.00
Dimension 4. Social networks
Small networks 1.78 1.16 – 2.73 .008
Other size or without networks 1.00
Family networks for economic support 1.72 1.03 – 2.87 .03
Without family networks 1.00
Dimension 5. Child health care
Number of visits to the Well Child Program:
Less than two 2.43 1.58 – 3.74 .0001
Two or more 1.00
Breast feeding: longer than six months 2.23 0.98 – 5.10 .05
six or less months 1.00
Additionally, two of the child's characteristics showed significance (data not presented in the table): child's age (rural area, OR 4.5, CI95% 2.59 – 8.11; urban area, OR 1.98, CI95% 1.29 – 2.92), and child's sex (urban area, OR 1.45, CI95% 0.98 – 2.14; rural area not significant)
Table 5 presents the final explanatory models for each area after adjusting for some established risk factors (child's age and sex, maternal literacy, and household income). In the rural area, the model included only three variables belonging to dimensions two, three, and four. Dimension two (family income) showed an increased risk when the father's occupation was farmer. Dimension three (household allocation of resources and family organisation) showed that one covariate related to the mother's activities (the child being cared for exclusively by the mother) had a protective effect. Dimension four (social networks) showed that having family networks to provide care for the child entailed a higher risk.
Table 5 Variables associated with stunting, multivariate analysis*
Variable Odds Ratio Confidence interval (95%) p value
RURAL AREA**
Father's occupation: farmer 1.77 0.98 – 3.18 .05
Child care provided exclusively by the mother 0.30 0.13 – 0.69 .004
Family networks for child care 2.31 1.28 – 4.15 .005
URBAN AREA**
Duration of parents' union longer than two years 1.89 1.01 – 3.54 .04
Parents migration from rural to urban area 1.57 0.95 – 2.59 .07
Father worked at the same place for 2 years or less 3.23 1.88 – 5.56 .0001
Small networks 2.11 1.27 – 3.49 .004
Less than 2 visits to the Well Child Program 2.57 1.54 – 4.30 .0001
Breast feeding longer than six months 1.71 0.62 – 4.73 .29
* Unconditional logistic regression analysis
**Adjusted by age and sex of the child, maternal literacy, and per capita family income in both areas
The explanatory model for the urban area included several covariates as risk factors in most dimensions. Dimension 1: migration of parents from rural to urban area, and duration of the parents' union longer than two years. Dimension 2: the father being in the same employment for two years or less. Dimension 4: having small family networks. Dimension 5: child having fewer than two visits to the Well Child Program activities during the past 6 months, and breast-feeding longer than six months. Age showed to be a significant confounder, and other characteristics of the child such as sex or birth weight, or those characteristics related to the parents such as literacy or per capita family income did not change the significance of the model when adjusted.
Discussion
The role of the family as an important influence for nutritional status of children has been increasingly emphasized during the past few years, [29] and reinforced by the household production function perspective [30]. This is defined as "a dynamic process that occurs within the household to allow family members to combine their knowledge, resources and patterns of behaviour, either to promote, recover, or maintain health status" [31].
Nutritional status is an indicator of well being and malnutrition is the result of a complex process within which coexist a number of variables. The results in this study showed that the family related factors for stunting were different in each context -urban or rural-. The model highlighted the importance of identifying, [rather arbitrarily] a number of conceptually grounded dimensions that could be associated with stunting.
As mentioned earlier, rural and urban environments are different. Regarding the results of the rural area, within the family income dimension, the variable father's occupation (farming) was found to be a risk factor for child's stunting. In this area, availability of food depends upon local production, which in turn is related to father's occupation. Maize is the staple, and there is lack of food variability or inability to produce sufficient food for the family's nutritional requirements [32].
In the urban area the instability of the father's employment (working in the same place for less than 2 years) was found to be a risk factor. Lack of stable employment is a common problem among unskilled workers; their income is low and irregular, thus affecting their capacity to purchase goods and food, which in turn affects child nutrition [33]. Nevertheless, the other important variable included in this dimension, per capita family income, which was statistically significant in the bivariate analysis, did not show significance in the multivariate analysis. This finding could be interpreted as the ability of the family to cope with a difficult environment. Analysis of microeconomic variables shows the need of further studies to confirm plausibility of our findings and its association with malnutrition.
The dimension of family structure, which included socio-demographic characteristics of the parents, migration, and literacy attainment, has been repeatedly associated with the nutritional status of children. [2,5,20,23,34,35] In the urban area, some indicators of this dimension were associated with higher risk of malnutrition, particularly the longer duration of parents' union. It is possible that the economic and social burden on poor urban families with several children led the mother to give less attention to her younger children, whose nutritional status suffered in consequence. This result highlights the importance of considering the reproductive health sphere when developing health programs.
An interesting finding in this study was the relationship between social networks and children's nutritional status. It is often assumed that extended families have extensive social networks, and that this can be advantageous for the care of young children. However, we found quite the opposite in the rural area: the presence of family networks was associated with stunting, while exclusive provision of child-care by the mother showed a protective effect. These findings stress the role of the mother as primary caregiver, at least for young children; such findings also suggest that when a mother is present to care for the child, some of the effects of living in a poor community can be ameliorated.
In the urban area, conditions seem to be different, and the sizes of social networks influenced the nutritional status of the children positively. Urban families living in a more aggressive environment might be more likely to obtain health benefits if they have greater family or social support.
Use of health services, based on preventive measures from the Well Child Program, was also analyzed. Few visits to the Well Child Program activities were associated with stunting in the urban area. Restricted access to health care services in deprived areas is an important shortcoming with regard to monitoring child's growth and health [36]. Based on these findings, promotion of comprehensive services for children living in extreme poverty is relevant. Most health workers carry out preventive activities, among which nutritional status monitoring is essential. Therefore, it seems advisable to train these workers to recognise families showing some of the risk factors described in this study, so malnourished or at-risk children can be identified.
Migration was also explored. In the final explanatory model addressing the urban area, parents' migration from rural to urban area was found to be a risk factor for stunting. Migration from rural to urban settings is frequent among young people looking for improving their living conditions. However, families' adaptation to the new situation is a long process, given that such families live in a hostile environment and they have limited access to information regarding how to care for the child. In contrast, migration of one member of the family from one urban setting to another did not show a relationship with nutritional status of children. Adult male family members, mainly the heads of household, migrate to other states in Mexico or to the U.S. in search of better income. Perhaps the influence of this variable could be identified by a qualitative approach. Therefore, further research on this aspect is necessary.
The study has some limitations. Firstly, the study groups were of unequal size; this is a consequence of including all children aged from 6 to 23 months of age identified by the census. Additionally, since children of the control group were significantly younger than cases, some of them could have become stunted by the time they were of comparable age to the cases. To solve this limitation, the covariate age was adjusted in the multivariate analysis. However, it is possible that an age-matched cases-control design would have been preferable.
Secondly, the fact that almost 25% of eligible children in the rural area were not included due to migration of their families could have led to selection bias.
Another limitation is the lack of precise information regarding food practices among participating families. The interview allowed identifying general aspects such as duration of breast-feeding and age of weaning. Breast-feeding beyond 6 months of age entailed a risk of stunting, but only in the urban area. This issue has been controversial; some authors suggest that the relationship between prolonged breast-feeding and malnutrition may indicate a maternal decision to continue breast-feeding to a nutritionally disadvantaged child [37-39], rather than being the direct effect of prolonging breast-feeding on the child's nutritional status [40]. Further research in this area will contribute to the knowledge of cultural preferences about breast-feeding and its consequences on children's nutrition.
Conclusions
This study suggests the influence of the family on the nutritional status of children under two years of age living in extreme poverty areas. Factors associated with stunting were different in rural and urban communities.
Therefore, developing and implementing health programs to tackle malnutrition should take into account such differences that are consequence of the social, economic, and cultural contexts in which the family lives.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HR and AS conceived, designed and co-ordinated the study. RPC participated in the design of the study, statistical analysis and interpretation of data, and drafted the manuscript. RC co-ordinated the acquisition of data and participated in the statistical analysis. SD collaborated in the data analysis and drafted the manuscript. JIS and GG participated in the conception and design of the study, interpretation of data and critical revision for important intellectual content. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank Drs. Carlos de la Peña, Director, and Cesar Piña, Chief Medical Advisor, of the Ministry of Health in the State of Guerrero, Mexico, for their permission to carry out the study and for support of the fieldwork. Financial support for this research was provided by the Consejo Nacional de Ciencia y Tecnología (CONACYT), Mexico, Grant # 29193-M.
==== Refs
Comisión Nacional en Favor de la Infancia. Programa Nacional de Acción 1995–2000: Evaluación 1998 México 1999
Delpeuch F Traissac P Martin-Prevel Y Massamba JP Maire B Economic crisis and malnutrition: socioeconomic determinants of anthropometric status of preschool children and their mothers in an African urban area Public Health Nutr 2000 3 39 47 10786722
Garret JL Ruel MT Are determinants of rural and urban food security and nutritional status different? Some insights from Mozambique World Development 1999 27 1955 1975 10.1016/S0305-750X(99)00091-1
Li Y Guo G Shi A Li Y Anme T Ushijima H Prevalence and correlates of malnutrition among children in rural minority areas of China Pediatr Int 1999 41 549 556 10530071 10.1046/j.1442-200X.1999.01103.x
Kikafunda JK Walker F Collette D Tumwine JK Risk factors for early childhood malnutrition in Uganda Pediatrics 1998 102 E451 E458 10.1542/peds.102.4.e45
Chang Y Xian ZD He W Chang SY Ma HJ Chen CM Child malnutrition in China – present status and changing trend Biomed Environ Sci 1996 9 164 80 8886327
Ray SK Biswas AB Kumar S A study of dietary pattern, household food security and nutritional profile of under-five children of a community of West Bengal J Indian Med Assoc 2000 98 517 519 11291783
Ballard TJ Neumann CG The effects of malnutrition, parental literacy and household crowding on acute lower respiratory infections in young Kenyan children J Trop Pediatr 1995 41 8 13 7723139
Fawzi WW Herrera MG Spiegelman DL el Amin A Nestel P Mohamed KA A prospective study of malnutrition in relation to child mortality in the Sudan Am J Clin Nutr 1997 65 1062 1069 9094894
Schelp FP Nutrition and infection in tropical countries – implications for public health intervention – a personal perspective Nutrition 1998 14 217 222 9530650 10.1016/S0899-9007(97)00436-X
Skull SA Ruben AR Walker AC Malnutrition and microcephaly in Australian aboriginal children Med J Aust 1997 166 412 414 9140346
Tupasi TE Mangubat NV Sunico ME Magdangal DM Navarro EE Leonor ZA Lupisan S Medalla F Lucero MG Malnutrition and acute respiratory tract infections in Filipino children Rev Infect Dis 1990 12 S1047 1054 2270404
Vazir S Naidu AN Vidyasagar P Nutritional status, psychosocial development and the home environment of Indian rural children Indian Pediatr 1998 35 959 966 10216718
Ricci JA Becker S Risk factors for wasting and stunting among children in Metro Cebu, Philippines Am J Clin Nutr 1996 63 966 975 8644694
Ruben AR Walker AC Malnutrition among rural aboriginal children in the Top End of the Northern Territory Med J Aust 1995 162 400 403 7746170
Rivera JR González-Cossío T Flores M Hernández M Lezana MA Sepúlveda-Amor J Stunting and emaciation in children under 5 in distinct regions and strata in Mexico Salud Publica Mex 1995 37 95 107 7618120
Secretaría de Salud, Instituto Nacional de Salud Pública. Instituto Nacional de Estadística, Geografía e Informática: Encuesta Nacional de Nutrición 1999 México 2000
Secretaría de Desarrollo Social: Programa de Educación, Salud y alimentación México; 2000
Sepúlveda-Amor J Lezana MA Tapia-Conyer R Valdespino JL Madrigal H Kumate J Nutritional status of pre-school children and women in Mexico: results of a probabilistic National Survey Gac Med Mex 1990 126 207 224 2094622
Bouvier P Papart JP Wanner P Picquet M Rougemont A Malnutrition of children in Sukasso (Mali): prevalence and socio-economic determinants Soz Praventivmed 1995 40 27 34 7900432
Brown K Solomons N Nutritional problems of developing countries Infect Dis Clin North Am 1991 5 297 317 1869811
Lipsky S Stephenson PA Koepsell TD Gloyd SS López JL Bain CE Breastfeeding and weaning in rural Mexico Nutr Health 1994 9 255 263 8065664
Rikimaru T Yartey JE Taniguchi K Kennedy DO Nkrumah FK Risk factors for the prevalence of malnutrition among urban children in Ghana J Nutr Sci Vitaminol 1998 44 391 407 9742460
Frongillo EA de Onis M JrHanson KM Socioeconomic and demographic factors are associated with worldwide patterns of stunting and wasting of children J Nutr 1997 127 2302 2309 9405578
Infante-Castañeda C Bases for the study of the interaction of family, social networks and health service utilization Salud Publica Mex 1988 30 175 196 3045992
Waterlow JC Buzina R Keller W Lane JM Nichaman MZ Tanner JM The presentation and use of height and weight data for comparing the nutritional status of groups of children under the age of 10 years Bull World Health Organ 1977 55 489 498 304391
National Center for Health Statistics NCHS Growth Curves for Children 0–18 Years 1977 United StatesWashington
Schlesselman J Case-control studies Design, conduct, analysis 1982 New York: Oxford University Press
Bordon LG Fernandez AE Stunting risk factors in 5-year-old children Rev Fac Cien Med Univ Nac Cordoba 1996 54 33 37 9336061
Bentley M Pelto G The household production of nutrition Soc Sci Med 1991 33 1101 1102 1767279 10.1016/0277-9536(91)90225-2
Berman P Kendall C Bhattacharyya K The household production of health: integrating social science perspective on micro-level health determinants Soc Sci Med 1994 38 205 215 8140447 10.1016/0277-9536(94)90390-5
Kigutha HN Van Starveren WA Veerman W Hautvast J Effects of seasonality on household food availability among smallholder rural households in Kenia: impact of land size, household size, and composition Ecol Food Nutr 1998 37 73 100
Cedraz LNC Carvallho FM Family income and malnutrition in the Coast of Camacari, Brazil Arch Latinoam Nutr 1990 40 323 332 2134136
Jeyaseelan L Lakshman M Risk factors for malnutrition in south Indian children J Biosoc Sci 1997 29 93 100 9881122 10.1017/S002193209700093X
Agueh VD Makoutode M Diallo P Soton A Ouendo EM Infant malnutrition and associated maternal factors in a secondary city south of Benin, Ouidah Rev Epidemiol Sante Publique 1999 47 219 228 10422116
Reyes H Tomé P Gutiérrez G Rodríguez L Orozco M Guiscafré H Mortality due to diarrhea in Mexico: a problem of accessibility or quality of care? Salud Publica Mex 1996 40 316 323 9774900
Simondon KB Costes A Delaunay V Diallo A Simondon F Children's height, health and appetite influence mothers' weaning decisions in rural Senegal Int J Epidemiol 2001 30 476 481 11416068 10.1093/ije/30.3.476
Getaneh T Assefa A Tadesse Z Protein-energy malnutrition in urban children, prevalence and determinants Ethiop Med J 1998 36 153 166 10214456
Caulfield LE Bentley ME Ahmed S Is prolonged breast feeding associated with malnutrition? Evidence from nineteen demographic and health surveys Int J Epidemiol 1996 25 693 703 8921445
Simondon KB Simondon F Mothers prolong breastfeeding of undernourished children in rural Senegal Int J Epidemiol 1998 27 490 494 9698141 10.1093/ije/27.3.490
| 15571622 | PMC539253 | CC BY | 2021-01-04 16:28:47 | no | BMC Public Health. 2004 Nov 30; 4:57 | utf-8 | BMC Public Health | 2,004 | 10.1186/1471-2458-4-57 | oa_comm |
==== Front
BMC Pulm MedBMC Pulmonary Medicine1471-2466BioMed Central London 1471-2466-4-121556657410.1186/1471-2466-4-12Research ArticlePredictors of mortality of patients with acute respiratory failure secondary to chronic obstructive pulmonary disease admitted to an intensive care unit: A one year study Khilnani GC [email protected] Amit [email protected] SK [email protected] Department of Medicine, All India Institute of Medical Sciences, New Delhi-110029, India2004 27 11 2004 4 12 12 18 3 2004 27 11 2004 Copyright © 2004 Khilnani et al; licensee BioMed Central Ltd.2004Khilnani 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
Patients with acute exacerbation of chronic obstructive pulmonary disease (COPD) commonly require hospitalization and admission to intensive care unit (ICU). It is useful to identify patients at the time of admission who are likely to have poor outcome. This study was carried out to define the predictors of mortality in patients with acute exacerbation of COPD and to device a scoring system using the baseline physiological variables for prognosticating these patients.
Methods
Eighty-two patients with acute respiratory failure secondary to COPD admitted to medical ICU over a one-year period were included. Clinical and demographic profile at the time of admission to ICU including APACHE II score and Glasgow coma scale were recorded at the time of admission to ICU. In addition, acid base disorders, renal functions, liver functions and serum albumin, were recorded at the time of presentation. Primary outcome measure was hospital mortality.
Results
Invasive ventilation was required in 69 patients (84.1%). Fifty-two patients survived to hospital discharge (63.4%). APACHE II score at the time of admission to ICU {odds ratio (95 % CI): 1.32 (1.138–1.532); p < 0.001} and serum albumin (done within 24 hours of admission) {odds ratio (95 % CI): 0.114 (0.03-0.432); p = 0.001}. An equation, constructed using the adjusted odds ratio for the two parameters, had an area under the ROC curve of 91.3%. For the choice of cut-off, sensitivity, specificity, positive and negative predictive value for predicting outcome was 90%, 86.5%, 79.4% and 93.7%.
Conclusion
APACHE II score at admission and SA levels with in 24 hrs after admission are independent predictors of mortality for patients with COPD admitted to ICU. The equation derived from these two parameters is useful for predicting outcome of these patients.
==== Body
Background
Chronic obstructive pulmonary disease (COPD) is characterized by irreversible airway obstruction that leads to chronic disability. Patients with COPD have a longstanding downhill course that is interspersed with episodes of exacerbations requiring hospitalization. COPD is known to be a common disease. There is lack of recent data regarding the burden of this disease from India, with only study on prevalence of COPD published in 1981 [1]. Data from United States indicate that incidence of disease is on the rise [2]. During the year 2000, approximately 24 million adults in United States had evidence of obstructive airway disease. COPD was responsible for 1.5 million emergency department visits, 726,000 hospitalizations, and 119,000 deaths [2]. It is obvious that this disease puts an enormous economic burden on the society. Andersson and coworkers estimated that almost 35-45% of the total per capita health-care costs for COPD are account for by exacerbations alone [3]. Severe exacerbations requiring hospitalizations are responsible for a large share of these costs and among these, treatment cost for those who require intensive care unit (ICU) admission is highest. In most of the third world countries, large number of ICU beds are occupied by patients with critical illnesses secondary to various infectious diseases, most of which are reversible.
It is important to identify patients at the time of admission who are likely to have poor outcome, so that such patients can be managed aggressively. Many prognostic scoring systems have been devised for the same purpose. These scoring systems help to segregate patients who are the sickest and are likely to die from those who are expected to have better outcome and survive. Most of these scoring systems have been devised for a broad range of critically ill patients. The present study was planned to determine the predictors of mortality in patients with exacerbation of COPD admitted to ICU over a one-year period. An attempt was made to develop a scoring system using the predictors of mortality that would help to identify patients at high risk of dying.
Methods
Prospectively collected data of patients with acute respiratory failure secondary to COPD admitted to medical ICU of All India Institute of Medical sciences, New Delhi, India (a tertiary care center in north India) over a one-year period (January 2002 to December 2002) was reviewed. Diagnosis of COPD was based upon the characteristic findings on history and examination with typical radiographic abnormalities [4]. Patients admitted to the ICU with COPD but due to any other primary reason such as those with poisoning or acute coronary event were excluded. Similarly, patients in whom the primary cause of respiratory failure was bronchiectasis, bronchial asthma, pulmonary edema or pulmonary embolism were not included. Finally, 82 patients with a primary admission diagnosis of acute respiratory failure secondary to COPD were included. All patients were documented cases with prior pulmonary function test confirmation of irreversible airway obstruction and had been receiving a combination of various bronchodilators.
Management of the patients was the primary responsibility of the ICU team. A treatment strategy was individualized for each patient and was the sole prerogative of the treating physician. All patients received regular nebulized bronchodilators including salbutamol (as frequently as 5 mg every 15 minutes to every 8 hours), ipratropium bromide (as frequently as 0.5 mg every 15 minutes to 0.25 mg every 8 hours), and intravenous corticosteroids. Most patients also received antibiotics (n = 75, 91.5%). Oxygen therapy (2-3 lt/min) was administered to spontaneously breathing patients. The decision to institute ventilatory support was taken by the treating physician. Wherever feasible non-invasive ventilation (NIV) was used as the initial strategy. Endotracheal intubation was done for usual indications such as respiratory arrest, deteriorating level of consciousness, rising PaCO2 despite maximal pharmacological treatment and deteriorating acidemia. Initiation of weaning from mechanical ventilation was considered as soon as the patients were considered capable of breathing spontaneously. Method of weaning trials included t-piece trials, gradual reduction of synchronized intermittent mandatory ventilation (SIMV) breaths and pressure support ventilation (PSV).
Clinical and demographic profile at the time of admission to ICU including age, sex, smoking status, history of previous hospital admissions, history of previous intubation and/or ventilatory support, prior evidence of cor pulmonale with or without congestive heart failure were recorded. Findings on clinical examination including heart rate, respiratory rate and mean blood pressure were recorded. Acute physiology and chronic health evaluation II (APACHE II) score and Glasgow coma scale (GCS) were recorded at the time of admission to the ICU. Acid-base abnormalities at the time of presentation were analyzed by recording the arterial blood gas analysis and serum electrolytes (estimations done on AVL 995S). Renal functions, liver functions and serum albumin (SA) done at the time of admission were also recorded. Development of complications during mechanical ventilator such as pneumothorax and ventilator associated pneumonia (VAP) were recorded. Development of acute respiratory distress syndrome (ARDS), sepsis and multi-organ failure was also documented. ARDS was defined as presence of bilateral pulmonary infiltrates on chest radiograph in presence of hypoxemia with PaO2 / FiO2 ratio less than 200 without any evidence of left atrial hypertension (American-European Consensus Conference) [5]. Sepsis was defined as the presence of a clinically identified site of infection (eg, pneumonia) and two or more of the following: temperature > 38°C or < 36°C; heart rate > 90 beats/min; respiratory rate > 20 breaths/min or PaCO2 < 32 mm Hg; and WBC count > 12 × 109/L, < 4.0 × 109/L, or > 0.10 immature forms (ie, bands) (American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference) [6]. Days on ventilator, days of ICU stay and days of hospital stay were recorded for all the patients. Primary outcome measure was hospital mortality.
Statistical analysis
Data were double entered to minimize errors and managed on an 'Excel' master sheet. Analysis was done using the statistical software 'SPSS version 10.0' (SPPS Corp, Chicago, IL, USA). Descriptive analysis consisted of mean with standard deviation and range for various parameters. Study group was split on the basis of final outcome. Various parameters were compared between the two groups to identify the predictors of mortality. Continuous variables were analyzed using student's t-test whereas Fisher's exact test was used to compare the ordinal variables. Baseline parameters significant on univariate analysis at p < 0.1 were identified as potential predictor variables. These parameters were evaluated using multivariate logistic regression analysis (backward stepwise method) to determine independent predictors of mortality. An equation was constructed using the independent predictors based on the adjusted odds ratios and a diagnostic rule was defined. To evaluate the predictive capability of the variables and the equation, receiver-operator characteristic (ROC) curves were constructed with sensitivity (on X-axis) and 1-specificity (on Y-axis) for various cut-offs. Significance was considered at p < 0.05 (only two tailed) for the present study.
Results
Baseline characteristics
Demographic and baseline clinical and laboratory profile of the study group are presented in Table 1. Almost all patients had type II respiratory failure (n = 74, 90.2%) and showed acute on chronic respiratory acidosis. Study cohort mostly consisted of critically ill patients as suggested by a high mean APACHE II score. History of smoking could be elicited in 65 patients (79.3%). A significant number of patients had history of previous hospitalization as well as intubation (39% and 18.3% respectively). Almost 55% of the patients (n = 45) had evidence of underlying cor pulmonale. Fifteen patients (18.3%) had underlying diabetes mellitus whereas 12 patients (14.6%) were on treatment for hypertension. None of the patients suffered from any other co-morbid condition. An attempt was made to define the cause of exacerbation for all patients. There was evidence of pneumonia in 67% (n = 55) of patients whereas pneumothorax was responsible for decompensation in 3 patients (3.7%). No obvious cause could be found in 24 patients (29%). Only one patient had evidence of sepsis, but none had ARDS at the time of admission to the ICU.
Table 1 Descriptive profile of the study group (n = 82)
Minimum Maximum Mean ± Std. Deviation
Age (years) 35 85 60 ± 10
APACHE II score 3 33 13 ± 6
PR (per minute) 46 166 105 ± 19
RR (per minute) 10 46 27 ± 10
MBP (mmHg) 20 126 89 ± 19
GCS 3 15 12.1 ± 3
pH 6.87 7.44 7.25 ± 0.19
PaCO2 (mmHg) 40.7 130.7 76.6 ± 23.5
PO2 (mmHg) 31.5 142.3 83.9 ± 41.7
HCO3 (mmHg) 5.4 55.4 32.3 ± 8.7
Serum Na (mEq/L) 115 152 136 ± 7
Serum K (mEq/L) 2.00 6.80 4.2 ± 0.9
Serum Albumin (gm%) 1.7 4.4 3.2 ± 0.7
Days on ventilator 1 33 8.7 ± 4.6
Days of ICU stay 1 35 9.6 ± 6.2
Days of hospital stay 1 63 16.3 ± 10.4
RR: Respiratory rate, PR: Pulse rate, MBP: Mean blood pressure, GCS: Glasgow coma scale, Serum Na: Serum sodium, Serum K: Serum potassium.
Hospital course
Non Invasive Ventilation (NIV) was used as initial strategy in 17 patients (20.7%). This strategy had a success rate of 59% (n = 10). Sixty-nine patients (84.1%) received invasive ventilation (including seven patients who failed NIV and had to be intubated). Sepsis developed in 11 patients (13.4%) and all these patients eventually died. Parameters associated with development of sepsis were high APACHE II score (18 vs. 12, p = 0.005) and low SA (2.6 gm/dL% vs. 3.3 gm/dL, p < 0.001). VAP developed in 6 patients (8.7%) and was associated with an increased stay in the ICU (18 days vs. 10 days, p = 0.021) as well as increased stay in the hospital (30 days vs. 15 days, p = 0.005). Outcome was not significantly affected by development of VAP (50% versus 42.8%).
Outcome
Hospital mortality was 36.6% (n = 30). Various parameters were compared for survivors and non-survivors (table 2). In addition to demographic characteristics (age and sex), presence of cor pulmonale and cause of exacerbation of COPD, baseline parameters significantly different between the two groups on univariate analysis were included in a multivariate equation. APACHE II score at admission to the ICU {odds ratio (95 % CI): 1.32 (1.138-1.532); p < 0.001} and SA (done within 24 hours of admission) {odds ratio (95 % CI): 0.114 (0.03-0.432); p = 0.001} emerged as the independent predictors of mortality. ROC curve showed that both these variables have good predictive capability with area under the ROC curve (AUC) of 86.9% for APACHE II score (Figure 1) and 82.2% for SA (Figure 2). Best cut-off, taken as the value on the ROC curve at the point where curve sharply angulated, was 13.5 for APACHE II score and that for SA was 3.05 gm/dL. Following equation was determined by combining the two variables using the adjusted odd ratio: Score = (0.278 × APACHE II score) - (2.17 × SA), where APACHE II score is the score at the time of admission and SA (gm/dL) is the level with in the first 24 hours. ROC curve for this equation showed an AUC value of 91.2% (Figure 3). We chose a cut-off of -2.97 for the equation. That is, a patient with a score above -2.97 is likely to die whereas the one with below -2.97 likely to survive. This diagnostic rule had a specificity of 86.5% with a sensitivity of 90%. Positive predictive value for this variable was 79.4% whereas negative predictive value was 93.7%. A cut-off of -0.45 was 100% specific for hospital mortality but sensitivity was only 40%. On the other hand a cut-off of -5.5 gave a sensitivity of 100% with specificity of 33%.
Table 2 Predictors of mortality for patients with exacerbation of COPD
Parameter Survivors (n = 52) Non-survivors (n = 30) p value
Mean ± SD Mean ± SD
APACHE II score 10.6 ± 4.3 17.5 ± 5.7 0.001
GCS 12.8 ± 2.1 10.8 ± 3.7 0.003
MBP (mmHg) 93 ± 13.6 82.5 ± 24.7 0.015
PR (per minute) 110.7 ± 16.9 102.2 ± 21.1 0.049
Serum Albumin (gm%) 3.5 ± 0.5 2.7 ± 0.6 0.001
PaCO2 (mmHg) 81.2 ± 20.8 68.7 ± 25.8 0.018
HCO3 (mmol/L) 33.8 ± 8.1 29.6 ± 9.1 0.035
Need of reintubation 35.3% 4.4% 0.001
Renal Failure Nil 16.7% 0.002
Sepsis Nil 36.7% <0.001
GCS: Glasgow coma scale, MBP: Mean blood pressure, PR: Pulse rate.
Figure 1 Receiver operator characteristic (ROC) curve plotted for studying the diagnostic utility of Serum Albumin in predicting outcome of patients. The choice of cut-off is shown by an arrow (3.05 g/dL).
Figure 2 Receiver operator characteristic (ROC) curve plotted for studying the diagnostic utility of APACHE II score in predicting outcome of patients. The choice of cut-off is shown by an arrow (13.5).
Figure 3 Receiver operator characteristic (ROC) curve plotted for studying the diagnostic utility of score derived form equation in predicting outcome of patients. The choice of cut-off is shown by an arrow (-2.97).
Discussion
Primary outcome measure of the present study was hospital mortality. Overall mortality rate was 36.6%. There was a high incidence of need of MV (84.1%). In studies that have taken into account all the patients with COPD requiring hospitalization, mortality rate has been to the tune of 6-42% [7-10]. Weiss & Hudson [11] reviewed 11 studies carried out to study outcome of patients with exacerbation of COPD and found the combined mortality rate to be 20.3%. Selection bias in the inclusion of patients for the present study precludes the generalization of these figures for patients with exacerbation of COPD requiring hospitalization from India. Only a fraction of all the patients with exacerbation of COPD admitted to our hospital are managed in ICU. Many other patients with acute exacerbation of COPD, especially those who do not require ventilatory support, are managed in the wards only. Because of this fact, by including patients who were admitted to ICU the sickest group of patient with exacerbation of COPD was selected.
Various physiological parameters estimated at the time of presentation were analyzed to find predictors of mortality. Only two parameters, namely APACHE II score at admission to ICU and SA in the first 24 hours of admission, were found to be independent predictors of hospital mortality. The same two parameters also predicted development of sepsis on bivariate analysis. Some of the earlier studies have found blood gas parameters like pH [12] and PaCO2 [13] to be useful in predicting outcome in COPD patients, whereas others [14-16] did not. In the present study, although PaCO2 and HCO3 were not independent predictors of mortality they tended to be lower in patients who died and the difference was statistically significant on bivariate analysis. Also, mean pH was similar for the two groups. This has not been reported in the earlier studies and investigators in the past have mostly found high PaCO2 levels to be associated with worse outcome. A possible reason for this finding is that patients with hypercapnia with concordantly high HCO3 are usually taken care of by mechanical ventilation. On the other hand, low mean PaCO2 and HCO3 levels in non-survivors probably reflected underlying metabolic acidosis. It has been reported earlier also that, for similar level of acidosis, patients with respiratory failure resulting in respiratory acidosis have better outcome as compared to patients with metabolic acidosis, that is commonly secondary to associated non-pulmonary organ failure [17]. Mean pH was similar for both survivors and non-survivors but survivors comprised predominantly of patients with respiratory acidosis (higher PaCO2 as well as HCO3) whereas non-survivors consisted of patients with metabolic acidosis (lower PaCO2 and HCO3 but similar pH). Another finding that corroborates the same fact is that all patients, who had associated renal failure and/or sepsis, died. The incidence of these two complications was significantly higher in non-survivors (renal failure 16.7% vs nil, p = 0.002; sepsis 36.7% vs nil, p < 0.001). Patients with both these complications commonly have associated metabolic acidosis.
Prognostic utility of APACHE II score has been extensively investigated. It has been found useful for prognosticating critically ill patients across a wide array of diagnostic categories. Earlier studies have also found APACHE II score to be useful in predicting mortality in COPD patients with acute exacerbation [18-21] although the timing of scoring after admission has varied in different studies. For example Nevins & Epstein [18] found APACHE II score at 6 hrs after initiation of ventilation to be a useful predictor of mortality. In the present study, APACHE II scoring done at the time of admission to medical ICU was analyzed. SA estimated with in first 24 hrs of admission was also found to be a strong predictor of mortality. SA has also been reported to be of good prognostic value in the past [21-23]. Utility of prognostic value of SA in patients with COPD is interesting. Albumin has a long half-life of approximately 18 days and because of this fact it is unlikely to change with development of acute respiratory failure in patients with COPD. On the other hand SA is known to reflect the underlying nutritional status and to be affected by the severity of chronic illness. These factors are of obvious significance in deciding the outcome of these patients.
An important purpose of the present study was to define predictors, which could help to identify patients that are likely to have worse outcome. This would help us to segregate patients who need to be managed aggressively from the very beginning. We looked at individual predictive utility of the parameters (SA and APACHE II score) that were found to be independent predictors of mortality. Both these parameters had good predictive value as evidenced by high AUC values. To improve the predictive utility, an equation was constructed using the adjusted odds ratio of the two parameters. ROC curve for this equation had a superior AUC value of 0.912. A good prognostic marker needs to be highly specific so that false positives remain low. On the other hand, good sensitivity is also desirable so that false negatives are not too high. A cutoff value of -2.97 has been suggested, which is associated with good specificity (86.5%) as well as sensitivity (90%) for predicting mortality of these patients. Cut-off that were associated with 100% specificity and sensitivity were also determined as different ICU's across the globe may have different priorities at different times.
Although prospective studies are required to validate the findings of present study, an equation devised by combination of APACHE II score and SA appears to make sense. Estimation of APACHE II score makes use of various physiological variables but does not include SA levels. Also, the chronic physiology score in APACHE II fails to stratify patients according to varying severity of chronic illnesses. This tends to happen in patients with COPD as well. Use of SA, which predominantly reflects the severity of chronic illness, in the equation seems to complement the predictive capability of APACHE II score. The results of the present study reflect the complex interplay of factors that occurs in patients with exacerbation of COPD. In these patients, an acute insult in the form of exacerbating illness develops on top of a chronic smoldering illness. Severity of both acute insult as well as the underlying disease in the background of the level of nutritional status tends to determine the outcome of these patients.
Although the equation is useful in to identifying patients with exacerbation of COPD who are likely to have poor outcome, it cannot be looked at in isolation. Other particulars of these patients such as associated illnesses and co-morbidities must be kept in mind before taking a final decision. It cannot be overemphasized that given the sensitivity and specificity of the equation, certain patients with a score below the suggested cut-off may also be sick. Also, the state of patients with exacerbation of COPD tends to remain in a constant flux and need constant monitoring. In spite of having a low score at presentation many of these patients may deteriorate during hospital stay.
It is concluded that APACHE II score at admission and SA levels with in first 24 hrs after admission are independent predictors of mortality for patients with exacerbation of COPD. The equation derived by combining these two parameters is useful for identifying patients that are likely to have poor outcome.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GCK: concept and design of study, management of patients, preparation of the manuscript. AB: concept of the study, management of patients, statistical analysis, preparation of the manuscript. SKS: management of patients and critical review of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Kinare SG Dave KM Sheth RA Incidence of emphysema in Bombay Indian J Med Res 1981 74 273 282 7309187
Mannino DM Homa DM Akinbami LJ Ford ES Redd SC Chronic obstructive pulmonary disease surveillance – United States, 1971-2000 Respir Care 2002 47 1184 1199 12354338
Andersson F Borg S Jansson SA Jonsson AC Ericsson A Prutz C Ronmark E Lundback B The costs of exacerbations in chronic obstructive pulmonary disease (COPD) Respir Med 2002 96 700 708 12243316 10.1053/rmed.2002.1334
American Thoracic Society Standards for the diagnosis and care of patients with chronic obstructive pulmonary disease and asthma Am Rev Respir Dis 1987 136 225 244 3605835
Bernard GR Artigas A Brigham KL Carlet J Falke K Hudson L Lamy M Legall JR Morris A Spragg R The American-European Consensus Conference on ARDS Am J Respir Crit Care Med 1994 149 818 824 7509706
Bone RC Balk RA Cerra FB Dellinger RP Fein AM Knaus WA Schein RM Sibbald WJ ACCP/SCCM Consensus Conference Chest 1992 101 1644 1655 1303622
Costello R Deegan P Fitzpatrick M McNicholas WT Reversible hypercapnoea in chronic obstructive pulmonary disease: a distinct pattern of respiratory failure with a favourable prognosis Am J Med 1997 102 239 244 9217591 10.1016/S0002-9343(97)00017-X
Burk RH George RB Acute respiratory failure in chronic obstructive pulmonary disease. Immediate and long-term prognosis Arch Intern Med 1973 132 865 868 4757258 10.1001/archinte.132.6.865
Asmundsson T Kilburn KH Survival after acute respiratory failure. 145 patients observed for 5 to 8 and one half years Ann Intern Med 1974 80 54 57 4810350
Martin TR Lewis SW Albert RK The prognosis of patients with chronic obstructive pulmonary disease after hospitalization for acute respiratory failure Chest 1982 82 310 314 7105857
Weiss SM Hudson LD Outcome from respiratory failure Crit Care Clin 1994 10 197 215 8118729
Warren PM Flenley DC Millar JS Avery A Respiratory failure revisited: acute exacerbations of chronic bronchitis between 1961-68 and 1970-76 Lancet 1980 1 467 470 6102193 10.1016/S0140-6736(80)91008-9
Breen D Churches T Hawker F Torzillo PJ Acute respiratory failure secondary to chronic obstructive pulmonary disease treated in the intensive care unit: a long term follow up study Thorax 2002 57 29 33 11809986 10.1136/thorax.57.1.29
Menzies R Gibbons W Goldberg P Determinants of weaning and survival among patients with COPD who require mechanical ventilation for acute respiratory failure Chest 1989 95 398 405 2914493
Portier F Defouilloy C Muir JF and the French task force for acute respiratory failure in chronic respiratory insufficiency Determinants of immediate survival among chronic respiratory insufficiency patients admitted to an intensive care unit for acute respiratory failure. A prospective multicenter study Chest 1992 101 204 210 1729072
Kaelin RM Assimacopoulos A Chevrolet JC Failure to predict 6-month survival of patients with COPD requiring mechanical ventilation by analysis of simple indices: a prospective study Chest 1987 92 971 978 3677842
Esteban A Anzueto A Frutos F Alia I Brochard L Stewart TE Benito S Epstein SK Apezteguia C Nightingale P Arroliga AC Tobin MJ Mechanical Ventilation International Study Group Characteristics and outcomes in adult patients receiving mechanical ventilation: a 28-day international study JAMA 2002 287 345 355 11790214 10.1001/jama.287.3.345
Nevins ML Epstein SK Predictors of outcome for patients with COPD requiring invasive mechanical ventilation Chest 2001 119 1840 1849 11399713 10.1378/chest.119.6.1840
Vitacca M Clini E Porta R Foglio K Ambrosino N Acute exacerbations in patients with COPD: predictors of need for mechanical ventilation Eur Respir J 1996 9 1487 93 8836664 10.1183/09031936.96.09071487
Knaus WA Prognosis with mechanical ventilation: the influence of disease, severity of disease, age, and chronic health status on survival from an acute illness Am Rev Respir Dis 1989 140 S8 S13 2669589
Connors AF JrDawson NV Thomas C Harrell FE JrDesbiens N Fulkerson WJ Kussin P Bellamy P Goldman L Knaus WA Outcomes after acute exacerbation of severe chronic obstructive lung disease: the SUPPORT investigators Am J Respir Crit Care Med 1996 154 959 967 8887592
Anon JM Garcia de Lorenzo A Zarazaga A Gomez-Tello V Garrido G Mechanical ventilation of patients on long-term oxygen therapy with acute exacerbations of chronic obstructive pulmonary disease: prognosis and cost-utility analysis Intensive Care Med 1999 25 452 457 10401937 10.1007/s001340050879
Menzies R Gibbons W Goldberg P Determinants of weaning and survival among patients with COPD who require mechanical ventilation for acute respiratory failure Chest 1989 95 398 405 2914493
| 15566574 | PMC539254 | CC BY | 2021-01-04 16:30:10 | no | BMC Pulm Med. 2004 Nov 27; 4:12 | utf-8 | BMC Pulm Med | 2,004 | 10.1186/1471-2466-4-12 | oa_comm |
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-4-291557595210.1186/1472-6750-4-29Research ArticleOver expression of the selectable marker blasticidin S deaminase gene is toxic to human keratinocytes and murine BALB/MK cells Bento Fernanda Mara [email protected] Daniela [email protected] Chester Bittencourt [email protected] Tamara Rocha [email protected] Monica Beatriz [email protected] Adriana Karaoglanovic [email protected] Sang Won [email protected] Department of Biophysics & Center for Gene Therapy, UNIFESP – EPM, São Paulo, SP, Brazil2 Instituto de Pesquisas Energéticas e Nucleares – CNEN, São Paulo, SP, Brazil2004 2 12 2004 4 29 29 30 3 2004 2 12 2004 Copyright © 2004 Bento 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 blasticidin S resistance gene (bsr) is a selectable marker used for gene transfer experiments. The bsr gene encodes for blasticidin S (BS) deaminase, which has a specific activity upon BS. Therefore, its expression is supposed to be harmless in cells. The work reported on herein consisted of experiments to verify a possible toxicity of bsr on mammalian cells, which include several cell lines and primary cultures.
Results
Murine keratinocyte BALB/MK and human primary keratinocyte cells transduced with the retroviral vector LBmSN, which has an improved expression system of bsr, namely bsrm, died in five days after the transduction. Meanwhile the control vector LBSN, which expresses bsr, did not provoke cell death. The lethal activity of bsrm was observed only in human keratinocytes and BALB/MK cells among the cell types tested here. Death appears to be mediated by a factor, which is secreted by the BALB/MK transduced cells.
Conclusion
By our study we demonstrated that the expression of bsrm gene is toxic to human keratinocytes and BALB/MK cells. It is likely over expression of BS deaminase gene is responsible for the death.
==== Body
Background
Blasticidin S (BS) is a nucleoside antibiotic isolated from Streptomyces griseochromogenes, and has been used as a fungicide against rice blast disease [1]. BS inhibits protein synthesis in both prokaryotes and eukaryotes [1]. Later, a gene that provides resistance against BS (bsr) was isolated from Bacillus cereus K55-S1 strain [2,3]. The bsr gene possesses only 420 bp [4] and codes for an enzyme of 15 kDa, which is usually present in its dimer form [5]. The enzyme acts upon BS converting it into a deaminohydroxy derivative [5,6], thus it is named BS deaminase.
BS antibiotic is highly toxic to mammalian cells; even 2 μg/ml is enough to kill HeLa cells in a few days. However, the cells transfected with bsr could resist against BS at several fold higher concentrations [7]. Hence, the BS/bsr pair has been used as an efficient selection system for gene transfer experiments in many different cell types.
Recently, we reported that some modifications introduced into the non-coding regions of the bsr gene (bsrm) resulted in an increase (several fold) of bsr gene expression, and consequently, NIH3T3 cells transduced by retroviral vectors could be selected with higher concentrations of BS in just a few days [8]. Even with such extensive use of the bsr/BS selection system, no side effects in response to bsr gene expression have been observed.
Using the murine keratinocyte cell line BALB/MK [9] and human primary keratinocytes, we report here a surprising death of the keratinocytes provoked by the expression of bsrm. A detailed investigation about the death of keratinocytes, which was mediated by an unknown molecule and secreted by the BALB/MK cells transduced with LBmSN, is discussed below.
Results
Sensitivity of BALB/MK cells to BS
To determine the range of BS concentrations and the time required for BALB/MK cell death to occur, 1 × 104 cells/well were incubated in a 24 well plate and 2 days later the media were replaced with a fresh one containing various concentrations of BS for 9 days. A BS concentration of at least 2 μg/ml was necessary to kill the cells in 5 days. At concentrations of BS greater than 8 μg/ml the majority of cells died within 3 days.
Effect of bsrm gene and BS on BALB/MK cells
To verify the effect of bsrm gene in BALB/MK cells, the cells were transduced with the LBmSN retroviral vector, which expresses the bsrm and neoR genes, and then incubated with BS. The bsrm gene was obtained by modifying the bsr gene at a non-coding region and, consequently, there was no alteration of the amino acid sequence [8]. The addition of BS at concentrations of ≤ 4 μg/ml or 500 μg/ml of G418 to cell media resulted in the death of all transduced BALB/MK (Figure 1G,1H,1I,1J,1K,1L,1M,1N,1O). In contrast, the transduced cells incubated with BS at concentrations of ≥ 8 μg/ml reached confluence within 12 days of culture (Figs. 1P,1Q,1R). As an experimental control, BALB/MK cells transduced with the LXSN vector (which expresses only the neoR gene) [10] were incubated in the presence or absence of 500 μg/ml of G418. As expected, the cells reached partial or complete confluence within 12 days (Figs. 1D and 1E), as also observed with the non-transduced BALB/MK cells (Figure 1A).
BALB/MK cells not expressing the bsrm gene died in the presence of 8 μg/ml of BS (Figs. 1C and 1F) as expected. However, the aspect of the BALB/MK cells which died after the expression of bsrm (Figs. 1G,1H,1I,1J,1K,1L,1M,1N,1O) was different from the aspect of the non-transduced cells (Figs. 1B and 1C) or cells transduced with LXSN (Figure 1F) which were killed by antibiotics alone. BALB/MK cells which died after expressing the bsrm gene had a reduced cell volume when compared with normal BALB/MK or transduced BALB/MK incubated with 8 μg/ml BS (Figure 2). Additionally, the cells, which died after bsrm expression, remained on the plate after washing with PBS, whilst the cells killed by the antibiotics, BS or G418, were easily removed after washing with PBS (Figure 1). The death of BALB/MK cells was confirmed by staining with Trypane blue (Figure 1S).
Removal of BS from the medium of the cells selected with 8 μg/ml BS resulted in cell death within a week (not shown). Thus, the BS antibiotic counteracted the death effect of bsrm and therefore has a vital role to BALB/MK cells transduced with LBmSN.
Effects of cell density and virus concentration upon induction of BALB/MK cell death
Based on the above observations two variables were analyzed to assess their influences on the cell death: virus concentration and cell density. The total BALB/MK cells, transduced and non-transduced ones, were seeded at 1 × 103 to 4 × 104 cells in 25 cm2 flasks and the cell death was monitored by optical microscope. The flasks containing a higher number of cells had faster cell death, even if the ratio of virus per cell was maintained constant in all flasks (Table 1). However, the absolute number of bsrm-transduced BALB/MK cells was higher in the flasks with higher number of seeded cells; consequently cell death was directly related to the presence of the number of bsrm-transduced cells.
The virus concentration used to transduce BALB/MK cells was evaluated by infecting the cells with 1 × 102 to 1 × 105 cfu (colony forming units) of LBmSN vector. The transduced cells incubated with 8 μg/ml of BS produced resistant colonies proportionally to the used virus concentration (Figure 3). LBmSN-transduced BALB/MK cells that did not undergo selection died at all virus concentrations (Figs. 3G,3H,3I,3J), although the cells transduced with 1 × 105 cfu died 2 days earlier than the cells transduced with 1 × 102 to 1 × 103 cfu. This is an extremely important observation since even in those wells containing less virus than cells (Figs. 3G and 3H) cell death occurred simultaneously in each cell. This result suggests the existence of intercellular signaling of death. To confirm this hypothesis, we seeded the LBmSN transduced and non-transduced BALB/MK cells together with or without BS (Table 2). A clear induction of death in BALB/MK cells by the BALB/MK cells transduced with LBmSN was observed. Changes of cell morphology in each colony occurred within 5 days, as was seen in all experiments.
To investigate whether the death signaling is mediated by a secreted factor, we tested the supernatant of the BALB/MK cells transduced with LBmSN on BALB/MK cells (Figure 4). A just two-day old medium was sufficient to induce death of normal BALB/MK cells. This result indicates that cell death was mediated by a soluble factor, secreted by the LBmSN-transduced cells, acting on both transduced and non-transduced BALB/MK cells. This factor we denominated DOKEB (Death factor Obtained from Keratinocytes Expressing Bsrm) to ease our discussion. DOKEB appears to be secreted only by LBmSN-transduced BALB/MK cells, because the 5 day old-medium from the LBmSN-transduced NIH3T3 cells had no death activity upon BALB/MK or NIH3T3 cells (data not shown).
Effect of bsrm on mammalian cell lines
The lethal effect provoked by bsrm was firstly observed in the murine keratinocyte cell line BALB/MK as described above. To verify this lethal effect in other cell types, we chose the cells originated from the skin or epithelium (NIH3T3, HeLa, LISP-A10, LISP-E11, HCT-8 and B16F10), because of the origin of the BALB/MK cells [11]. In addition, the rat vascular smooth muscle cells, which are useful for gene therapy experiments [12], were also tested.
Until 7 days post-infection none of the above cells, which were transduced with LBmSN retroviral vector, did not die, whereas the control cell line BALB/MK died 5 days after the transduction (not shown). As the viral transduction rate is essential to analyze the possible death effect by the expression of bsrm, the cells were transduced with a ten-fold higher number of viruses than cells. Even in such conditions no cell types suffered with the expression of bsrm.
To ensure the transduction and expression of bsrm in the cells, those transduced cells were selected with 8 μg/ml of BS from the non-transduced ones that die in 4 days. The selected cells were distributed to two plates, and in one plate the initial concentration of BS was maintained and in from the other plate the BS was removed. During the 7 days of observation no death was observed in both plates (not shown), which confirm the previous result that the bsrm gene is not lethal to those cell types.
We also compared death activity of LBmSN and LBSN, which express bsrm and bsr respectively, on BALB/MK and NIH3T3 cells. Transduction of LBSN vector on BALB/MK or NIH3T3 cells did not cause cell death; meanwhile LBmSN caused cell death as expected (Table 3). In the presence of BS both cell lines transduced with LBmSN or LBSN did not die, which is a demonstration of BS deaminase gene expression, and also protection of the LBmSN transduced BLAB/MK cells against death as seen before. These results infer that over expression of BS deaminase gene could be responsible for the death of BALB/MK cells expressing bsrm.
Effect of the BS analogs on BALB/MK cells
The analogs of BS, cytidine, 5'-deoxycytidine, uridine and 5'-deoxyuridine, were tested in the culture of the BALB/MK cells transduced with LBmSN at 1 μM to 10 mM concentrations (Figure 4). Interestingly all analogs with 10 mM protected the transduced cells during 5 days of observation. Changing the medium with a fresh one containing 10 mM of each analog at every five days, the cells could be maintained alive for several passages (not shown).
Effect of the bsrm gene and BS upon human keratinocytes
The transduced human keratinocytes, which were modified with the virus producing cell clone PA317/LBmSN as a feeder-layer, did not grow during 8 days of culture in the absence of BS (Figure 6). Incubating the transduced keratinocytes with BS at 0.05 to 2 μg/ml, which are tolerant concentrations by the cells, also resulted in the absence of cell growth (not shown). However, the presence of 8 μg/ml of antibiotic, in the medium, which is a lethal concentration for the cells, resulted in the formation of many keratinocyte colonies. The number of these BS resistant colonies decreased as the BS concentration increased (Figure 6).
The BS selected cells could be maintained alive even with the expression of bsrm if the medium is replaced every two days with a fresh one containing the initial concentration of BS. Nevertheless, if those cells were seeded on a new plate without BS they died in few days (Figure 7). This result confirms the vital role of BS in bsrm expressing keratinocytes, as was observed with murine keratinocyte BALB/MK cells.
The vector LBmSN also expresses neoR, which provides resistance against geneticin. However, the transduced keratinocytes died after the incubation with 500 μg/ml of geneticin (Figure 6) even if the antibiotic was neutralized by aminoglycoside phosphotransferase. This result corroborates previous data that in keratinocytes the expression of bsrm gene leads to cell death. The keratinocytes transduced with the LXSN vector [10], which was used to construct the LBmSN vector, presented normal growth reaching complete confluence within 8 days (not shown).
Discussion
Here we report a surprising death caused by the transduction of the murine keratinocyte cell line BALB/MK cells with the retroviral vector LBmSN. This vector expresses bsrm which is a modified form of bsr only in the non-coding region; consequently both genes express the same BS deaminase. As the vector LBSN, which expresses bsr much lower than LBmSN [8], did not kill the cells an over expression of bsr could be responsible for the death. However, we can not discard other possibilities caused by interactions of the bsrm gene product (mRNA or protein) with intracellular molecules.
In some cases the Moloney murine retrovirus can cause cell death [13]. However in our case the control retroviral vector LXSN, which was used to construct LBmSN and LBSN does not carry any viral genes [8], did not induce cell death in the same culture conditions (Figure 1). Additionally the wide range of viral concentrations and cell densities tested here (Table 1 and Figure 3) did not affect cell death. These last results corroborate the above conclusion that the death of BALB/MK cells was caused by the expression of bsrm.
An interesting phenomenon of the death of BALB/MK cells expressing bsrm was that those cells can be rescued if a lethal concentration of BS (8 μg/ml to 32 μg/ml of BS were tested) is added in the medium before three days after transduction. It was a paradox, since the toxic antibiotic, BS, was able to rescue the murine keratinocyte BALB/MK induced to die by the apparently inoffensive bacterial bsrm gene.
BS-rescuing process of the bsrm-transduced keratinocytes could be understood as a consequence of inhibition of the death factor (DOKEB) production by BS. Because those transduced keratinocytes could be maintained for long period (at least two months) simply by changing the medium every two days for a fresh one containing 8 μg/ml of BS (not shown). However, we do not know if the inhibition of DOKEB production is caused by the inhibition of protein synthesis by BS or just occupation of the active site of the BS deaminase by BS, and consequently inhibiting the binding of the first target molecule which is a responsible for the induction of the death process. We believe more in the last explanation, because if the analogs of BS were added to the medium at concentrations higher than 1 mM, cell death can be avoided (Figure 4). The requirement of higher concentration of the analogs of BS than BS, which requires only 19 μM to protect the bsrm expressing BALB/MK cells, is likely due to the low affinity of analogs for BS deaminase as expected [5].
Even though we have demonstrated that the expression of the bsrm gene in BALB/MK is lethal, the bsrm/BS selection system can still be used in keratinocytes. During the selection of LBmSN-transduced keratinocytes, the initial concentration of BS in the medium should determine a strict range of LBmSN transduced keratinocytes, which express not more and not less than certain levels of bsrm gene (Figure 6). Thus, for survival of those transduced cells AND selected with 8 μg/ml BS, the medium should be changed every 2 days to maintain active BS concentration in the medium; otherwise, the low BS concentration will allow the synthesis of DOKEB and in turn trigger the death mechanism. The BALB/MK cells transduced with LBmSN and selected with 8 μg/ml of BS should not be challenged with concentrations much higher than those used, since the cells expressing higher levels of bsrm gene should have died during the previous selection. Thus, the selected BALB/MK cells exist in a precarious situation where either apoptosis or necrosis can be easily activated at any unfavorable moment.
We evaluated if the lethal effect provoked by bsrm occurs exclusively in BALB/MK cells or if this phenomenon is general for all cell types. In this study we included normal human primary keratinocytes and several cell lines. The human keratinocytes are resistant to BS at concentrations lower than 2 μg/ml and at concentrations higher than 8 μg/ml the cells die within 5 days (not shown). However, the keratinocytes transduced with LBmSN behaved in an opposite way. In the absence or presence of BS at low concentrations (lower than 2 μg/ml) the transduced cells died, as it was observed with the BALB/MK cells transduced with LBmSN. Additionally, the protection against the death of the BALB/MK cells expressing bsrm by the addition of a lethal concentration of BS was also observed with the human keratinocytes expressing bsrm (Figs. 6, 7). These results indicate that the death process triggered by the expression of bsrm in keratinocytes should follow the same way.
Interestingly the lethal effect provoked by bsrm appears to be specific to keratinocytes, because none of the cell types tested here died in our experimental conditions, except for the human and murine keratinocytes. As the gene transfer and expression of bsrm are an essential step to access the lethal effect of bsrm, an alternative strategy used to verify the gene expression and its effect was selecting the transduced cells with 8 μg/ml BS and exposing the cells to a fresh medium without BS. Even in such conditions the bsrm gene did not cause any morphological alterations to those cell types, whereas the BALB/MK cells and the human primary keratinocytes transduced with LBmSN and selected with BS died in a few days after removal of the antibiotic (Figure 7). Therefore, we conclude that the lethal activity of bsrm is specific to those keratinocytes.
The analysis of DOKEB through exclusion molecular chromatography showed that the factor has a molecular weight equivalent of two amino acids (not shown). Therefore, DOKEB should not be BS deaminase, or even any protein. Further purification and molecular analysis are in progress.
In this study we demonstrated only in vitro that the expression of the reporter gene bsrm has a lethal effect on keratinocytes. However as most of the gene therapy experiments using keratinocytes are carried out ex vivo with retroviral vectors, our finding has a very important meaning. Because the cells transduced with retroviral vector carrying bsrm and selected with BS can survive until the antibiotic is maintained in the medium, but when those cells are returned to the own organism, which has no BS in it, DOKEB will be produced and can provoke serious lesion in the body.
By this study we also point out the danger of using heterologous genes, in particular those isolated from the microorganisms, in gene transfer and gene therapy experiments without proper controls.
Conclusions
We demonstrated in this study that the expression of the reporter gene bsrm has a lethal effect on the murine BALB/MK cell line and human primary keratinocytes. It is likely over expression of the BS deaminase gene is responsible for the death. The death appears to be mediated by a factor, which is secreted by the BALB/MK transduced cells. By this study we point out the danger of using heterologous genes, in particular those isolated from the microorganisms, in gene transfer experiments without proper controls.
Methods
Retroviral vectors
The retroviral vectors used in the present study are based on the Moloney murine leukemia virus: LXSN [10], LBSN [8] and LBmSN [8]. The letters L, X, S, N, B, Bm of those vectors represent retroviral LTR promoter, cloning site, promoter of simian virus SV40, neomycine resistance gene (neoR), bsr and bsrm, respectively. The LBSN and LBmSN vectors were constructed inserting the bsr and bsrm genes into the Hpa I site of LXSN, which is located in the cloning site [8].
Cell line culture
The amphotropic retrovirus producing cell clones PA317/LBmSN, PA317/LBSN and PA317/LXSN [8], the murine fibroblast NIH3T3, HeLa, the human colorectal carcinoma cell lines LISP-A10 and LISP-E11 [14] were cultured in Dulbecco's modified Eagle medium (DMEM) with high glucose (4.5 g/ml), supplemented with 2 mM glutamine, 200 U/ml penicillin, 200 μg/ml streptomycin and 10 % fetal bovine serum (FBS) (InVitrogen, São Paulo, Brazil) at 37°C in a humidified atmosphere with 5 % CO2. The mouse fibroblast CCL-92 (ATCC) was cultured as above, except that the FBS was replaced with the bovine calf serum (InVitrogen, São Paulo, Brazil). For the culture of the murine keratinocyte BALB/MK cells (kindly provided by Dr Stuart A. Aaronson, The Derald H. Ruttenberg Cancer Center, New York, NY), EMEM (Biofluids, Rockville, MD) containing 0.05 mM CaCl2, 10 ng/ml of EGF and 10 % FBS was used.
The human colorectal carcinoma cell line HCT-8 and the murine melanoma cell line B16F10 [15] were cultured in RPMI 1640 (InVitrogen, São Paulo, Brazil) supplemented with 0.2 % NaHCO3, HEPES 10 mM, pH 7.3, 40 μg/ml garamicine and 10 % FBS at 37°C in a humidified atmosphere with 5 % CO2.
Rat primary smooth muscle cell culture and viral transduction
A primary culture of rat smooth muscle cells was prepared as previously described [12], digesting the Wistar isogenic rat aortas enzymatically. These cells were characterized immunocytochemically using antibodies against α-actin (Boehringer Mannheim, São Paulo, Brazil) for SMC positive staining and von Willebrand factor (Boehringer Mannheim, São Paulo, Brazil) for SMC negative and endothelial cell positive staining [12]. Only early-passage smooth muscle cells were exposed for 24 h to virus harvested from PA317/LBmSN cells for a period of 24 h in the presence of Polybrene (8 μg/ml, Sigma)
Transduction of mammalian cell lines with retroviral vectors
The target cells were seeded on a 24 well plate at 1 × 104 cells per well with an appropriate medium as mentioned above. In parallel, 1 × 106 of virus producing cells (PA317/LBmSN, PA317/LBSN or PA317/LXSN) were seeded in a 25 cm2 flask. After 24 h, the media from the target cells and the virus producing cells were replaced with a fresh one used for target cells. On the next day, the media of the target cells were replaced with 500 μl of the virus solution collected from the supernatant of the PA317/LBmSN cell culture and filtered in 0.45 μm syringe filter. Polybrene was added to the virus solution at the final concentration of 8 μg/ml. One day after the infection, the media were replaced with a fresh one, maintained in the CO2 incubator and the cells were observed using a microscope everyday.
In parallel, after two days of the infection, a new set of the transduced cells was split and only 1/10 part of the cells was maintained in the same well. The BS antibiotic was added to the wells at concentrations between 0 to 32 μg/ml. When the cells reached confluence they were split and transferred to two wells of a 12-well plate. To one well, BS was added at the concentration used for selection, and another well was maintained without BS. The cells were observed under the microscope everyday.
Human primary keratinocytes culture and viral transduction
Normal human keratinocytes from healthy adult volunteers were obtained by biopsy, cut in small pieces and incubated in a trypsin solution (0.05 %) containing 0.01 % EDTA at 37°C for 3 h under constant agitation. Every 30 min the detached cells were transferred to a new 75 cm2 flask containing 2 × 106 cells of the irradiated CCL-92 cells (60 Gy) as a feeder-layer. The medium used for the culture of the keratinocytes was composed of DMEM and Ham's F12 (2:1) containing 10 % FBS, 4 mM glutamine, 50 IU/ml streptomycin- penicillin, 0.18 mM adenine, 5 μg/ml insulin, 5 μg/ml transferrin, 0.4 μg/ml hydrocortisone, 0.1 nM choleric toxin and 20 pM triiodotyronin [11]. The medium was replaced every 2 to 3 days with a fresh one containing 10 ng/ml EGF. The cells were maintained in a humidified atmosphere with 5 % CO2.
For retroviral transduction, the packaging clones PA317/LBmSN and PA317/LXSN were irradiated with 30 Gy and used as a feeder-layer for the prepared previously keratinocyte cultures.
Author's contribution
FMB performed experiments with NIH3T3, keratinocytes and smooth muscle cells, and CBS with HCT-8, B16F10 and BALB/MK cells. DT and AKC performed purification and characterization of DOKEB. TRM tested analogs of BS in BALB/MK and BALB/MK/LBmSN cells. MBM participated in the preparation of the human primary keratinocytes. SWH drafted the manuscript and conducted all experiments.
Acknowledgments
This study (00/14639-3) was supported by Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) in Brazil and FM Bento, CB Sacramento and Machado TR were recipients of FAPESP fellowships (98/02006-4, 01/09774-1 and 02/02032-2) and D Takeshita of CNPq.
Figures and Tables
Figure 1 Effect of BS upon the BALB/MK cells transduced with LBmSN On day 1, BALB/MK cells were seeded on a 24-well plate at 1 × 104 cells/well, and 1 × 106 cells of the virus producing cell lines PA317/LBmSN (2 × 105 cfu/ml) and PA317/LXSN (5 × 106 cfu/ml) were seeded in 25 cm2 flasks. On day 2, the media of PA317 cell clones were replaced with EMEM without EGF. On day 3, the media of PA317 cells containing virus were harvested, centrifuged at 10000 rpm per 1 min in Eppendorf centrifuge. The media of BALB/MK cells were replaced with 0.5 ml of virus solution and supplemented with 5 ng/ml EGF and 8 μg/ml Polybrene. On day 4, the media were replaced with fresh EMEM in the following conditions. After 12 days, the cells were stained with Coomassie Blue and photographed. The results are representative of at least three experiments, which gave essentially the same results. (A) BALB/MK; (B) BALB/MK + 500 μg/ml G418; (C) BALB/MK + 8 μg/ml BS; (D) BALB/MK/LXSN + 500 μg/ml G418; (E) BALB/MK/LXSN; (F) BALB/MK/LXSN + 8 μg/ml BS; (G) BALB/MK/LBmSN; (H) BALB/MK/LBmSN + 500 μg/ml G418; (I) BALB/MK/LBmSN + 0.05 μg/ml BS; (J) BALB/MK/LBmSN + 0.1 μg/ml BS; (K) BALB/MK/LBmSN + 0.2 μg/ml BS; (L) BALB/MK/LBmSN + 0.5 μg/ml BS; (M) BALB/MK/LBmSN + 1 μg/ml BS; (N) BALB/MK/LBmSN + 2 μg/ml BS; (O) BALB/MK/LBmSN + 4 μg/ml BS; (P) BALB/MK/LBmSN + 8 μg/ml BS; (Q) BALB/MK/LBmSN + 16 μg/ml BS; (R) BALB/MK/LBmSN + 32 μg/ml BS(S) Five days after the LBmSN transduction, the cells were detached with trypsin and stained with Trypan blue. Left and right sides represent BALB/MK (control) and BALB/MK/LBmSN cells respectively.
Figure 2 Phase contrast microscopic appearance of the BALB/MK cells transduced with LBmSN The cells were cultured and transduced as described above. After 8 days of the transduction, the cells were photographed (X 200). The results are representative of at least five experiments, which gave essentially the same results. (A) BALB/MK; (B) BALB/MK/LBmSN + 8 μg/ml BS; (C) BALB/MK/LBmSN
Figure 3 The effect of the concentration of the virus LBmSN upon BALB/MK cells The cells were cultured and transduced as described in Figure 1, however the virus concentration in each plate varied as follows. After 10 days of the transduction, the cells were stained with Coomassie Blue and photographed. The results are representative of at least three experiments, which gave essentially the same results. (A) BALB/MK + 8 μg/ml BS; (B) BALB/MK + 102 cfu of LBmSN + 8 μg/ml BS; (C) BALB/MK + 103 cfu of LBmSN + 8 μg/ml BS; (D) BALB/MK + 104 cfu of LBmSN + 8 μg/ml BS; (E) BALB/MK + 105 cfu of LBmSN + 8 μg/ml BS; (F) BALB/MK; (G) BALB/MK + 102 cfu of LBmSN; (H) BALB/MK + 103 cfu of LBmSN; (I) BALB/MK + 104 cfu of LBmSN; (J) BALB/MK + 105 cfu of LBmSN
Figure 4 Effect of analogs of BS upon BALB/MK cells transduced with LBmSN The cells were cultured and transduced with LBmSN as described in Figure 1. On day 4 the media were replaced with a fresh one and BS-analogs were added at the indicated concentrations. This time is indicated as day zero in the figure.
Figure 5 The lethal effect of the supernatant from the BALB/MK cells transduced with LBmSN On day 1, BALB/MK cells were seeded on a 6 well plate at 4 × 104 cells/well, and 1 × 106 cells of the virus producing cell lines PA317/LBmSN and PA317/LXSN were seeded in 25 cm2 flasks. On day 2, the media of the PA317 cell clones were replaced with the EMEM supplemented with 10 % FBS and without EGF. On day 3, the media of PA317 cells containing virus were harvested, centrifuged at 10,000 rpm per 1 min in Eppendorf centrifuge. The media of BALB/MK cells were replaced by 2 ml of virus solution and complemented with 5 ng/ml of EGF and 8 μg/ml of Polybrene. In parallel, 1 × 104 BALB/MK cells were seeded on a 24 well plate. On day 4, the media were replaced by fresh ones. On day 5, the medium of BALB/MK cells was replaced by 0.5 ml of the medium of the LBmSN transduced BALB/MK cells, which was previously centrifuged at 10,000 rpm/ 1 min. After 7 days, the cells were photographed (X 200). The results are representative of at least three experiments, which gave essentially the same results. (A) The BALB/MK cells incubated with supernatant of BALB/MK cells transduced with LXSN; (B) The BALB/MK cells incubated with supernatant of BALB/MK cells transduced with LBmSN.
Figure 6 Effect of BS upon the human keratinocytes transduced with LBmSN vector On day 1, the irradiated PA317/LBmSN cells (30 Gy) and the human keratinocytes were seeded together at 3.3 × 105 and 5 × 104 cells per well respectively in a 12 well plate. Two days later, the medium was replaced with a fresh one containing the indicated concentrations of antibiotics. The cells were stained with Rhodamine B after 8 days. (A) no antibiotic; (B) 800 μg/ml G418; (C) 8 μg/ml BS; (D) 12 μg/ml BS; (E) 16 μg/ml BS; (F) 32 μg/ml BS
Figure 7 Effect of BS upon the keratinocytes transduced with LBmSN and selected with BS The keratinocytes were cultured and selected with 8 μg/ml BS as described in Figure 5. When the keratinocytes reached 70 % confluence, the cells were divided into two parts and transferred to 2 wells of a 12-well plate that contained the irradiated NIH3T3 cells. In one well the cells were maintained without BS (A) and in another with 8 μg/ml BS (B). The cells were stained with Rhodamine B after 10 days.
Table 1 The effect of cell density on inducing death
Cell concentration Time till cell death1 (days)
1.0 × 103 8 (0)
5.0 × 103 7 (1)
1.0 × 104 7 (0)
1.5 × 104 6 (1)
2.0 × 104 5 (1)
4.0 × 104 5 (1)
On day 1, 1 × 105 BALB/MK cells and 1 × 106 PA317/LBmSN cells were seeded in 25 cm2 flasks. On day 2, the medium from the PA317/LBmSN cells was replaced with EMEM supplemented with 10 % FBS and without EGF. On day 3, the medium of the PA317/LBmSN cells containing virus was harvested, centrifuged at 10,000 rpm per 1 min in an Eppendorf centrifuge. The medium of the BALB/MK cells was replaced with 4 ml of the virus solution and complemented with EGF and Polybrene at a final concentration of 5 ng/ml and 8 μg/ml respectively. On day 4, the cells were trypsinized and seeded in the 24 well plate at the indicated number of cells containing 8 μg/ml BS.
1 Results are the means ± S.D. of three experiments.
Table 2 Effect of the BALB/MK cells expressing bsrm upon BALB/MK cells
BALB/MK BALB/MK/LBmSN BS1 EFFECT
+ + - D
+ + + L
- + - D
- + + L
+ - - L
Both cell lines, BALB/MK and BALB/MK/LBmSN, were seeded in a 24 well plate at 1 × 104 of each one/well or 2 × 104 cells of one cell line/well. Two days later, the media were replaced by fresh ones, and after 6 days the cells were stained with Coomassie Blue for microscopic analysis. The results are representative of at least two experiments, which gave essentially the same results.
1 [BS] = 8 μg/ml
D = dead
L = alive
+ = presence
- = absence
Table 3 Comparison of the vectors LBmSN and LBSN upon BALB/MK and NIH3T3 cells
BALB/MK NIH3T3
BS1 No antibiotic BS No antibiotic
LBmSN L D L L
LBSN L L L L
No vector D L D L
The experiment was carried out as described in the legend of the Figure 1. After 8 days of the transduction the cells were stained with Coomassie Blue for microscopic analysis. The results are representative of three independent experiments, which gave essentially the same results.
1 [BS] = In the media of the cells transduced with LBmSN and LBSN were added 8 μg/ml and 2 μg/ml of BS respectively, because the cells transduced with LBSN do not support concentrations of BS higher than 2 μg/ml [8].
D = dead
L = alive
==== Refs
Takeuchi S Hirayama K Ueda K Sakai H Yonehara H Blasticidin S, a new antibiotic J Antibiot (Tokyo) 1958 11 1 5 13525246
Kamakura T Kobayashi K Tanaka T Yamaguchi I Endo T Cloning and expression of a new structural gene for Blasticidn S deaminase, a nucleoside aminohydrolase. Agric Biol Chem 1987 51 3165 3168
Endo T Kobayashi K Nakayama N Tanaka T Kamakura T Yamaguchi I Inactivation of blasticidin S by Bacillus cereus. II. Isolation and characterization of a plasmid, pBSR8, from Bacillus cereus J Antibiot (Tokyo) 1988 41 271 273 2833485
Kobayashi K Kamakura T Tanaka T Yamaguchi I Endo T Nucleotide sequence of the bsr gene and N-terminal amino acid sequence of blasticidin S deaminase from blasticidin S resistant Escherichia coli TK121 Agric Biol Chem 1991 55 3155 3157 1368770
Nawa K Tamura Y Sato K Hattori J Shimotohno KW Endo T Inactivation of blasticidin S by Bacillus cereus. V. Purification and characterization of blasticidin S-deaminase mediated by a plasmid from blasticidin S resistant Bacillus cereus K55-S1 Biol Pharm Bull 1995 18 350 354 7742811
Yamaguchi I Seto H Misato T Substrate binding by blasticidin S deaminase, an aminohydrolase for novel 4-aminopyrimdine nucleosides. Pest Biochem Physiol 1986 25 54 62 10.1016/0048-3575(86)90031-3
Izumi M Miyazawa H Kamakura T Yamaguchi I Endo T Hanaoka F Blasticidin S-resistance gene (bsr): a novel selectable marker for mammalian cells Exp Cell Res 1991 197 229 233 1720391 10.1016/0014-4827(91)90427-V
Freitas AC Bento FM Ramesh N Osborne WR Han SW Modified blasticidin S resistance gene (bsrm) as a selectable marker for construction of retroviral vectors J Biotechnol 2002 95 57 62 11879712 10.1016/S0168-1656(01)00442-4
Weissman B Aaronson SA Members of the src and ras oncogene families supplant the epidermal growth factor requirement of BALB/MK-2 keratinocytes and induce distinct alterations in their terminal differentiation program Mol Cell Biol 1985 5 3386 3396 2427928
Miller AD Rosman GJ Improved retroviral vectors for gene transfer and expression Biotechniques 1989 7 980 6, 989 2631796
Green H Kehinde O Thomas J Growth of cultured human epidermal cells into multiple epithelia suitable for grafting Proc Natl Acad Sci U S A 1979 76 5665 5668 293669
Beltrao-Braga PC Koh IH Silva MR Gutierrez PS Han SW Vascular adventitia is a suitable compartment to transplant transduced vascular smooth muscle cells for ex vivo gene expression Cell Transplant 2002 11 583 592 12428748
Kim HT Tasca S Qiang W Wong PK Stoica G Induction of p53 accumulation by Moloney murine leukemia virus-ts1 infection in astrocytes via activation of extracellular signal-regulated kinases 1/2 Lab Invest 2002 82 693 702 12065679
Solimene AC Carneiro CR Melati I Lopes JD Functional differences between two morphologically distinct cell subpopulations within a human colorectal carcinoma cell line Braz J Med Biol Res 2001 34 653 661 11323753
Fidler IJ Activation in vitro of mouse macrophages by syngeneic, allogeneic, or xenogeneic lymphocyte supernatants J Natl Cancer Inst 1975 55 1159 1163 1206741
| 15575952 | PMC539255 | CC BY | 2021-01-04 16:02:57 | no | BMC Biotechnol. 2004 Dec 2; 4:29 | utf-8 | BMC Biotechnol | 2,004 | 10.1186/1472-6750-4-29 | oa_comm |
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-4-311557921110.1186/1472-6750-4-31Research ArticleA sugar beet chlorophyll a/b binding protein promoter void of G-box like elements confers strong and leaf specific reporter gene expression in transgenic sugar beet Stahl Dietmar J [email protected] Dorothee U [email protected] Reinhard [email protected] PLANTA Angewandte Pflanzengenetik und Biotechnologie GmbH, Grimsehlstr. 31, D-37555 Einbeck, Germany2 Institut für Genetik – Biozentrum, Technische Universität Braunschweig, Spielmannstr. 7, D-38106 Braunschweig, Germany3 Genedata GmbH, Lena-Christ Strasse 50, D-82152 Martinsried, Germany2004 5 12 2004 4 31 31 13 7 2004 5 12 2004 Copyright © 2004 Stahl 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
Modification of leaf traits in sugar beet requires a strong leaf specific promoter. With such a promoter, expression in taproots can be avoided which may otherwise take away available energy resources for sugar accumulation.
Results
Suppression Subtractive Hybridization (SSH) was utilized to generate an enriched and equalized cDNA library for leaf expressed genes from sugar beet. Fourteen cDNA fragments corresponding to thirteen different genes were isolated. Northern blot analysis indicates the desired tissue specificity of these genes. The promoters for two chlorophyll a/b binding protein genes (Bvcab11 and Bvcab12) were isolated, linked to reporter genes, and transformed into sugar beet using promoter reporter gene fusions. Transient and transgenic analysis indicate that both promoters direct leaf specific gene expression. A bioinformatic analysis revealed that the Bvcab11 promoter is void of G-box like regulatory elements with a palindromic ACGT core sequence. The data indicate that the presence of a G-box element is not a prerequisite for leaf specific and light induced gene expression in sugar beet.
Conclusions
This work shows that SSH can be successfully employed for the identification and subsequent isolation of tissue specific sugar beet promoters. These promoters are shown to drive strong leaf specific gene expression in transgenic sugar beet. The application of these promoters for expressing resistance improving genes against foliar diseases is discussed.
==== Body
Background
Sugar beet (Beta vulgaris L.) is a biennial plant, a member of the Chenopodiaceae family [1]. In the first year after germination, a rosette of leaves develops while the taproot swells and accumulates sucrose. In the second year, flower initiation is induced after vernalization in the preceding winter. Beets are harvested at the end of the first year when sugar content of the taproot is high. Transgenic approaches towards modification of specific traits comprise the increase of pathogen resistance, the increase of sugar content and the improvement of sugar storage. These approaches require promoters that direct gene expression in a timely and spatial manner which is determined by the desired expression profile of the transgene.
In many cases improvement of transgenic traits in plants were achieved by using specific promoters [2,3]. Furthermore, to accomplish high tissue specific protein production in transgenic plants, often promoters from photosynthetic or storage specific genes are employed [4,5].
For the identification of desired promoters, a subtractive approach to enrich differentially expressed genes or a large scale approach to identify these genes in cDNA libraries may be employed prior to promoter isolation. One possible way to identify nonredundant clones in a cDNA library is the method of oligonucleotide fingerprinting (ofp) which was recently applied to sugar beet [6]. With this approach different cDNAs can be identified on a large scale basis within a cDNA library. While a large scale ofp approach is a feasible method to identify differentially expressed genes in different cDNA libraries, this method is very cost intensive and hence not applicable for many research groups.
A straight forward approach for the isolation of differentially expressed genes was achieved by the "Suppression Subtractive Hybridization" method (SSH) [7]. SSH is a cDNA- and a PCR-based technique that includes a step for the equalization of the abundance of different cDNA fragments during subtractive hybridization. Combined with suppression PCR, selective amplification of differentially expressed cDNA sequences was achieved without the application of physical separation methods [7]. This method was recently applied to isolate taproot expressed genes from sugar beet [8].
Here we have employed SSH for the isolation of leaf expressed sugar beet genes. Among the genes isolated was a cDNA fragment for a light-harvesting chlorophyll a/b binding protein (CAB). It is shown that sugar beet genotypes harbor either one or two cab genes that are both expressed. To investigate the use of the cab promoters for gene expression, the 5' regulatory sequences were isolated and linked to reporter genes. Transient reporter gene assays indicate that both promoters are expressed in sugar beet leaves. In transgenic sugar beet both promoters are expressed in green tissue. Sequence analysis revealed that the cab11 promoter, in contrast to the cab12 promoter, is void of G-box like regulatory sites with a palindromic ACGT core sequence. A leaf specific promoter in transgenic sugar beets can be employed for biotechnological applications.
Results
Identification of leaf expressed genes from sugar beet
To isolate cDNA fragments corresponding to leaf expressed genes, poly(A)+RNAs from leaf and taproot were isolated and subjected to cDNA synthesis. Suppression subtractive hybridization was performed as described in Methods. A total of 23 cDNA clones specific for the subtraction for leaf expressed genes were isolated and sequenced (data not shown). Fourteen out of 23 cDNAs were found to be different. Table 1 shows that 10 out of these 14 clones have homology mainly to nuclear encoded genes involved in the calvin cycle and in photosynthesis. Fragments L6 and L11 detect homology to the same gene but do not overlap. Four clones do not have any sequence homology to a known gene.
To further confirm that the enriched cDNA fragments correspond to tissue specific genes, Northern blot hybridizations to total RNA isolated from leaves (L), taproots (R), stems (S), and inflorescences (I) were performed with 7 cDNA fragments. Figure 1 shows that all but one of the cDNA fragments hybridize specifically or much stronger to RNA from leaves than to RNA from taproots. Clone L5 hybridizes with similar intensity to RNA from all four tissues.
The expression profile of the gene for SSH fragment L2 encoding a chlorophyll a/b binding protein (CAB) was analysed in more detail. RNA from different tissues and from different developmental stages were used in RNA blot experiments. RNA was isolated from different organs of sugar beets from a field 4, 6, 10, 12, 16 and 19 weeks after sowing. This interval covers the lifetime of a sugar beet in central Europe. Figure 2 shows that the gene is expressed only in above ground tissue like petiole, sink and source leaf regardless of the developmental stage. Since no expression is observed in below ground tissue like primary and lateral root the promoter of the cab gene was isolated and analysed with reporter gene fusions. Compared to the other genes, the expression profile of the gene for SSH fragment L2 showed the strongest tissue specificity during the different time points analysed (data not shown).
Isolation of two promoters for the light-harvesting chlorophyll a/b binding protein
For the isolation of a sugar beet promoter corresponding to the cab gene, a complete cDNA clone was isolated (GenBank Acc. Nr. AJ579711, see Methods). A homology search with the encoded 252 amino acid long protein reveals a 87% identity to the cab11 and 85% identity to the cab12 gene from tomato encoding chlorophyll a/b binding protein [9].
Prior to the isolation of genomic clones, a gene copy number analysis was performed. DNA from sugar beet genotypes 1K0088 and 4B5421 was digested with different restriction enzymes and hybridized to the L2 fragment (Table 1). Figure 3 shows that in genomic DNA from genotype 1K0088 three (EcoRI, HindIII) and four (PstI) hybridizing fragments are detected, while only one (EcoRI) or two (PstI, HindIII) fragments are detected in genotype 4B5421. From this result it is concluded that genotype 4B5421 harbors one and genotype 1K0088 harbors two copies of the cab gene.
Genomic clones for the two different genes were isolated (Methods). Sequence comparison between the cDNA and both genomic clones indicate a very high degree of sequence identity within the coding region. The CAB11 and CAB12 amino acid sequence differ only in one position (data not shown).
From both genes the promoter regions were subcloned into plasmid vectors and sequenced (Methods). Sequence of 1148 and 3049 base pairs, respectively, containing most of the upstream region were deposited to GenBank (Acc. Nr. AX449166 and AX449167). The 1148 bp promoter fragment is designated cab11 promoter and the 3049 bp fragment cab12 promoter. Both fragments harbor 51 base pairs coding region of the CAB protein and 113 (cab11) and 70 (cab12) base pairs upstream untranslated sequence. Upstream of the untranslated region only about 300 bp are homologous between the two promoters while the rest of the sequence is highly divergent (data not shown). Because the cDNA clone isolated before originates from the genotype 4B5421 and corresponds to the cab11 gene, it was investigated if the second gene is also transcribed. Towards these ends 5' RACE amplifications were performed with RNA from genotype 1K0088 and sequenced. This analysis revealed that the cab12 gene is also transcribed (data not shown).
Transient expression assays in sugar beet leaves
To test whether the isolated promoters confer reporter gene expression in sugar beet leaves, a transient assay was performed. Towards these ends translational fusions of promoter fragments with the luciferase gene from Photinus pyralis in the vector pGEM-luc were constructed (see Methods). Table 2 summarizes the relative gene expression strength obtained with the different promoter reporter gene constructs when transformed into sugar beet leaves. Relative to the CaMV 35S promoter, the strength of all promoter fragments is approximately 20% and no significant drop in gene expression is observed between the largest and smallest promoter fragments. In summary, it can be concluded that the 1097 bp long fragment from the cab11 gene and the 342 bp fragment from the cab12 gene harbor all cis-regulatory sequences required for gene expression in sugar beet leaves at least under the conditions of the transient bioassay.
The promoter of two chlorophyll a/b binding protein genes confers leaf specific and light inducible gene expression in transgenic sugar beet plants
To investigate if the cab11 and cab12 promoters from sugar beet drive reporter gene expression in transgenic plants, promoter reporter gene fusions were introduced via Agrobacterium mediated transformation in sugar beet (see Methods). The length of the cab11 promoter fragment is 1097 bp and the length of the cab12 promoter fragment 2998 bp. Leaves of transgenic lines transferred to the greenhouse were analysed for reporter gene activity. The cab11 promoter of 12 independent transformants showed a specific β-glucuronidase (GUS) activity from 9 to 40599 pmol Mu × min-1 × mg-1, respectively (Fig. 4A). The cab12 promoter of 4 transformants showed a specific activity from 223 to 11656 pmol Mu × min-1 × mg-1, respectively (Fig. 4B). These results indicate that the 1097 bp cab11 promoter fragment and the 2998 bp cab12 promoter fragment are sufficient to confer promoter activity to transgenic sugar beet leaves. Furthermore, the cab11 promoter seems to be stronger than the cab12 promoter although more transgenic lines were analysed for cab11 than for cab12.
In order to analyse if the cab11 and cab12 promoters confer tissue specific expression to sugar beet, the roots of three transgenic cab11 and three cab12 promoter lines were analysed. According to the strength of the cab11 and cab12 promoters in leaves (Fig. 4A and 4B) transgenic lines were selected which show low, moderate or high GUS activity in leaves. None of the lines showed GUS activity in the roots which was above the background level of nontransgenic control plants (Fig. 4C). Therefore the promoter activity of the cab11 and cab12 regulatory element is restricted to the above ground tissue of sugar beet and absent in roots. This result is consistent with the observations that transcripts of the cab genes are not detectable in the below ground tissue by Northern blot hybridization (Fig. 2).
To analyse cab11 and cab12 gene regulation in response to light, the reporter gene activtity of etiolated and green transgenic sugar beets was compared. In vitro shoots of the transgenic lines C1-121, C1-122, C2-50 and C2-52 were etiolated for 40 days in the dark. The GUS activity of the etiolated leaflets of one half of the plants was determined. Seven days after illumination the GUS activity of the remaining plants was determined after greening of the leaves. In two independent experiments the cab11 and cab12 promoter plants showed a 4,3 to 8,3 fold and a 95 to 118 fold induction during greening, respectively (Table 3). Although the GUS activity of the etiolated plants was comparable in the two assays, the greening plants showed a much stronger reporter gene activity in the second experiment. Apparently, different time points during the differentiation of etioplasts to chloroplasts were analysed. Time after illumination seems to have a strong influence on the level of promoter induction. Furthermore, gene expression is much stronger in these experiments compared to the analysis in transgenic plants (Table 3, Fig. 4). Finally, these results show that the cab11 and cab12 promoter are activated during the light induced plastid development.
The cab11 promoter lacks G-box elements with a palindromic ACGT core sequence
Gene expression is mainly regulated by the binding of transcription factors to specific cis-regulatory elements. Because the two sugar beet promoters show a similar expression profile, it was investigated if there are any differences or similarities in the composition of cis-regulatory sequences in both promoters. Towards these ends, a database-assisted approach was employed [10]. The Patch™ program [11] was used to inspect both promoter sequences for the occurence of plant transcription factor (TF) binding sites that are annotated to the TRANSFAC® database [11]. The results reveal a large number of putative TF binding sites in both promoters (data not shown). Upon closer inspection of the results it was striking that the cab11 promoter, in contrast to the cab12 promoter was completely void of putative G-box like binding sites that contain a conserved ACGT core sequence and are recognized by bZIP transcription factors. Using the program Patch™ and entering a lower score boundary of 100% to detect only experimentally verified binding sites, only two putative binding sites for bZIP transcription factors were found in the cab11 promoter. Figure 5 shows the positions of these two motifs that both lack the ACGT core sequence. The sequence motifs at position -749 and -489 relative to the translation start site were found in the rice glutelin-B1 promoter and are bound by the rice bZIP transcription factor family RISBZ [12]. Both sites were also found to be bound by the tobacco bZIP transcription factor TGA1a in a pea lectin promoter [13].
Inspecting the sequence of the cab12 promoter for the ACGT core sequence of bZIP factor binding sites reveals 12 positions for this motif (data not shown). Using the program Patch™ six experimentally verified binding sites for bZIP factors were detected among these twelve sites that harbor the ACGT core in the cab12 promoter (Fig. 5). The motif at position -2104 is also present in the glutathione-S transferase 6 gene promoter of Arabidopsis where it is bound by the factor OBF4 [14]. The same site and the sites at position -1608 and -1247 occur in the embryonic abundant protein 1 promoter of rice and are recognized by the factors OSBZ8 and TRAB1 [15-17]. The sites at position -1767 and -1599 were recognized as bZIP binding sites in many other systems. The sequence TGACGT is part of the as-1 element of the CaMV 35S promoter that was shown to be bound by tobacco TGA1a, TGA1b, and TGA2.2 [18,19]. The site at position -659 is also present in the CaMV 35S promoter where it was shown to be bound by the wheat nuclear factor HBP-1 [20].
The observation that the cab11 promoter lacks G-box like elements with a conserved ACGT core sequence indicates that such sites are not required for leaf specific gene expression.
Discussion
The chlorophyll a/b binding proteins CAB11 and CAB12 from sugar beet belong to the light harvesting complex I – 730 (LHCI-730)
Subtractive hybridization was used to isolate leaf expressed genes from sugar beet. The goal was the identification of a promoter that drives leaf specific gene expression in transgenic sugar beet plants. Among seven analysed genes a cDNA fragment corresponding to a chlorophyll a/b binding protein gene was shown by RNA gel blot hybridization to be highly specific for green tissue (Fig. 1 and 2). Genomic DNA blot hybridizations indicate that the two sugar beet genotypes investigated harbor either one or two copies of the gene designated Bvcab11 and Bvcab12 (Fig. 3). A complete cDNA for the gene from genotype 4B5421 was isolated and encodes a protein of 252 amino acids that shows the highest homology (87%) to the cab11 gene from the light harvesting complex I (LHCI) in tomato [9]. This and homologies to other LHCI proteins indicate that the sugar beet gene belongs to the type IV LHCI complex [21]. Further support for this classification comes from the observation that the intron positions between cab11 from tomato and Bvcab11 from sugar beet are identical (data not shown).
LHCI can be subdivided into at least two different chlorophyll-protein complexes, one of which appears to be responsible for the 730 nm fluorescence of PSI (LHCI-730) and the other complex (LHCI-680) fluoresces at lower wavelength [21]. In barley the LHCI-730 complex was isolated as a heterodimer composed of the type I and type IV polypeptides [22]. Furthermore, tomato type I and type IV LHCI polypeptides (Lhca1/cab6a and Lhca4/cab11) expressed in E. coli form a heterodimer in vitro that closely resembles the native LHCI-730 dimer from tomato leaves [23]. Therefore, the sugar beet CAB11 and CAB12 proteins may be part of the LHCI-730 complex.
G-box like elements are not a prerequisite for leaf specific gene expression
The promoters for both sugar beet cab genes were isolated and linked to reporter genes. Transient gene expression studies in sugar beet indicated that 1097 bp upstream of the ATG from the Bvcab11 gene and 342 bp upstream of the ATG from the Bvcab12 gene are sufficient for leaf specific gene expression in sugar beet (Table 2).
Promoter reporter gene constructs for Bvcab11 and Bvcab12 were stably transformed into sugar beet (Beta vulgaris, var. VRB). In sugar beet both promoters are expressed in leaves (Fig. 4).
When the promoter sequences of both cab genes where analysed for putative transcription factor binding sites, a striking difference was observed. The Bvcab11 promoter lacks G-box like sequences with a palindromic ACGT core. Are G-boxes required for light or leaf specific gene expression? A 268 bp fragment of the wheat cab-1 promoter functions as a light responsive and organ specific enhancer in transgenic tobacco [24]. Most notably the three regions that interact with nuclear factors and that were able to enhance gene expression of a 90 bp CaMV 35S minimal promoter did not contain a G-box sequence [24]. The requirement of G-box sequences for light specific gene expression has also been analysed directly [25]. A trimer of the G-box motif found in the spinach ribulose-1,5-bisphosphate carboxylase small subunit-1 promoter was fused to a 90 bp CaMV 35S minimal promoter. While a mutant of this G-box did not confer gene expression to the minimal promoter in the dark and under different light conditions, the G-box increased reporter gene expression under these conditions [25]. Reporter gene expression in the dark was comparably higher than under different light conditions. This is similar to the finding that a G-box like sequence in the cab1R gene of rice is necessary for high level transient expression of a reporter gene in tobacco leaf tissue [26]. Taken together, this indicates that the presence of G-box sequences may have a quantitative effect but may not be a prerequisite for green tissue specific gene expression in sugar beet.
Biotechnological applications of leaf specific promoters in sugar beet
The major goal of this work was the isolation of a strong leaf specific sugar beet promoter that can be used for biotechnological applications. Disease control is one of the most important goals for biotechnological approaches towards improving sugar beet performance. There are many leaf spot diseases that are detrimental to the plant. For example, Cercospora leaf spot is one of the most widespread and destructive foliar diseases of sugar beet [27]. Expressing resistance improving genes in a strong and specific manner against pathogens causing foliar diseases may require a strong leaf specific promoter. With such a promoter, expression in taproots can be avoided which may otherwise take away available energy resources for sugar accumulation.
The work here shows that two promoters, Bvcab11 and Bvcab12, have been isolated that drive highly leaf specific gene expression in sugar beet (Fig. 4). No expression above background levels was detected for both promoters in sugar beet roots (Fig. 4C).
Based on the expression strength in transgenic plants, the Bvcab11 promoter may be suitable for biotechnological applications because it achieves a reporter gene activity comparable to the strong CaMV 35S promoter. CaMV 35S-mediated GUS activities in transgenic tobacco plants were reported as 113000 U (average of 10 plants, [28]) and 9000 U (average of 15 plants, [29]) in which 1 Unit refers to pmol 4-Mu produced min-1 × mg protein-1 [30].
The highest level of cab11 derived GUS expression is 40599 pmol Mu x min-1 × mg-1 which is comparable with the expression strength of the strong CaMV 35S promoter in tobacco.
Conclusions
In summary, this work presents the isolation and expression analysis of two cab promoters from sugar beet. Particularly, the Bvcab11 promoter may be useful to drive strong and specific gene expression in transgenic host plants. The lack of bZIP binding sites harboring the ACGT core sequence could also be advantageous for transient analysis of bZIP transcription factors when using a Bvcab11 reporter gene construct as a transformation control. Furthermore these promoters may be useful to express resistance improving genes against foliar diseases such as Cercospora leaf spot.
Methods
Preparation of RNA and genomic DNA
Two different methods for RNA preparation were employed. To isolate RNA for cDNA subtraction, the procedure described below was followed. For some of the Northern blot analyses a method described earlier was employed [31].
For RNA isolation plant material was homogenized in liquid nitrogen and resuspended in a solution containing 4 M guanidinthiocyanat, 25 mM Tris-HCl, pH8 und 100 mM β-mercaptoethanol. After centrifugation (4°C, 10 min. at 3300 rcf) nucleic acids in the supernatant were precipitated by addition of 0.03 volume sodium acetate (3 M, pH5) and 0.75 volume ethanol (100%) and incubation over night at -20°C. After centrifugation (4°C, 10000 g, 10 min.) the nucleic acid containing pellet was dissolved in 20 ml 100 mM NaCl, 10 mM EDTA pH8, 50 mM Tris-HCl pH8, and 0.2% SDS. Afterwards, a phenol:chloroform (1:1) and a chloroform:isoamylalcohol (24:1) extraction was performed. The pH of the aqueous solution was adjusted to about 5 with acidic acid and nucleic acids were precipitated by addition of 0.6 volume isopropanol and 0.05 volume 4 M NaCl and incubation for 2 hrs at -20°C. After centrifugation (20–30 min., 10000 g at 4°C) the nucleic acids containing pellet was resuspended in 10 ml H2ODEPC containing 0.1% SDS. Total RNA was precipitated by addition of 0.25 volume 8 M LiCl and incubation for at least 15 hrs at 4°C with subsequent centrifugation for 20 min at 4°C, 10000 g. Total RNA was resuspended in 400 μl H2ODEPC. After ethanol precipitation (addition of 0.1 volume sodium acetate, 3 M pH4.8, and 2.5 volume ethanol) total RNA was resuspended in a volume of 50–100 μl H2ODEPC. The isolation of poly(A)+ RNA was carried out with the Oligotex Kit according to the manufacturers protocol (Qiagen; Hilden, Germany). Measurements of RNA yield and electrophoretic separation on formaldehyde gels were done following standard protocols [[32], modified].
Genomic DNA was isolated from sugar beet genotypes 1K0088 and 4B5421 according to a previous published method [33]. For recombinant DNA work standard techniques were employed [32].
Suppression subtractive hybridization
The synthesis of cDNA was performed using the CLONTECH PCR-Select™ cDNA Subtraction Kit (Heidelberg, Germany). Each synthesis was carried out with 8 μg poly(A)+ RNA from sugar beet isolated either from leaves or taproots. Subtractive hybridization was done following the user manual (PT1117-1) of the CLONTECH PCR-Select™ cDNA Subtraction Kit. After the second PCR the amplified fragments from the forward and the reverse subtraction were cloned into the PCR cloning vector pCR®2.1. For each microgram PCR product approximately 300 recombinant plasmids were obtained. For the cloning of PCR products the Invitrogen T/A Cloning® Kit was employed (Karlsruhe, Germany). Prior to ligation into pCR®2.1 the subtracted PCR cDNA products were subjected to an additional incubation of 1 hour at 72°C with dATP and Taq polymerase (TaKaRa; Gennevilliers, France) to ensure that the majority of the PCR fragments contain "A-overhangs" for an efficient cloning into the T/A cloning vector.
DNA sequence analysis
The inserts of the plasmids were sequenced with fluorescently labeled M13 reverse and forward (-20) primers using the AutoRead Sequencing Kit (Pharmacia) and the Automated Laser Fluorescent A.L.F.™ DNA Sequencer from Pharmacia LKB (Freiburg, Germany). The DNA sequence analysis of the genomic and full-length cDNA clones was done by the custom sequencing service of MWG Biotech AG (Ebersberg, Germany). Sequences were subjected to data bank analysis using the BLAST algorithms [34] and analysed with the PILEUP programme of the GCG Wisconsin Analysis Package. For further promoter analysis the TRANSFAC® database was employed [11]. DNA sequences were also processed and analysed on a Macintosh computer using DNA Strider 1.3 [35] and a PC computer using Vector NTI Suite 8.0 (Informax).
Southern and Northern blot hybridizations
Radioactive probes were generated by the method of random hexamer priming with the Amersham Multiprime DNA Labelling System (Freiburg, Germany). Southern and Northern hybridizations were carried out following standard protocols [32,36].
For genomic Southern blot hybridizations 10 μg of DNA from sugar beet genotypes 1K0088 and 4B5421 was digested with different restriction enzymes. Electrophoretic separation, transfer to Hybond nylon membranes (Amersham Pharmacia Biotech, Freiburg), hybridization to radioactive probes, and exposure of the membrane to X-ray films were done according to standard protocols [32]. Radioactive probes were generated by labelling 20 ng of DNA with 50 μCi P32-dATP (6000 Ci/mMol, Amersham Pharmacia Biotech, Freiburg).
Isolation of cDNA and genomic clones
A leaf specific, directional cDNA library from sugar beet genotype 4B5421 was synthesized by the custom cDNA library service of GIBCO BRL (Rockville, USA) and cloned into the plasmid vector pCMV Sport 6.0. Screening of the library was done according to standard protocols [32]. Seven positive cab cDNA clones were identified after screening of 10000 clones using the SSH fragment L2 as a probe (Table 1). The longest cDNA is 1062 bp long, harbors a 114 bp non-translated leader, a 756 bp long reading frame, a 177 bp 3' nontranslated leader, and a 15 bp poly A tail (data not shown, GenBank Acc. Nr. AJ579711).
A genomic library from sugar beet genotype 1K0088 was generated in the lambda vector EMBL3 SP6/T7 and screened using standard protocols [32]. Genomic clones for two different cab loci were isolated.
The promoter for the gene cab11 is present on a ClaI fragment that was subcloned into a plasmid vector and completely sequenced. The fragment is 6294 bp long and contains 51 bp from the coding region of the gene. The resulting plasmid was designated pC1a. Additionally, a 6026 bp large SalI/ClaI fragment was released from the phage clone and subcloned into a Bluescript plasmid and designated pC1b. The promoter for the gene cab12 is present on a PstI fragment that was also subcloned into a plasmid vector and completely sequenced. The fragment is 4002 bp long and the harboring plasmid was designated pC2.
From both genomic clones 1148 and 3049 base pairs containing most of the upstream region were deposited in GenBank (Acc. Nr. AX449166 and AX449167). The 1148 bp promoter fragment is designated cab11 promoter and the 3049 bp fragment cab12 promoter.
Promoter reporter gene constructs
For transient gene expression assays, promoter fragments were linked as translational fusions to the luciferase reporter gene from Photinus pyralis in the reporter gene vector pGEM-luc (Promega, Mannheim). To introduce a plant polyA addition signal into pGEM-luc the respective fragment was isolated from pBI101.3 (Clontech, Heidelberg) by EcoRI digestion, followed by a Klenow fill in reaction and by redigestion with SacI. This released a 260 bp DNA fragment from the nopaline synthase (nos) gene containing the polyA addition signal. To directionally clone this fragment into pGEM-luc, this plasmid was first linearised with SfiI, treated with T4-polymerase to generate blunt ends and subsequently redigested with SacI. After inserting the nos fragment the resulting plasmid was designated pLuc-nos2. To insert the cab11 promoter fragment, a SalI(fill in)-AviII fragment was cloned into the ApaI linearised and T4-polymerase treated plasmid pLuc-nos2. This plasmid harbors 1145 bp from the cab11 promoter including the coding sequence for the first 16 amino acids of the cab11 gene. This plasmid was designated pC1L-1097. In this plasmid the luciferase gene is translationally fused with the first 16 amino acids from the cab11 gene. A second plasmid was generated which harbors additional upstream sequences. Towards these ends a 6099 bp KpnI fragment was released from the plasmid pC1b (see above) and the ends treated with T4-polymerase. The fragment was redigested with NotI and the desired fragment was directionally cloned as a KpnI(blunt end)-NotI fragment upstream of the cab11 fragment in pC1L-1097. To generate compatible ends pC1L-1097 was digested with HindIII treated with T4-polymerase and redigested with NotI. The resulting plasmid was designated pC1L-7126.
To clone the promoter for gene cab12 upstream to the luciferase coding region, the promoter fragment from pC2 was released by NotI/EcoRI digestion and subsequently subjected to a partial digestion with AviII. A 3100 bp long NotI/AviII fragment was purified and subcloned into pLuc-nos2. The plasmid pLuc-nos2 was digested with ApaI, the ends treated with T4-polymerase and redigested with NotI. After ligation the resulting plasmid was designated pC2L-2998. In this plasmid the luciferase gene is translationally fused with the first 16 amino acids from the cab12 gene. To generate 5' promoter deletions pC2L-2998 was (1) digested with KpnI/NotI, T4-polymerase treated, and religated to yield pC2L-1827, (2) digested with SmaI and religated to yield pC2L-989, and (3) digested with NotI and SalI (partial), Klenow polymerase treated, and religated to yield pC2L-342.
For stable transformation the cab11 and cab12 promoters were cloned 5' to the β-glucuronidase gene (uidA). Towards these ends a 1.17 kb HindIII/BamHI fragment was released from pC1L-1097 and cloned into the binary vector pBI101.3 (Clontech, Heidelberg). The resulting plasmid pC1G-1097 harbors a translational fusion between the first 16 amino acids of the cab11 gene and the uidA gene. Similarly, the cab12 promoter was released as a PstI fragment from plamid pC2L-2998, treated with T4-polymerase then digested with BamHI and subcloned into pBI101.3 which was linearised with SalI, ends filled in with Klenow and redigested with BamHI. The resulting plasmid was named pC2G-2998.
Transient and stable gene expression assays
The luciferase expression from plasmids pC1L-1097, pC1L-7126, pC2L-2998, pC2L-1827, pC2L-989, and pC2L-342 were measured in sugar beet leaves after biolistic transformation [37]. For biolistic transformation the PDS-1000/He Particle Delivery System (BioRad, München, Germany) was used. Microcarrier was gold powder type 200-03 (Heraeus, Hanau, Germany) with a diameter of 1.09–2.04 micrometer. The transformation protocol supplied by the manufacturer of the particle delivery system was followed. Equimolar amounts of plamids pC1L-1097 and pC1L-7126 were used. Similarly, equimolar amounts of plamids pC2L-2998, pC2L-1827, pC2L-989, and pC2L-342 were used. To quantify gene expression the transformation control plasmid p70Sruc harboring the luciferase gene from Renilla reniformis under the control of the doupled CaMV 35S promoter was employed as a second reporter gene [38]. For each reporter gene construct three (pC1L-series) or four (pC2L-series) bombardments were made, gene expression strength of both luciferases measured and normalised relative to the luciferase expression of p70Sruc (see below). For each bombardment 13 leaf discs of equal diameter were cut out of sugar beet leaves and preincubated for 6 hours in petri dishes on MS-media containing 0.4 M mannitol at 25°C. The particle delivery conditions were 1550 psi, 9 cm distance and 27 Hg low pressure. After bombardment the petri dishes with the leaf discs were incubated for 16 h at 25°C under constant light. The Photinus and Renilla luciferase activity were measured with the dual-luciferase reporter assay system (Promega, Mannheim, Germany) in a Lumat 9501 luminometer (PE Biosystem) according to the protocol of the supplier.
For the generation of transgenic plants pC1G-1097 and pC2G-2998 were directly transformed into Agrobacterium tumefaciens strain GV2260 [39]. Agrobacterium tumefaciens mediated transformation techniques were performed with the binary T-DNA plasmids on sugar beet (Beta vulgaris, var. VRB) according to [40]. Selection of the transgenic plants was carried out on kanamycin. β-Glucuronidase (GUS) activity in crude leaf extracts was determined as described by Jefferson et al. [41] using 4-methylumbelliferone beta-glucuronide as a substrate. The concentration of the product 4-methylumbelliferone (Mu) was determined with a multiwell fluorescence plate reader (Millipore CytoFluor 2350). Protein content was measured by the method of Bradford (BioRad protein assay kit). Enzyme activity was calculated as pmol Mu × min-1 × mg-1.
Authors' contributions
DS isolated the cDNA and genomic clones, analysed the transcription during different developmental stages and tissues, generated promoter reporter gene constructs, transformed sugar beet, and performed the quantitative reporter gene assays. DUK performed the suppression subtractive hybridization and analysed the homology and tissue specificity of the cDNA fragments. RH identified cis-regulatory elements. RH and DS conceived of the study, and participated in its design and coordination.
Acknowledgements
We thank Maike Baumeister, Iris Grocholl, Jeanette Kurrasch, Dorothea Pralle and Corinna Rohlf for excellent technical assistance, Klaus Schmidt for providing the plasmid p70Sruc, and Frank Breuer for helpful suggestions on the manuscript.
Figures and Tables
Figure 1 Northern blot analysis with cDNA fragments enriched for leaf expressed genes. Ten micrograms total RNA from leaves (L), taproots (R), stems (S), and inflorescences (I) were hybridized with the indicated cDNA fragments (Table 1). The sizes of the hybridizing transcripts are given in kilo base pairs (kb). The RNA gels were stained prior to blotting to confirm equal loading (data not shown).
Figure 2 Organ specific expression of the cab gene in the above ground parts of sugar beet during development. RNA from root, lateral root, hypocotyl, cotyledon, petiole, sink and source leaf of sugar beets grown in the field was isolated 4, 6, 10, 12 16 and 19 weeks after sowing. Ten micrograms total RNA were separated in each lane and hybridized with the SSH fragment L2 (Table 1). The RNA gels were stained prior to blotting to confirm equal loading (data not shown).
Figure 3 The cab gene is encoded by one or two copies in different sugar beet genotypes. Ten micrograms genomic DNA of genotype 4B5421 (lanes 1–3) and genotype 1K0088 (lanes 4–6) were restricted with different enzymes and hybridized with SSH fragment L2 (Table 1). Lane 1: PstI, lane 2: HindIII, lane 3: EcoRI, lane 4: PstI, lane 5: HindIII, lane 6: EcoRI, lane 7: size standard in kilo base pairs.
Figure 4 GUS activity of cab11 and cab12 promoter reporter gene constructs in leaves and roots of transgenic sugar beets. GUS activity was assayed using total protein prepared from the leaf or root. Each bar represents the activity of an individual transformant. The data are presented as the mean of three individual plants per transformant. The error bar describes the standard deviation of the mean. A. Activity of the cab11 promoter in leaves of transgenic sugar beet plants. B. Activity of the cab12 promoter in leaves of transgenic sugar beet plants. C. Comparison of the activity of the cab11 and cab12 promoter in leaves and roots of sugar beet. The GUS activity of the leaf extracts is shown by the left column and the activity of the root extracts by the right column.
Figure 5 Schematic representation of the cab11 and cab 12 promoters with putative G-box elements. Positions of putative G-box like elements are given relative to the translation start site in the cab11 and cab12 promoters. All binding sites except the single one at position -489 in the cab11 promoter are detected in the upper strand.
Table 1 Homology of cDNA fragments enriched for leaf expressed genes.
cDNA-fragment Base pairs Nearest homologa
L1 862 Flaveria pringlei mRNA for glycine hydroxymethyltransferase (371 bp, 84 %) Acc. Z25859; [42]
L2 479 Nicotiana tabacum mRNA for light-harvesting Chl a/b binding protein (337 bp, 86 %) Acc. X82497
L3 424 Spinacia oleracea psaH mRNA for photosystem I reaction centre subunit VI (361 bp, 90 %) Acc. X16858 [43]
L4 508 unknown
L5 587 Spinacia oleraceae ALDCHL mRNA for fructose-1,6-bisphosphate aldolase (495 bp, 87 %) Acc. X66814 [44]
L6 383 Beta vulgaris clone RUB109UNI ribulose bisphosphate carboxylase, small subunit gene, partial sequence (139 bp, 98 %) Acc. AF173667 [45]
L7 451 unknown
L8 221 Arabidopsis thaliana mRNA for hydroxypyruvate reductase, complete cds (203 bp, 83 %) Acc. D85339
L9 520 Spinacia oleracea, ferredoxin-thioredoxin reductase A2 mRNA (151 bp, 82 %) Acc. X77163 [46]
L10 643 unknown
L11 338 Beta vulgaris clone RUB109UNI ribulose bisphosphate carboxylase, small subunit gene, partial sequence (206 bp, 93 %) Acc. AF173667 [45]
L12 532 unknown
L13 265 Spinacia oleraceae rubisco activase mRNA, complete cds (256 bp, 89 %) Acc. J03610 [47]
L14 540 Spinacia oleraceae psaL mRNA for subunit XI of photosystem I reaction center (409 bp, 88 %) Acc. X64445 [48]
aThe genes that gave the highest homology in a BLASTN gene bank comparison are listed. In parenthesis, the contiguous stretch of base pairs showing the indicated percent homology is given.
Table 2 Transient gene expression analysis of two CAB promoters in sugar beet leaves. The plasmids harbor promoter fragments of cab11 (pC1L) and cab12 (pC2L).
Promoter Reporter Gene Constructa Relative expression level in sugar beet leavesb
pC1L-7126 7.0 +/- 3.4c
pC1L-1097 9.6 +/- 4.6
pC2L-2998 8.0 +/- 0.2
pC2L-1827 4.5 +/- 0
pC2L-989 6.4 +/- 2.4
pC2L-342 6.5 +/- 1.9
CaMV 35S 30.2 +/- 16.6
aThe number following the identifier of the promoter relates to fragment length
bThe relative expression level was calculated as:
(Photinus value(construct) - Photinus value (without DNA) / Renilla value (construct) - Renilla value (without DNA)) × 100.
cAverage of at least three assays +/- standard deviation
Table 3 Induction of two cab promoters seven days after illumination of etiolated sugar beet plants. The results for two cab11 (pC1) and two cab12 (pC2) promoter lines are shown.
Experiment 1 Experiment 2
Etiolated leaves, specific activitya Green leaves, specific activity Light induction Etiolated leaves, specific activity Green leaves, specific activity Light induction
non- transgenic 10 0 0 16 45 2,8
C1-121 155 1287 8,3 42 4126 98
C1-122 36 153 4,3 50 5611 112
C2-50 30 181 6 44 5207 118,4
C2-52 n.db n.d n.d 43 4107 95,5
apmol Mu × min-1 × mg-1
bnot determined
==== Refs
Elliott MC Weston GD Cooke DA, Scott RK Biology and physiology of the sugar-beet plant The sugar beet crop: science into practice 1993 London: Chapman and Hall 37 66
Graham MW Craig S Waterhouse PM Expression patterns of vascular-specific promoters RolC and Sh in transgenic potatoes and their use in engineering PLRV-resistant plants Plant Mol Biol 1997 33 729 735 9132064 10.1023/A:1005726918110
Zhang J Van Toai T Huynh L Preiszner J Development of flooding-tolerant Arabidopsis thaliana by autoregulated cytokinin production Mol Breed 2000 6 135 144 10.1023/A:1009694029297
Outchkourov NS Peters J De Jong J Rademakers W Jongsma MA The promoter-terminator of chrysanthemum rbcS1 directs very high expression levels in plants Planta 2003 216 1003 1012 12687368
Jaeger GD Scheffer S Jacobs A Zambre M Zobell O Goossens A Depicker A Angenon G Boosting heterologous protein production in transgenic dicotyledonous seeds using Phaseolus vulgaris regulatory sequences Nat Biotechnol 2002 20 1265 1268 12415287 10.1038/nbt755
Herwig R Schulz B Weisshaar B Hennig S Steinfath M Drungowski M Stahl D Wruck W Menze A O'Brien J Lehrach H Radelof U Construction of a 'unigene' cDNA clone set by oligonucleotide fingerprinting allows access to 25 000 potential sugar beet genes Plant J 2002 32 845 857 12472698 10.1046/j.1365-313X.2002.01457.x
Diatchenko L Lukyanov S Lau YF Siebert PD Suppression subtractive hybridization: a versatile method for identifying differentially expressed genes Methods Enzymol 1999 303 349 380 10349654
Kloos DU Oltmanns H Dock C Stahl D Hehl R Isolation and molecular analysis of six taproot expressed genes from sugar beet J Exp Bot 2002 53 1533 1534 12021303 10.1093/jexbot/53.373.1533
Schwartz E Shen D Aebersold R McGrath JM Pichersky E Green BR Nucleotide sequence and chromosomal location of Cab11 and Cab12, the genes for the fourth polypeptide of the photosystem I light-harvesting antenna (LHCI) FEBS Lett 1991 280 229 234 2013317 10.1016/0014-5793(91)80299-I
Hehl R Wingender E Database-assisted promoter analysis Trends Plant Sci 2001 6 251 255 11378466 10.1016/S1360-1385(01)01954-9
Matys V Fricke E Geffers R Gossling E Haubrock M Hehl R Hornischer K Karas D Kel A E Kel-Margoulis OV Kloos DU Land S Lewicki-Potapov B Michael H Münch R Reuter I Rotert S Saxel H Scheer M Thiele S Wingender E TRANSFAC: transcriptional regulation, from patterns to profiles Nucleic Acids Res 2003 31 374 378 12520026 10.1093/nar/gkg108
Onodera Y Suzuki A Wu CY Washida H Takaiwa F A rice functional transcriptional activator, RISBZ1, responsible for endosperm-specific expression of storage protein genes through GCN4 motif J Biol Chem 2001 276 14139 14152 11133985
de Pater S Katagiri F Kijne J Chua NH bZIP proteins bind to a palindromic sequence without an ACGT core located in a seed-specific element of the pea lectin promoter Plant J 1994 6 133 140 7920707 10.1046/j.1365-313X.1994.6020133.x
Chen W Chao G Singh KB The promoter of a H2O2-inducible, Arabidopsis glutathione S-transferase gene contains closely linked OBF- and OBP1-binding sites Plant J 1996 10 955 966 9011080 10.1046/j.1365-313X.1996.10060955.x
Hobo T Asada M Kowyama Y Hattori T ACGT-containing abscisic acid response element (ABRE) and coupling element 3 (CE3) are functionally equivalent Plant J 1999 19 679 689 10571853 10.1046/j.1365-313x.1999.00565.x
Hobo T Kowyama Y Hattori T A bZIP factor, TRAB1, interacts with VP1 and mediates abscisic acid-induced transcription Proc Natl Acad Sci U S A 1999 96 15348 15353 10611387 10.1073/pnas.96.26.15348
Nakagawa H Ohmiya K Hattori T A rice bZIP protein, designated OSBZ8, is rapidly induced by abscisic acid Plant J 1996 9 217 227 8820608 10.1046/j.1365-313X.1996.09020217.x
Katagiri F Lam E Chua NH Two tobacco DNA-binding proteins with homology to the nuclear factor CREB Nature 1989 340 727 730 2528073 10.1038/340727a0
Niggeweg R Thurow C Kegler C Gatz C Tobacco transcription factor TGA2.2 is the main component of as-1-binding factor ASF-1 and is involved in salicylic acid- and auxin-inducible expression of as-1-containing target promoters J Biol Chem 2000 275 19897 19905 10751419 10.1074/jbc.M909267199
Mikami K Sakamoto A Takase H Tabata T Iwabuchi M Wheat nuclear protein HBP-1 binds to the hexameric sequence in the promoter of various plant genes Nucleic Acids Res 1989 17 9707 9717 2602142
Green BR Pichersky E Kloppstech K Chlorophyll a/b-binding proteins: an extended family Trends Biochem Sci 1991 16 181 186 1882419 10.1016/0968-0004(91)90072-4
Knoetzel J Svendsen I Simpson DJ Identification of the photosystem I antenna polypeptides in barley. Isolation of three pigment-binding antenna complexes Eur J Biochem 1992 206 209 215 1587270
Schmid VH Cammarata KV Bruns BU Schmidt GW In vitro reconstitution of the photosystem I light-harvesting complex LHCI-730: Heterodimerization is required for antenna pigment organization Proc Natl Acad Sci U S A 1997 94 7667 7672 11038558 10.1073/pnas.94.14.7667
Gotor C Romero LC Inouye K Lam E Analysis of three tissue-specific elements from the wheat Cab-1 enhancer Plant J 1993 3 509 518 8220458 10.1046/j.1365-313X.1993.03040509.x
Lübberstedt T Bolle CEH Sopory S Flieger K Herrmann RG Oelmüller R Promoters from genes for plastid proteins possess regions with different sensitivities toward red and blue light Plant Physiol 1994 104 997 1006 8165263 10.1104/pp.104.3.997
Luan S Bogorad L A rice cab gene promoter contains separate cis-acting elements that regulate expression in dicot and monocot plants Plant Cell 1992 4 971 981 1392604 10.1105/tpc.4.8.971
Duffus JE Ruppel EG Cooke DA, Scott RK Diseases The sugar beet crop: science into practice 1993 London: Chapman and Hall 347 427
Benfey PN Ren L Chua NH The CaMV 35S enhancer contains at least two domains which can confer different developmental and tissue-specific expression patterns EMBO J 1989 8 2195 2202
Sanger M Daubert S Goodman RM Characteristics of a strong promoter from figwort mosaic virus: comparison with the analogous 35S promoter from cauliflower mosaic virus and the regulated mannopine synthase promoter Plant Mol Biol 1990 14 433 443 2102823 10.1007/BF00028779
Gatz C Chemical control of gene expression Annu Rev Plant Physiol Plant Mol Biol 1997 48 89 108 15012258 10.1146/annurev.arplant.48.1.89
Logemann J Schell J Willmitzer L Improved method for the isolation of RNA from plant tissue Anal Biochem 1987 163 16 20 2441623 10.1016/0003-2697(87)90086-8
Ausubel FM Brent R Kingston RE Moore DD Seidman JG Smith JA Struhl K Current protocols in molecular biology 1988 New York, Greene and Wiley Interscience
Saghai-Maroof MA Soliman KM Jorgensen RA Allard RW Ribosomal DNA spacer-length polymorphisms in barley: mendelian inheritance, chromosomal location, and population dynamics Proc Natl Acad Sci U S A 1984 81 8014 8018 6096873
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Marck C 'DNA Strider': a 'C' program for the fast analysis of DNA and protein sequences on the Apple Macintosh family of computers Nucleic Acids Res 1988 16 1829 1836 2832831
Church GM Gilbert W Genomic sequencing Proc Natl Acad Sci U S A 1984 81 1991 1995 6326095
Sanford JC The biolistic process Trends Biotechnol 1988 6 299 302 10.1016/0167-7799(88)90023-6
Schmidt K Identifizierung eines Cercospora beticola responsiven Bereichs innerhalb des Kernpromotors des pal-Gens der Zuckerrübe sowie eines daran bindenden, neuen Transkriptionsfaktors 2003 Dissertation, Technische Universität Braunschweig, Germany
An G Binary Ti vectors for plant transformation and promoter analysis Methods Enzymol 1987 153 292 305
Lindsey K Gallois P Eady C Regeneration and transformation of sugar beet by Agrobacterium tumefaciens Plant Tissue Culture Manual B7 1991 Kluwer Academic Publishers 1 13
Jefferson RA Kavanagh TA Bevan MW GUS fusions: β-glucuronidase as a sensitive and versatile gene fusion marker in higher plants EMBO J 1987 6 3901 3907 3327686
Kopriva S Bauwe H Cloning and sequencing of two isoforms of serine hydroxymethyltransferase from Flaveria pringlei (accession nos. Z25859 and Z25860) Plant Physiol 1998 116 1603 9599133 10.1104/pp.116.4.1603
Steppuhn J Hermans J Nechushtai R Herrmann GS Herrmann RG Nucleotide sequences of cDNA clones encoding the entire precursor polypeptide for subunit VI and of the plastome-encoded gene for subunit VII of the photosystem I reaction center from spinach Curr Genet 1989 16 99 108 2688927 10.1007/BF00393402
Pelzer-Reith B Penger A Schnarrenberger C Plant aldolase: cDNA and deduced amino-acid sequences of the chloroplast and cytosol enzyme from spinach Plant Mol Biol 1993 21 331 340 8425060 10.1007/BF00019948
Schneider K Borchardt DC Schäfer-Pregl R Nagl N Glass C Jeppsson A Gebhardt C Salamini F PCR-based cloning and segregation analysis of functional gene homologues in Beta vulgaris Mol Gen Genet 1999 262 515 524 10589840 10.1007/s004380051113
Falkenstein E von Schaewen A Scheibe R Full-length cDNA sequences for both ferredoxin-thioredoxin reductase subunits from spinach (Spinacia oleracea L.) Biochim Biophys Acta 1994 1185 252 254 8167141
Werneke JM Zielinski RE Ogren WL Structure and expression of spinach leaf cDNA encoding ribulosebisphosphate carboxylase/oxygenase activase Proc Natl Acad Sci U S A 1988 85 787 791 3277181
Flieger K Oelmüller R Herrmann RG Isolation and characterization of cDNA clones encoding a 18.8 kDa polypeptide, the product of the gene psaL, associated with photosystem I reaction center from spinach Plant Mol Biol 1993 22 703 709 8343606 10.1007/BF00047411
| 15579211 | PMC539256 | CC BY | 2021-01-04 16:02:56 | no | BMC Biotechnol. 2004 Dec 5; 4:31 | utf-8 | BMC Biotechnol | 2,004 | 10.1186/1472-6750-4-31 | oa_comm |
==== Front
BMC Oral HealthBMC Oral Health1472-6831BioMed Central London 1472-6831-4-41556656710.1186/1472-6831-4-4Research ArticleEndodontic flare-ups: comparison of incidence between single and multiple visit procedures in patients attending a Nigerian teaching hospital Oginni Adeleke O [email protected] Christopher I [email protected] Department of Restorative Dentistry, Faculty of Dentistry, College of Health Sciences, Obafemi Awolowo University, Ile-Ife, Osun State, Nigeria2004 26 11 2004 4 4 4 6 7 2004 26 11 2004 Copyright © 2004 Oginni and Udoye; licensee BioMed Central Ltd.2004Oginni and Udoye; 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
Until recently the most accepted technique of doing root canal treatment stresses multiple visit procedure. Most schools also concentrated upon teaching the multi-visit concept. However, it has now been reported that the procedure of single visit treatment is advocated by at least 70% of schools in all geographical areas. It was therefore the aims of the present study to find the incidence of post-obturation flare-ups following single and multiple visit endodontic treatment procedures, and to establish the relationship between pre-operative and post-obturation pain in patients referred for endodontic therapy in a Nigerian teaching Hospital.
Methods
Data collected included pulp vitality status, the presence or absence of pre-operative, inter-appointment and post-obturation pain. Pain was recorded as none, slight, or moderate/severe. Flare-ups were defined as either patient's report of pain not controlled with over the counter medication or as increasing swelling. The patients were recalled at three specific post-obturation periods, 1st, 7th and 30th day. The presence or absence of pain, or the appropriate degree of pain was recorded for each recall visits and the interval between visits. The compiled data were analysed using chi-square where applicable. P level ≤ 0.05 was taken as significant.
Results
Ten endodontic flare-ups (8.1%) were recorded in the multiple visit group compared to 19 (18.3%) flare-ups for the single visit group, P = 0.02. For both single and multiple visit procedures, there were statistically significant correlations between pre-operative and post-obturation pain (P = 0.002 and P = 0.0004 respectively). Teeth with vital pulps reported the lowest frequency of post-obturation pain (48.8%), while those with nonvital pulps were found to have the highest frequency of post-obturation pain (50.3%), P = 0.9.
Conclusion
The present study reported higher incidences of post-obturation pain and flare-ups following the single visit procedures. However, single visit endodontic therapy has been shown to be a safe and effective alternative to multiple visit treatment, especially in communities where patients default after the first appointment at which pain is relieved.
==== Body
Background
Until recently the most accepted technique of doing endodontic treatment stresses multiple visit procedures. Most schools also concentrated upon teaching the multi-visit concept. However, it has now been reported that the procedure of single visit treatment is advocated by at least 70% of schools in all geographical areas [1].
Some of the problems of root canal treatment are post-obturation pain, inter-appointment pain and swelling. Although these in most cases do not last long, but could be a source of embarrassment to the dentist and annoying for the patient, more so if the tooth was symptomless before the commencement of treatment. Literature review revealed varied opinions on the incidence and severity of post-obturation pain. Some authors reported slightly more post-obturation pain following single visit than with multiple visit procedures [2,3]. Others found no significant differences in the post-obturation pain experienced by patients following single or multiple visit treatment procedures [4]. O'Keefe [4] however proposed a correlation between pretreatment pain and post-obturation discomfort. The rate of endodontic flare-ups was reported to be more following multiple visits than for the single visit [5-7], Imura & Zuolo [7] also reported a positive correlation between flare-ups and multiple appointment, retreatment cases, peri-radicular pain prior to treatment and presence of radioluscent lesions. They reported no correlation between post-obturation flare-ups and the status of the pulp. However, Sim [8] reported a significantly higher incidence of flare-ups in necrotic teeth than in vital teeth (p = 0.01). Fox et al. [9] in their study showed that female patients had more post operative pain than did males. Factors of age, bacteriologic status, tooth position and type of filling material showed no clear effect upon post-operative results.
Endodontic treatment in Nigeria is carried out in the department of restorative dentistry of the four dental schools and in few private dental clinics located in major cities, health centres and general hospitals. Previous report revealed that few cases of root canal treatment were undertaken and that root canal treatment was completed in multiple visits, specifically three visits for about half the teeth treated [10]. Reasons for the reported few cases of root canal treatment included patient's preference for extraction, which is a cheaper option (the cost of root canal treatment is about twice that of extraction). Also because most of the patients had to travel a considerable distance for the treatment, they prefer extraction, which is completed in a single visit (except in complicated cases). However, it has been recently observed that the acceptability of endodontic treatment is on the increase among Nigerian patients, with more people desiring to keep their teeth. Despite the desire, they present for treatment late only after the onset of pain. Also some patient do not come back to complete the treatment after the first appointment at which pain is relieved. Hence more dentists are embracing the single visit procedure particularly in the Teaching Hospitals. It was therefore the aims of the present study to find the incidence of post-obturation flare-ups following single and multiple visit endodontic treatment procedures. Establish the relationship between pre-operative and post-obturation pain. Find the incidence and degree of pain at the 1st, 7th, and 30th post-obturation days, and to compare these results with those reported in previous studies.
Methods
Consenting patients referred to the department of Restorative Dentistry for root canal therapy within a period of twelve months were randomly assigned for either single visit or multiple visit procedures. For the multiple visit procedures, Patients that defaulted after the first appointment (incomplete treatment) were excluded from the study. For each tooth treated, the clinical factors and conditions existing before, during and after the completion of treatment were recorded. This data included pulp vitality status, the presence or absence of pre-operative pain, post-obturation flare-ups and degree of post-obturation pain. For patients requiring root canal treatment on more than one tooth, the treatment of each tooth was separated by a period of at least four weeks to allow for proper evaluation. The pulp vitality was determined by an electric pulp tester (Parkell pulp vitality tester, Farmingdale, NY 111735) in combination with the presence of pulpal haemorrhage.
The patients were recalled at three specific post-obturation periods, the 1st, 7th and 30th day. At each post-obturation recall visit, the patients were interviewed to determine whether or not there were symptoms at the present visit and whether or not there had been symptoms during the interval between the present visit and the previous one. The presence or absence of pain, or the appropriate degree of pain was recorded for each recall visit and the interval between visits. Pain was recorded as none, slight, or moderate/severe. Slight pain was defined as any discomfort no mater how brief in duration that did not require medication and that did not impair masticatory function in any way. Moderate/severe pain was defined as pain requiring medication or other palliative treatment. Impairment of masticatory function (discomfort in chewing) was recorded as moderate/severe pain. Endodontic flare-ups were defined as either patient's report of pain not controlled with over the counter medication and or increasing swelling.
The root canals were obturated with multiple gutta-percha cones and a zinc oxide-eugenol based sealer, using the lateral condensation technique. The compiled data were analysed using chi-square where applicable. P level ≤ 0.05 was taken as significant.
Results
Two hundred and eighty three (283) teeth in 255 patients were treated in all, given a ratio of 1.11 teeth per patient. Of these 56 were excluded from the study due to non-availability of patients at post-obturation recall visit. These exclusions were randomly distributed between treatment groups, with no differential loss to follow-up (25 from the single visit group, 31 from the multiple visit group). The treatment groups were fairly comparable, with similar distribution of tooth types between treatment groups, Table 1. Two hundred and fourty three (243) were available for check-up on the 1st post-obturation day, of these 107 were completed in single visit and 136 were completed in multiple visit. Eighty-six (86) had vital pulps and 157 had nonvital pulp canal contents. Two hundred and twenty seven (227) reported for check-up on the 7th post-obturation day, of these 104 was completed in single visit and 123 completed in multiple visit. Two hundred and twenty two reported for check-up on the 30th post-obturation day, 102 completed in single visit and 120 in multiple visit.
Table 1 Tooth distribution between treatment groups.
Single visit Multiple visit
Tooth types No (%) No (%)
Maxillary incisors 40 (38.5) 49 (39.8)
Maxillary canines 3 (2.9) 3 (2.4)
Maxillary premolars 20 (19.2) 22 (17.9)
Maxillary molars 6 (5.8) 9 (7.3)
Mandibular incisors 13 (12.5) 12 (9.8)
Mandibular canines 1 (1.0)* 2 (1.6)
Mandibular premolars 9 (8.6) 13 (10.6)
Mandibular molars 12 (11.5) 13 (10.6)
Total 104 (100.0) 123 (100.0)
*Rounded up percentage
Percentage is based on total number in treatment group.
Each interval between visits and subsequent interview were combined and considered as a single post-obturation period. The highest degree of pain reported in either the interval or at the subsequent interview was recorded as the degree of pain for the specific post-obturation period. Ten flare-ups (8.1%), that is patients presenting with pain not controlled by over the counter medication and or increasing swelling, were recorded in the multiple visit group compared to 19 (18.3%) flare-ups for the single visit group. This shows a significant difference (Mantel Haenszel chi-square = 5.18, p = 0.02), Table 2. Of the 107 teeth whose treatments were completed in single visit 67 had pre-operative pain, out of which 50 (74.6%) reported post-obturation pain. Of the 40 teeth with no pre-operative pain, 8 (20%) had post-obturation pain (x2 = 9.04, p = 0.002). For the multiple visit procedures, 88 teeth presented with pre-operative pain out of which 55 (62.5%) reported post-obturation pain. 48 teeth had no pre-operative pain out of which 6 (12.5%) had post-obturation pain (x2 = 12.5, p = 0.0004). These show that for both single and multiple visit procedures, there were statistically significant correlations between pre-operative and post-obturation pain (Table 3).
Table 2 Incidence of post obturation flare-ups
Group Number in study No flare-ups Flare-ups present
No. (%) No. (%)
Single visit 104 85 (81.7) 19 (18.3)
Multiple visit 123 113 (91.9) 10 (8.1)
Mantel Haenszel Chi square = 5.18, df, = 1, p = 0.02.
Table 3 Relationship between pre-operative pain and pain on 1st post obturation day.
Group No preop. Pain Postob. Pain Preop. Pain Postob. Pain
No. (%) No. (%) No. (%) No. (%)
Single visit(n = 107) 40 (37.4) 8 (20.0) 67 (62.6) 50 (74.6)
x2 = 9.04, p = 0.002
Multi Visit(n = 136) 48 (35.3) 6 (12.5) 88 (64.7) 55 (62.5)
x2 = 12.5, p = 0.0004
Teeth with vital pulps reported the lowest frequency of pain (48.8%), while those with nonvital pulps were found to have the highest frequency of pain (50.3%), Table 4. The difference was however, not statistically significant (p = 0.90).
Table 4 Incidence of pain on 1st post obturation day: Vital and nonvital.
Group Number in Group None Slight Moderate/severe
No (%) No (%) No (%)
Vital 86 44 (51.2) 27 (31.4) 15 (17.4)
Nonvital 157 80 (51.0) 52 (33.1) 25 (15.9)
Incidence of pain x2 = 0.02, df = 1, p = 0.90.
Percentage incidence of pain, vital = 48.8.
Percentage incidence of pain, nonvital = 50.3.
The percentages of single visit patients who exhibited slight post-obturation pain on the 1st and 7th days respectively 35.5% and 16.3% were higher than those in the multiple visit group 30.2% and 9.8%. Chi square test indicated no statistically significant differences (Tables 5 &6). The same trend was recorded for moderate/severe pain on the 1st day post-obturation review. The percentage of patient with moderate/severe pain on the 7th day post-obturation was higher for the multiple visits than the single visit group (Table 6). No post-obturation pain persisted to the 30th day.
Table 5 Comparison of pain on 1st post obturation day: single and multiple visit.
Group Number in study None Slight Moderate/severe
No. (%) No. (%) No. (%)
Single visit 107 49 (45.8) 38 (35.4) 20 (18.7)
Multiple visit 136 75 (55.1) 41 (30.2) 20 (14.7)
Total 243 124 (51.0) 79 (32.5) 40 (16.5)
Incidence of pain: x2 = 1.74, df = 1, p = 0.19, degree of pain: x2 = 2.14, df = 2, p = 0.34.
Table 6 Comparison of pain on 7th post obturationday: single and multiple visits.
Group Number in study None Slight Moderate/severe
No. (%) No. (%) No. (%)
Single visit 104 87 (83.7) 17 (16.3) 0 (0.0)
Multiple visit 123 109 (88.6) 12 (9.8) 2 (1.6)
Total 227 196 (86.3) 29 (12.8) 2 (0.9)
Incidence of pain: x2 = 0.79, df = 1, p = 0.37.
Teeth with non-vital pulp recorded more post-obturation pain. There was however no significant difference in post-obturation pain between teeth treated (either by the single or multiple visit procedures) whose pulps were non-vital.
Discussion
Many authorities in the field of endodontics advice against the completion of root canal treatment in single visit in order to prevent post-obturation pain, especially in cases presenting with pre-operative pain [11,12].
In the present study more flare ups occurred in the single visit group (18.3%) than in the multiple visit group (8.1%), showing a disadvantage for single visit treatment at a 95% confidence level, (Table 2). This is in contrast with the findings of Eleazer & Eleazer [6] who reported fewer flare-ups for the single visit group (3.0%) and (8.0%) for the multiple visit group. Other studies also have reported lower incidence figures for endodontic flare-ups [7,13], Walton & Fouad [13] in the United States of America reported an incidence of 3.17%, while Imura & Zuolo [7] in Brazil reported a further lower figure of 1.58%. In Nigeria and possibly in most developing nations, patients do not present for treatment before the onset of severe pain. In most cases they would have tried self prescribed analgesics. These may explain the high incidence of flare-ups reported in the present study, since endodontic flare-ups have been reported to be positively correlated with more severe presenting symptoms and in patients on analgesics [7,13].
Previous studies have shown that there is a strong positive correlation between pre-operative and post-obturation pain [4,14,15]. The present study supports this correlation, in both the single and multiple visit groups there were statistically significant correlation between pre-operative and post-obturation pain, p = 0.002, p = 0.0004 respectively (Table 3). No significant correlation was found between pulp vitality and the reported incidence of post-obturation pain (p = 0.9), Table 4. This finding is in agreement with those of Roan, Dryden & Grimes [16], and Fox et al [9], who reported that whether a tooth pulp was vital or not had little effect on post-obturation pain. It is however in direct conflict with the traditional belief that only vital cases should be considered for single visit endodontics. Although the single visit patients seemed to experience more pain (slight, moderate/severe) than did the multiple visit patients during the first 24 hours, the differences were not statistically significant (Table 5). This finding is supported by that of Pekruhn [17] who also reported no statistically significant difference between the two groups when the total number of pain days was considered, but in contrast with the findings of Roan, Dryden & Grimes [16] that discloses a significant difference in the incidence of post-obturation pain between single and multiple visit.
Despite the high percentages of post-obturation pain reported on the first post-obturation day in both groups (Table 5), seven days after obturation, 83.7% and 88.6% of teeth treated by the single and multiple visit respectively were free of symptoms (Table 6). Also since no post-obturation pain persisted to the 30th day in both groups, these present a strong indication that practitioners should not overreact to early post-obturation symptoms by immediately initiating endodontic retreatment procedures or extraction of the involved tooth. Apart from the reported higher incidence of flare-ups in the single visit group, the post-obturation pain experienced by the patients in both groups compares favourably well with each other. Therefore the higher incidence should not be taken as condemnation for the single visit endodontic therapy, it should however stress the fact that a thorough understanding of the basic endodontic principles is important in considering each case on an individual basis before making a decision as to whether or not it can be completed in one visit [18]. As it is common with all hospital-based studies, the subjects in the present study may not be a true representation of the population. Therefore the ability to generalize the results is weak. However, a careful case selection and adherence to the basic principles of endodontic therapy will reduce the incidence of flare-ups and post-obturation pain and thus enhance a successful outcome.
Conclusions
The present study reported higher incidences of post-obturation pain and flare-ups following the single visit procedures. However, single visit endodontic therapy has been shown to be a safe and effective alternative to multiple visit treatment, especially in communities where patients default after the first appointment at which pain is relieved.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AOO conceived of the study, participated in its design, performed the statistics, and participated in the final write-up of the manuscript. CIU participated in the design, collected the data, drafted the initial manuscript, and participated in the final write-up of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to thank the dentists who worked in the Department of Restorative Dentistry, Obafemi Awolowo University Teaching Hospitals Complex, Ile-Ife, Nigeria during the time of the study.
==== Refs
Qualtrough AJ Whitworth JM Dummer PM Preclinical endodontology: an international comparison Int Endod J 1999 32 406 14 10551115 10.1046/j.1365-2591.1999.00253.x
Clem W Post treatment endodontic pain J Am Dent Assoc 1970 81 1166 70 5272982
Soltanoff WA A comparative study of single visit and multiple visit endodontic procedures J Endod 1978 4 278 81 283193
O' Keefe EM Pain in endodontic therapy: preliminary study J Endod 1976 2 315 19 1068208
Albashaireh ZS Alnegrish AS Postobturation pain after single- and multiple visit endodontic therapy. A prospective study J Dent 1998 26 227 32 9594474 10.1016/S0300-5712(97)00006-7
Eleazer PD Eleazer KR Flare-up rate in pulpally necrotic molars in one-visit versus two-visit endodontic treatment J Endod 1998 24 614 6 9922752
Imura N Zuolo ML Factors associated with endodontic flare-ups: a prospective study Int Endod J 1995 28 261 5 8626209
Sim CK Inter-appoitment emergencies in a Singapore private practice setting: a retrospective study of incidence and cause-related factors Singapore Dent J 1997 22 22 7 10597173
Fox J Atkinson JS Dinin AP Greenfield E Hechtman E Reemen CA Salkind M Todaro CJ Incidence of pain following one visit endodontic treatment Oral Surg 1970 30 123 30 5269799
Akpata ES Endodontic treatment in Nigeria Int Endod J 1984 17 139 51 6381322
Grossman LI Endodontic practice 1974 8 Lea & Febiger, Philadelphia 286
Nicholls E Endodonrics 1984 3 John Wright & Sons Ltd, Bristol 184 6
Walton R Fouad A Endodontic interappointment flare-ups: a prospective study of incidence and related factors J Endod 1992 18 172 7 1402571
Seltzer S Bender IB Ehrenreich J Incidence and duration of pain following endodontic therapy Oral Surg Oral Med Oral Pathol 1961 14 74 82 13749944
Genet JM Wesselink PR Thoden Van Velzen SK The incidence of preoperative and postoperative pain in endodontic therapy Int Endod J 1986 19 221 29 3473042
Roane JB Dryden JA Grimes EW Incidence of postoperative pain after single-and multiple- visit endodontic procedures Oral Surg 1983 55 68 72 6572350
Pekruhn RB Single-visit endodontic therapy: a preliminary clinical study J Am Dent Assoc 1981 103 875 7 6947006
Askenaz PJ One-visit endodontics Dent Clin North Am 1984 28 853 63 6594282
| 15566567 | PMC539257 | CC BY | 2021-01-04 16:29:56 | no | BMC Oral Health. 2004 Nov 26; 4:4 | utf-8 | BMC Oral Health | 2,004 | 10.1186/1472-6831-4-4 | oa_comm |
==== Front
BMC Complement Altern MedBMC Complementary and Alternative Medicine1472-6882BioMed Central London 1472-6882-4-181557596010.1186/1472-6882-4-18Research ArticleFactors associated with herbal use among urban multiethnic primary care patients: a cross-sectional survey Kuo Grace M [email protected] Sarah T [email protected] L Todd [email protected] Rajesh [email protected] Robert J [email protected] Department of Family and Community Medicine, Baylor College of Medicine, Houston, Texas, USA2 Division of General Medicine, University of Michigan and Ann Arbor VA Center for Practice Management and Outcomes Research, Ann Arbor, Michigan, USA3 University of Texas Health Science Center School of Public Health, Houston, Texas, USA2004 2 12 2004 4 18 18 20 3 2004 2 12 2004 Copyright © 2004 Kuo et al; licensee BioMed Central Ltd.2004Kuo 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 use of herbal supplements in the United States has become increasingly popular. The prevalence of herbal use among primary care patients varies in previous studies; the pattern of herbal use among urban racially/ethnically diverse primary care patients has not been widely studied. The primary objectives of this study were to describe the use of herbs by ethnically diverse primary care patients in a large metropolitan area and to examine factors associated with such use. The secondary objective was to investigate perceptions about and patterns of herbal use.
Methods
Data for a cross-sectional survey were collected at primary care practices affiliated with the Southern Primary-care Urban Research Network (SPUR-Net) in Houston, Texas, from September 2002 to March 2003. To participate in the study, patients had to be at least 18 years of age and visiting one of the SPUR-Net clinics for routine, nonacute care. Survey questions were available in both English and Spanish.
Results
A total of 322 patients who had complete information on race/ethnicity were included in the analysis. Overall, 36% of the surveyed patients (n = 322) indicated use of herbs, with wide variability among ethnic groups: 50% of Hispanics, 50% of Asians, 41% of Whites, and 22% of African-Americans. Significant factors associated with an individual's herbal use were ethnicity other than African-American, having an immigrant family history, and reporting herbal use by other family members. About 40% of survey respondents believed that taking prescription medications and herbal medicines together was more effective than taking either alone. One-third of herbal users reported using herbs on a daily basis. More Whites (67%) disclosed their herbal use to their health-care providers than did African-Americans (45%), Hispanics (31%), or Asians (31%).
Conclusions
Racial/ethnic differences in herbal use were apparent among this sample of urban multiethnic adult primary care patients. Associated factors of herbal use were non-African-American ethnicity, immigrant family history, and herbal use among family members. Whereas Hispanics and Asians reported the highest rates of herbal use, they were the least likely to disclose their use to health-care professionals. These findings are important for ensuring medication safety in primary care practices.
==== Body
Background
The use of complementary and alternative medicine (CAM) in the United States gained greater popularity in the 1990s. Two national telephone surveys of 1,539 and 2,005 adults, respectively, demonstrated an increasing trend in the use of CAM, including relaxation techniques, herbal medicine, massage, chiropractic, and acupuncture[1,2]. Specifically, the use of these unconventional treatments rose from 33.8% in 1990 to 42.1% in 1997. These surveys found that use of herbal medicine within the past year increased from 2.5% in 1990 to 12.1% in 1997[2]. CAM use was also found to be more frequent among females, persons 35 to 49 years of age, persons of ethnicities other than African-American, persons who were college educated, and persons whose annual income was greater than $50,000[2]. In a separate study also conducted in the 1990s, the American Botanical Council estimated that one-third of the nation's adults use herbal remedies[3].
Efficacy studies of herbal supplements are on the rise, but most data published to date are preliminary and do not provide strong evidence for the clinical effectiveness of herbs. Nevertheless, about 15 million American adults (18%) are thought to use prescription medications concurrently with herbal or vitamin products[4], and as many as 70% of persons who use herbal remedies do not discuss their use of such remedies with their physicians or pharmacists[1,5-7]. By not communicating about herbal use, they may put themselves at increased risk for adverse drug-herb interactions[8] and make it extremely difficult for health-care professionals to monitor them for such interactions[9]. Likewise, patients do not know what symptoms they should report to their health-care provider that indicate potential adverse effects of drug-herb interactions. Consequently, unintentional medication errors could occur.
The prevalence of herbal use among racially/ethnically diverse primary care patients varies from study to study[2,3,5-7,10-12], ranging from 30%[5,6] to 77%[7]. Since patients must interact with their primary care providers and pharmacists for illnesses to be diagnosed and quality medical care to be provided, a better understanding of variations in herbal use patterns among primary care patients is needed. To this end, we conducted a study with two objectives: 1) to describe the herbal use of ethnically diverse patients in a large metropolitan area and to examine factors associated with herbal use; and 2) to investigate perceptions about and patterns of herbal use among those patients.
Methods
Setting and study population
We implemented this cross-sectional study within the Southern Primary-care Urban Research Network (SPUR-Net) from September 2002 to March 2003. SPUR-Net is a practice-based research network in Houston, Texas, that consists of five constituent member organizations affiliated with a county health system, a managed care organization, or a private practice clinic. SPUR-Net clinicians provide care to patients from diverse ethnic and socioeconomic backgrounds, with approximately one million patient visits per year. A total of six primary care clinics were included in this study that varied according to socioeconomic status (SES) of their patients as measured by income level and insurance type. For the purposes of this study, we defined "clinic SES" according to the insurance status of the majority of patients; "high SES" means that most patients have insurance (i.e., private insurance and/or Medicare), and "low SES" means most patients are indigent (i.e., county health-care coverage and/or Medicaid). Human subject approvals were obtained from the Institutional Review Boards at all of the SPUR-Net constituent organizations. Permission to conduct the study was also obtained from the medical directors and applicable patient advisory groups at each of the six participating clinics.
To be eligible for participation in the study, patients had to be at least 18 years of age and to be visiting one of the participating clinics for routine, nonacute care. A target of 50 surveys in each of the six clinics was collected from a convenience sample of patients. The decision regarding the number of patients to be surveyed was limited by our resources, including availability in funding and personnel. A research assistant approached potential subjects in the clinic setting to determine their willingness to complete a 23-item questionnaire about herbal use in either English or Spanish. Those patients who consented to participate were either given the survey to complete on their own or had the survey administered to them by the research assistant. Research assistants were available on-site to answer any questions the patients had, helping to improve patients' understanding of the terms used in the survey. Recruitment methods were the same in all of the participating clinics. The research assistants stopped recruiting patients when a minimum of 50 surveys was collected in each clinic.
Survey instrument
Survey questions were adopted and modified from previously developed and validated surveys on CAM use, including national telephone surveys conducted by Eisenberg et al.[1,2,13], a family practice survey by Elder et al.[5], a research clinic survey by Johnson et al.[3], and a national mail survey by Astin et al[14]. We modified these questions for use among our multiethnic patient population; we also translated the survey questions into Spanish. The survey instrument was pilot tested with 54 English-speaking subjects and 10 Spanish-speaking subjects before the study. The survey was reviewed by several groups of patient representatives in the community health centers to ensure consistency in responses. For example, some members of a patient advisory group representing a homeless clinic perceived herbal use to be marijuana use; for this reason, we decided not to include this patient population in our study.
The final survey instrument had three components. First, all participating patients answered questions regarding sociodemographic characteristics (e.g., gender, age, race, ethnicity, education, immigrant family history, herbal use by other family members, spoken language other than English, and clinic location). Immigrant family history and spoken language were elicited with the following questions: "Are your family members immigrants to the United States (Y/N)?," and "Do you speak another language other than English?" After completing the demographic questions, respondents answered a series of questions regarding their belief in herbal use and their herbal information sources. The questions pertained to their personal use of herbs (Y/N); their belief in the benefit of herbal remedies (Y/N), the source of their herbal information (physician, pharmacist, family, friends, etc.); their preferred content of herbal information (e.g., effectiveness, side-effects, interactions with other medications), and their preferred methods for obtaining herbal information from physicians or pharmacists (e.g., handout, World Wide Web site, consultation). Patients who reported using herbal supplements answered additional questions related to their patterns of and reasons for herbal use. In open-ended questions, the participating patients were asked about the herbs they specifically used and the health conditions for which they took the herbal products. Related questions included frequency of herbal use (daily, frequently-few times/month, occasionally—few times/year); duration of use (< 1 year, 1–2 years, 3–5 years, > 5 years); expenditure on herbal products; reported concomitant use of prescription medications; disclosure of herbal use to physicians or pharmacists; and any experiences of adverse reactions from using herbs.
For the purposes of this study, we used the definition of dietary supplements stipulated in the 1994 Dietary Supplement and Health Education Act (DSHEA) to differentiate herbs from vitamins and minerals. Herbal use was defined as having ever used herbal products or natural medicines for health maintenance or treatment of health conditions. To measure herbal use, we asked the following question: "Do you use any of the following?" Response options included: herbs/herbal products or natural medicine (e.g., echinacea, St. John's wort, ginseng, ginkgo biloba, soy supplements), folk medicine or home remedy, vitamins, minerals, or none. Herbal use did not include the use of folk medicine, home remedies (such as honey), vitamins, or minerals.
Data analysis
Data from the paper-based survey were entered into an ACCESS database and were imported into SAS 9.1 for Windows. The study variables were summarized by using one-way frequencies to examine the sociodemographic characteristics of the study sample, the belief in and information source for herbal use, and the patterns of and reasons for herbal use among urban multiethnic primary care patients. The frequencies of use of specific herbs were counted, and the health conditions for which herbs were used were further coded into three types—acute, chronic, and health maintenance.
Based on findings from previous studies, we used the following independent variables as reference variables for both the univariate and multivariate logistic regression analyses: male gender, age less than 30 years, African-American ethnicity, less than a college education, no immigrant family history, no herbal use by other family members, and visiting a high SES clinic. A Chi-square test of proportions was used to determine the association between herbal use and each of the independent variables related to demographic characteristics; a p value ≤ 0.05 was considered to be statistically significant. In order to assess factors associated with herbal use, all hypothesized variables (age, gender, race and ethnicity, education, immigrant family history, herbal use by other family members, and clinic clientele stratified by SES) were included in both the univariate and the multivariate logistic regression analyses. These independent variables were entered as dichotomous variables in the model: gender (male vs. female), age (< 50 years, ≥ 50 years), ethnicity (African-American vs. other, including Whites and Hispanics), education (less than college vs. college and greater), immigrant family history (yes vs. no), herbal use by family members (yes vs. no), and clinic clientele (high SES vs. low SES). Significant variables identified by backward elimination of the main effects from the multivariate analysis were further evaluated in two-way interactions. Thus, the final model contained all of the significant main effects and the two-way interaction terms. Odds ratios and 95% confidence intervals were calculated to determine the effects of the significant variables on herbal use. Since the sample size for Asians was small, Asians were not included in the logistic regression analyses. Furthermore, the language variable was excluded from the regression analyses because the survey question was not clearly answered by many patients; for example, 10 Spanish-language forms had "no language other than English" indicated. In addition, some answers were possibly indicative of an exclusive language other than English instead of the bilingual capability of the respondent.
Results
Description of sample
Of the 327 patients who agreed to participate in the survey, only 322 completed the race/ethnicity information and were included in the analysis. The characteristics of the study sample are summarized in Table 1. Two-thirds of the patients were female, and approximately half of all the patients had less than a college education. More than a third (37%) of the patients reported having an immigrant family history, and 50 patients (15%) used the Spanish-language form to complete the survey.
Table 1 Descriptive Characteristics of the Study Sample (n = 322)
Variables White n (%) Hispanic n (%) African American n (%) Asian n (%)
Totals 68(21.1) 98(30.4) 136(42.2) 20(6.2)
Gender
Male 20(29.4) 34(34.7) 37(27.4) 7(35.0)
Female 48(70.6) 64(65.3) 98(72.6) 13(65.0)
Age (yrs)
< 30 13(19.1) 17(17.3) 27(19.8) 4(20.0)
30–49 34(50.0) 34(34.7) 47(34.6) 5(25.0)
50+ 21(30.9) 47(48.0) 62(45.6) 11(55.0)
Education
< High School 3(4.4) 52(53.0) 17(12.6) 0
High School 16(23.5) 23(23.5) 54(40.0) 5(25.0)
≥ College 49(72.1) 23(23.5) 64(47.4) 15(75.0)
Immigrant Family History
No 57(86.4) 41(41.8) 101(77.1) 0
Yes 9(13.6) 57(58.2) 30(22.9) 20(100.0)
Herbal Use by Other Family Members
No 43(63.2) 42(42.9) 95(69.9) 11(55)
Yes 25(36.8) 56(57.1) 41(30.1) 9(45)
Clinic Type
High SES Clinic 40(58.8) 14(14.3) 52(38.2) 9(45)
Low SES Clinic 28(41.2) 84(85.7) 84(61.8) 11(55)
Herbal use
Overall, 36% of our study sample reported ever using herbs. The proportions of herbal users varied across racial/ethnic groups, with use being reported by 50% of Hispanics, 50% of Asians, 41% of Whites, and 22% of African-Americans. Herbal use by other family members was reported to be 41% (57% among Hispanics, 45% among Asians, 37% among Whites, and 30% among African-Americans). Patients who reported using herbs indicated that they received information about those herbs mainly from family members and relatives. Nevertheless, most patients reported that they preferred receiving herbal information (e.g., on effectiveness, side-effects, and drug interactions) through handouts or brochures from their physicians or pharmacists, followed by having access to a consultation service or a Web site. About 40% of all of the survey respondents, but especially Asians (55%) and Whites (47%), believed that taking prescription medications and herbal medicines together was more effective than taking either alone. About 41% of Hispanic respondents believed that herbal medicines were superior to prescription medications, as compared to 12% of Whites. These differences in beliefs about herbal use among the ethnic groups were found to be statistically significant (p < 0.05). Nearly half of the patients who reported using herbs (46%), particularly Hispanics (63%) and Asians (57%), also reported taking prescription medications concomitantly with the herbs (Table 2). Since our survey question was designed to measure self-reported concomitant herbal use and prescription drug use, we cannot confirm whether or not those who reported taking both were actually using both.
Table 2 Patterns of and Reasons for Herbal Use Among Urban Multiethnic Primary Care Patients (n = 322)
Variables White n (%) Hispanic n (%) African American n (%) Asian n (%)
Herbal Use 28(41.2) 49(50.0) 30(22.1) 10(50.0)
Daily Herbal Use* 14(48.3) 13(22.8) 13(33.3) 4(30.8)
Herbal Use 3+ Years* 12(41.1) 45(78.9) 17(45.9) 7(53.8)
Report Taking Herbs and Prescription Medications for the Same Health Problems*
10(33.3) 36(63.2) 17(32.7) 8(57.1)
Told Physicians/Pharmacists About Herbal Use*
20(66.7) 17(30.9) 21(44.7) 4(30.8)
Had a Bad Reaction* 2(7.4) 1(2.0) 3(11.1) 0
Believed Both Prescription Medications and Herbal Medicines Are Better Than Either Alone**
Agree 32(47.1) 28(28.6) 54(40.6) 11(55.0)
Disagree 14(20.6) 44(44.9) 48(36.1) 6(30.0)
Neither 22(32.4) 26(26.5) 31(23.3) 3(15)
Believed Herbal Medicines Are Superior to Prescription Medications***
Agree 8(12.3) 40(41.2) 31(23.1) 6(30.0)
Disagree 36(55.4) 20(20.6) 62(46.3) 8(40.0)
Neither 21(32.3) 37(38.1) 41(30.6) 6(30.0)
Received Herbal Information (multiple)
Family or relatives 20(29.4) 60(61.2) 43(31.6) 10(50.0)
Magazines 24(35.3) 19(19.4) 38(27.9) 5(25.0)
Television 13(19.1) 24(24.5) 45(33.1) 0
Internet 12(17.7) 7(7.1) 9(6.6) 4(20)
Physician 10(14.7) 8(8.2) 12(8.8) 2(10.0)
Pharmacist 2(2.9) 2(2.0) 6(4.4) 0
Preferred Herbal Information (multiple)
Effectiveness 53(77.9) 72(73.5) 87(64.0) 9(45.0)
Side-effects 42(61.8) 76(77.6) 82(60.3) 12(60.0)
Interactions 46(67.7) 67(68.4) 75(55.2) 9(45.0)
Preferred Method for Obtaining Herbal Information (multiple)
Handout/Brochure 45(66.2) 80(81.6) 84(61.8) 11(55.0)
Website 25(36.8) 11(11.2) 20(14.7) 6(30.0)
Consultation 29(42.7) 20(20.4) 47(34.6) 5(25.0)
*Indicates only those patients who reported herbal use
**p = 0.008;*** p < 0.0001
Factors associated with herbal use
Variables demonstrating a significant univariate association (p < 0.05) with herbal use were ethnicities other than African-American, immigrant family history, and herbal use by other family members (Table 3). In the multivariate logistic regression model, non-African-American race/ethnicity (OR = 2.42, 95% CI, 1.33–4.40), immigrant family history (OR = 2.23, 95% CI, 1.20–4.14), and reported herbal use by other family members (OR = 7.98, 95% CI, 4.48–14.18) remained significant predictors of reported herbal use (p < 0.05). In addition, interactions between immigrant family history and herbal use by other family members were found to be significant terms in the model (Table 3). With the race/ethnicity variable adjusted, having an immigrant family history was associated with a 19 times greater likelihood of herbal use among those whose family members also use herbs. When the analyses were run with the Asian group included, the results did not change.
Table 3 Univariate Analysis of Factors Associated with Herbal Use Among Urban Multiethnic Primary Care Patients (n = 302)
Variables Total Herbal Use n (%) X2 p-value
Gender
Male 91 32(35.2) 0.9
Female 210 75(35.7)
Age (yrs)
< 30 106 34(32.1) 0.4
≥ 30 196 73(37.2)
Race/Ethnicity
African-American 136 30(22.1) <0.0001
White & Hispanic 166 77(46.4)
Education
< College 165 57(34.6) 0.8
≥ College 136 49(36.0)
Immigrant Family History
No 199 56(28.1) 0.0001
Yes 96 49(51.0)
Herbal Use by Other Family Members
No 180 31(17.2) <0.0001
Yes 122 76(62.3)
Clinic Type
High SES Clinic 106 35(33.0) 0.5
Low SES Clinic 196 72(36.7)
Perceptions about and patterns of herbal use
The reasons given by the study subjects for herbal use included faster resolution of symptoms (47%), the desire to try alternative therapies (33%), and preference for having their own methods to care for their health (20%). Among the herbal users, 32% reported taking herbs on a daily basis, and 60% reported using herbs for longer than three years. Usage varied by race/ethnicity; for example, 48% of Whites reported taking herbs on a daily basis, and 79% of Hispanics reported using herbs for longer than three years.
Even though Hispanics and Asians used herbs more frequently, they were the least likely to disclose their herbal use to their physicians or pharmacists. More Whites (67%) told their health-care professionals about their herbal use than did the African-Americans (45%), Hispanics (31%), or Asians (31%). The reasons given for nondisclosure generally fell into two main categories: 1) "They (the provider) never asked," and 2) "It wasn't important for them to know." While few respondents (5.3%) reported having experienced an adverse reaction to herbs, many of them (43%) did not inform their physicians of it.
The specific herbs used by the patients covered a wide spectrum and varied by ethnicity. The herbs used most commonly by White patients were echinacea (32.1%), St. John's wort (21.4%), ginkgo biloba (14.3%), and chamomile (14.3%). Hispanic patients most often reported using chamomile (61.2%), aloe vera (44.9%), and garlic (20.4%). African-American patients reported primarily using garlic (40%), ginseng (30%), and ginkgo biloba (10%). The herbs used by Asian patients were garlic (50%), ginkgo biloba (30%), and ginger (30%). Other herbs that were reported by patients—albeit infrequently—included Yun Zhi, black cohosh, dong quai, guggle phosphate, bee pollen, cat claws, and "a shot of whiskey." The patients who reported using herbs used them for a wide range of health problems, such as boosting the immune system, improving memory, and treating insomnia, depression, or diabetes. For conditions considered to be chronic, 44% of the White patients reported herbal use versus 32% of African-American patients. For conditions considered to be acute, 71% of Hispanic patients used herbs versus 10% of Asians. For health maintenance, 50% of Asian patients used herbs versus 16% of Hispanic patients.
Discussion
Our data show that herbal use is common (36%) among urban multiethnic primary care patients, but has a wide variability among racial/ethnic groups. Hispanics and Asians reported the highest rates of herbal use (50%), and African Americans reported the lowest (22%). Previous research conducted in the western United States found that the prevalence of herbal use among racially/ethnically diverse primary care patients varies[2,3,5-7,10-12], ranging from 30% among primary care patients residing in urban settings on the west coast of the United States[5,6] to 77% among primary care patients residing in the largest United States—Mexico border city[7].
As expected, factors associated with herbal use included race/ethnicity, having an immigrant family history, and herbal use by other family members. In addition, we found interactions between having an immigrant family history and herbal use by other family members. Previous studies did not examine such interactions and found age to be predictive of herbal use[2,6,7]. Unlike other investigators, we did not find a significant relationship between age and herbal use. Other investigators, however, did not account for interactions such as those addressed in our analysis. One study (n = 113) found no significant differences in the use of CAM therapies that could be attributable to gender, educational level, age, race, or clinic attended[5]. Another study (n = 542) found an association between the use of CAM therapies, high education level, and female gender[6]. In addition, a recent study conducted in a large United States—Mexico border city revealed that 77% of the residents surveyed (n = 547) use all modalities of CAM therapies and that such use was associated with a high education level[7]. When the residents reported specifically using herbal and home remedies (59%), however, herbal use was found to be associated with a low education level[7].
We found that nearly half of herbal users (46%) reported taking herbal medicines and prescription medications concomitantly. More importantly, 43% of herbal users reported not disclosing their herbal use to their physicians or pharmacists. Interestingly, Hispanics and Asians used herbs the most frequently but disclosed their herbal use to their physicians or pharmacists less often than did Whites and African Americans. This lack of communication about herbal use is an area of concern because of the potential for medication errors and untoward reactions to herb-drug interactions. Adverse drug-herb interactions pose a great danger for patients. For example, ginkgo biloba, garlic, and ginseng all may interact with Coumadin® (warfarin sodium) and cause an increase in bleeding time[15,16]. Echinacea, an immunostimulant, can counteract the action of the immunosuppressants (e.g., the corticosteroids prednisone, methotrexate, and cyclosporine) used to treat immune disorders[17,18]. The interaction between St. John's wort and cyclosporine—which is used to treat rheumatoid arthritis and psoriasis and to prevent the rejection of a transplanted organ—could result in decreased availability of cyclosporine and, consequently, to the worsening of arthritis or psoriasis or the rejection of a transplanted organ [19-23]. St. John's wort may also interact with antidepressants, such as monoamine oxidase inhibitors (e.g., Nardil®, Parnate®) and potentiate the effects of selective serotonin reuptake inhibitors (e.g., Paxil®, Prozac®, Zoloft®)[24]. Moreover, drug-herb interactions might adversely affect the monitoring of certain drug therapies and might even cause life-threatening complications. For example, ginseng, hawthorn, licorice, kyushin, plantain, and uzara root have the potential to interfere with the monitoring of Lanoxin® (digoxin)[25]. In addition, kava has been associated with hepatitis[26] and has resulted in coma when used with Xanax® (alprazolam)[27]. As these detrimental effects have been realized, concern about the increased use of herbal supplements has grown[2,28-33].
Two-thirds of the patients we surveyed reported wanting to receive information on herbal medicines from their physicians or pharmacists, preferably in the form of a handout or a brochure. These findings suggest that future studies are warranted to develop and test educational materials to 1) deepen our understanding of racial/ethnic variation in herbal use among primary care patients; 2) educate health-care professionals about the variations in the use patterns and the rationales for use that may help to reduce medication errors and increase the quality and safety of medical care; and 3) educate patients regarding evidence-based herbal information and encourage patients to communicate their herbal use to their physicians/pharmacists.
Our study results should be interpreted in the context of several limitations. First, our estimates of herbal use frequency are imprecise because we used a convenience sample instead of identifying patients by randomized sampling. Secondly, even though we adopted the DSHEA definition of herbs, some patients had difficulty understanding this definition. Specifically, a small group of patients thought that herbs were equivalent to prescription medications such as digoxin and aspirin; the patients' level of understanding of herbs was improved after the research assistants provided further explanation and clarification. Third, we discovered that asking questions, such as "What do you take when you run out of your medications?," was more effective in eliciting answers from the study subjects than when asking them, "Do you use herbs, herbal products or natural medicine?." For these reasons, we had research assistants on-site to help facilitate the survey process. Fourth, the patients surveyed reported their concomitant use of herbs with prescribed medications based on their perceptions and memories. Last, we did not include measures of quality of life or questions about patient satisfaction with herbal use, which would be helpful in future studies, especially when comparing multiethnic and socioeconomically diverse patient groups.
Conclusions
Despite these limitations, our findings confirm the increasing frequency of herbal use as reported in previous studies. Our study also gives a unique perspective by focusing on factors associated with reported herbal use among an urban multiethnic primary care patient population. In particular, we found that patients with immigrant family history—especially those with family members who use herbs—are most likely to report herbal use. Perhaps most disconcerting was our finding that while an increasing number of primary care patients report taking herbal medicines concomitantly with prescription medications, many of them do not disclose their herbal use to their physicians or pharmacists. These findings suggest that primary care clinicians need to understand the extent and patterns of herbal use by their multiethnic patients and efforts to elicit information from patients about herbal use may be warranted. Future studies are needed to develop effective interventions for primary care health-care professionals and patients to improve medication safety by eliminating potential adverse herb-drug interactions and medication errors.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GMK conceived of the study, designed the survey questionnaires, coordinated and managed the data-collection process, directed data analysis, and drafted the manuscript. STH participated in drafting the manuscript and helped with data analysis. LTW performed the statistical analysis and participated in drafting the manuscript. RB helped with data analysis. RJV reviewed the questionnaires and data analysis, and participated in drafting the manuscript. All authors read and approved the final manuscript.
Table 4 Multiple Logistic Regression Analysis of Factors Associated with Herbal Use Among Urban Multiethnic Primary Care Patients (n = 302)
Variable OR 95% CI
Main Effects
Gender
Male 1.00
Female 1.12 (0.60–2.11)
Age (yrs)
< 30 1.00
≥ 30 1.37 (0.74–2.55)
Race/Ethnicity
African-American 1.00
White & Hispanic 2.42* (1.33–4.40)
Education
< College 1.00
≥ College 1.11 (0.58–2.15)
Immigrant Family History
No 1.00
Yes 2.23* (1.20–4.14)
Herbal Use by Other Family Member
No 1.00
Yes 7.98* (4.48–14.18)
Clinic Type
High SES Clinic 1.00
Low SES Clinic 0.80 (0.40–1.60)
Interactions
Immigrant Family History * Herbal Use by Other Family Members
19.39 (8.11–46.38)
*p < 0.05
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgments
This project was supported in part by grants P20 HS11187 from the Agency for Healthcare Research and Quality and grant D12 HP00042 from Bureau of Health Professions of the Health Resources and Services Administration, which provided infrastructure support for the Southern Primary-care Urban Research Network (SPUR-Net).
The authors wish to acknowledge the following clinics for their participation in this study: Baylor Family Medicine, Casa de Amigos Community Health Center (CHC), Gulfgate CHC, Martin Luther King CHC, People's CHC, and Kelsey Seybold Clinic—Main Campus. We appreciate the support from members of the SPUR-Net Executive Committee. We thank the following individuals, Carlos Vallbona, MD, Thomas Gavagan, MD, MPH, and Anthony Greisinger, PhD, for helping us facilitate project approval processes at community health centers and the Kelsey Seybold Clinic. We also thank the efforts of the following individuals from the Department of Family and Community Medicine at Baylor College of Medicine: research assistants Joanne Wei, Nancy Cheak, Jana Davis, and Armandina Garza for administering the surveys at the participating clinics; Cai Wu and Carol Mansyur for their computer information systems support; and Pamela Paradis Tice, ELS(D) for her editorial assistance.
==== Refs
Eisenberg DM Kessler RC Foster C Norlock FE Calkins DR Delbanco TL Unconventional medicine in the United States: prevalence, costs, and patterns of use N Engl J Med 1993 328 246 252 8418405 10.1056/NEJM199301283280406
Eisenberg DM Davis RB Ettner SL Appel S Wilkey S Van Rompay M Kessler RC Trends in alternative medicine use in the United States, 1990–1997: results of a follow-up national survey JAMA 1998 280 1569 1575 9820257 10.1001/jama.280.18.1569
Johnston BA One-third of nation's adults use herbal remedies Herbalgram 1997 40 49
Smolinske SC Dietary supplement-drug interactions J Am Med Womens Assoc 1999 54 191 192 195 10531760
Elder NC Gillcrist A Minz R Use of alternative health care by family practice patients Arch Fam Med 1997 6 181 184 9075455 10.1001/archfami.6.2.181
Palinkas LA Kabongo ML The use of complementary and alternative medicine by primary care patients. A SURF*NET study J Fam Pract 2000 49 1121 1130 11132062
Rivera JO Ortiz M Lawson ME Verma KM Evaluation of the use of complementary and alternative medicine in the largest United States-Mexico border city Pharmacotherapy 2002 22 256 264 11837563 10.1592/phco.22.3.256.33543
Sorensen JM Herb-drug, food-drug, nutrient-drug, and drug-drug interactions: mechanisms involved and their medical implications J Altern Complement Med 2002 8 293 308 12165187 10.1089/10755530260127989
Bush M Visser A Complementary and alternative medicine: whose responsibility? Patient Educ Couns 2004 53 1 3 15062897 10.1016/j.pec.2004.03.001
Zeilmann CA Dole EJ Skipper BJ McCabe M Dog TL Rhyne RL Use of herbal medicine by elderly Hispanic and non-Hispanic white patients Pharmacotherapy 2003 23 526 532 12680482 10.1592/phco.23.4.526.32117
Dole EJ Rhyne RL Zeilmann CA Skipper BJ McCabe ML Dog TL The influence of ethnicity on use of herbal remedies in elderly Hispanics and non-Hispanic whites J Am Pharm Assoc (Wash) 2000 40 359 365 10853536
Giveon SM Liberman N Klang S Kahan E Are people who use "natural drugs" aware of their potentially harmful side effects and reporting to family physician? Patient Educ Couns 2004 53 5 11 15062898 10.1016/S0738-3991(03)00241-6
Eisenberg DM Kessler RC Van Rompay MI Kaptchuk TJ Wilkey SA Appel S Davis RB Perceptions about complementary therapies relative to conventional therapies among adults who use both: results from a national survey Ann Intern Med 2001 135 344 351 11529698
Astin JA Why patients use alternative medicine: results of a national study JAMA 1998 279 1548 1553 9605899 10.1001/jama.279.19.1548
Miller LG Herbal medicinals: selected clinical considerations focusing on known or potential drug-herb interactions Arch Intern Med 1998 158 2200 2211 9818800 10.1001/archinte.158.20.2200
Norred CL Brinker F Potential coagulation effects of preoperative complementary and alternative medicines Altern Ther Health Med 2001 7 58 67 11712472
Parnham MJ Benefit-risk assessment of the squeezed sap of the purple coneflower (Echinacea purpurea) for long-term oral immunostimulation Phytomedicine 1996 3 95 102
Wagner H Suppner H Schafer W Zenk M Immunologically active polysaccharides of Echinacea purpurea cell cultures Phytochemistry 1988 27 119 126 10.1016/0031-9422(88)80601-0
Barone GW Gurley BJ Ketel BL Abul-Ezz SR Herbal supplements: a potential for drug interactions in transplant recipients Transplantation 2001 71 239 241 11213066 10.1097/00007890-200101270-00012
Karliova M Treichel U Malago M Frilling A Gerken G Broelsch CE Interaction of Hypericum perforatum (St. John's wort) with cyclosporin A metabolism in a patient after liver transplantation J Hepatol 2000 33 853 855 11097498 10.1016/S0168-8278(00)80321-9
Barone GW Gurley BJ Ketel BL Lightfoot ML Abul-Ezz SR Drug interaction between St. John's wort and cyclosporine Ann Pharmacother 2000 34 1013 1016 10981246 10.1345/aph.10088
Breidenbach T Hoffmann MW Becker T Schlitt H Klempnauer J Drug interaction of St John's wort with cyclosporin Lancet 2000 355 1912 10866469
Ruschitzka F Meier PJ Turina M Luscher TF Noll G Acute heart transplant rejection due to Saint John's wort Lancet 2000 355 548 549 10683008 10.1016/S0140-6736(99)05467-7
Linde K Ramirez G Mulrow CD Pauls A Weidenhammer W Melchart D St. John's wort for depression: an overview and meta-analysis of randomized clinical trials BMJ 1996 313 253 258 8704532
Blumenthal M Industry alert: plantain adulterated with digitalis HerbalGram 1997 40 28 29
Escher M Desmeules J Giostra E Mentha G Drug points: hepatitis associated with kava, a herbal remedy for anxiety BMJ 2001 322 139 11159570 10.1136/bmj.322.7279.139
Almeida JC Grimsley EW Coma from the health food store: interaction between kava and alprazolam Ann Intern Med 1996 125 940 941 8967683
Eliason BC Kruger J Mark D Rasmann DN Dietary supplement users: demographics, product use, and medical system interaction J Am Board Fam Pract 1997 10 265 271 9228621
Fugh-Berman A Herbal medicinals: selected clinical considerations, focusing on known or potential drug-herb interactions [letter; comment] Arch Intern Med 1999 159 1957 1958 10493330 10.1001/archinte.159.16.1957
Fugh-Berman A Ernst E Herb-drug interactions: review and assessment of report reliability Br J Clin Pharmacol 2001 52 587 595 11736868 10.1046/j.0306-5251.2001.01469.x
Neal R Report by David M. Eisenberg, M.D., on complementary and alternative medicine in the United States: overview and patterns of use J Altern Complement Med 2001 7 S19 S21 11822631 10.1089/107555301753393760
Kessler RC Davis RB Foster DF Van Rompay MI Walters EE Wilkey SA Kaptchuk TJ Eisenberg DM Long-term trends in the use of complementary and alternative medical therapies in the United States Ann Intern Med 2001 135 262 268 11511141
Ernst E Harmless herbs? A review of the recent literature Am J Med 1998 104 170 178 9528737 10.1016/S0002-9343(97)00397-5
| 15575960 | PMC539258 | CC BY | 2021-01-04 16:31:44 | no | BMC Complement Altern Med. 2004 Dec 2; 4:18 | utf-8 | BMC Complement Altern Med | 2,004 | 10.1186/1472-6882-4-18 | oa_comm |
==== Front
BMC Med EthicsBMC Medical Ethics1472-6939BioMed Central London 1472-6939-5-810.1186/1472-6939-5-8DebateStoicism, the physician, and care of medical outliers Papadimos Thomas J [email protected] Department of Anesthesiology, Medical College of Ohio, 3000 Arlington Avenue, Toledo, Ohio 43614, USA2004 9 12 2004 5 8 8 6 9 2004 9 12 2004 Copyright © 2004 Papadimos; licensee BioMed Central Ltd.2004Papadimos; 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
Medical outliers present a medical, psychological, social, and economic challenge to the physicians who care for them. The determinism of Stoic thought is explored as an intellectual basis for the pursuit of a correct mental attitude that will provide aid and comfort to physicians who care for medical outliers, thus fostering continued physician engagement in their care.
Discussion
The Stoic topics of good, the preferable, the morally indifferent, living consistently, and appropriate actions are reviewed. Furthermore, Zeno's cardinal virtues of Justice, Temperance, Bravery, and Wisdom are addressed, as are the Stoic passions of fear, lust, mental pain, and mental pleasure. These concepts must be understood by physicians if they are to comprehend and accept the Stoic view as it relates to having the proper attitude when caring for those with long-term and/or costly illnesses.
Summary
Practicing physicians, especially those that are hospital based, and most assuredly those practicing critical care medicine, will be emotionally challenged by the medical outlier. A Stoic approach to such a social and psychological burden may be of benefit.
==== Body
Background
Medical outliers are defined in health care reimbursement, especially in the prospective payment system, as those patients who require an unusually long hospital stay or whose stay generates unusually high costs, i.e., the most severely ill [1]. According to Meadow et al., racial and Hispanic minorities are more likely to be outliers, as are urban dwellers, and those who live in counties in the USA that have poverty rates greater than 16.7% [2]. The ranks of medically uninsured in the USA have risen to the highest level since 1998 (45,000,000) while at the same time an additional 1.3 million people have fallen below the poverty line [3].
Financial concerns have overwhelmed the American medical educational system and hospitals [4,5]. Medical outliers have become a significant problem. The US Congress has developed a system of payments through Medicare to protect hospitals from high cost patient stays. Teaching hospitals and smaller public hospitals have a higher percentage of outliers. These hospitals are frequently subjected to patient "dumping" by larger, more powerful hospital systems trying to rid themselves of less lucrative clientele [6]. To make matters worse, teaching and public institutions may have their elective surgery schedules (normally profitable) impacted by the medical outlier problem [7,8].
Eric Cassell has pointed out that medical practice is now being shaped by commercial considerations; the issue of financing health care dominates all facets of medicine, including education, research, the relationships between physicians, and the relationships between physicians and their patients [9]. Nonetheless, "sick people are often cared for by physicians who, though burdened by the system in which they work, are dedicated to the sick and to medicine. Doctors who love their profession and who devote their lives to it are not rare" [9]. However, caring for medical outliers can be a burden for even the most dedicated physicians.
In a previous debate an argument was made, using the works of Kant and Hegel, that physicians have an obligation to treat medical outliers [10]. Here an argument is made using Stoic thought in the pursuit of a corollary to support that maxim (obligatory care of medical outliers by physicians). This corollary, that a physician's mental state (attitude), will determine his comfort in pursuing the above maxim, is supported by the Stoic view of determinism.
Stoics will argue that becoming a physician is predetermined by Nature, or God, and, that having occurred, it must also be predetermined that a physician will care for medical outliers. Thus, the only thing physicians can alter, or control, is their attitude. Zeno, the founder of Stoicism, explains,
"A man's excellence or virtue does not depend on his success in obtaining anything in the external world; it depends entirely on having the right mental attitude toward things" [11].
Stoicism teaches that the universe is rational, that it can be explained rationally, and organized rationally. Stoics taught that logos, the ability of humans to think, plan, and express themselves, was inherent in the cosmos. Therefore, logos is part of Nature, or God,
"God and man are related to each other at the heart of their being rational agents. If a man fully recognizes the implications of this relationship, he will act in a manner which wholly accords with human rationality at its best, the excellence of which is guaranteed by its willing agreement with nature. This is what it is to be wise...." [12].
Stoics believed that a cosmic thread related every event to another and that such a premise allows a human to live a life at one with Nature, or God. In acquiring such an understanding a human could strike an accord between his or her attitudes, actions, and the course of events.
The reason for having a good mental attitude and the understanding that a cosmic thread relates all actions and events, according to the Stoics, is that God rules all and nothing can occur unless God wills it. Accepting everything that happens to one's self will bring contentment,
"So the good man will accept everything, knowing that it is not only unalterable, since Fate determines all, but also the work of God, the perfect being; namely that his happiness depends entirely upon himself, and not at the mercy of other persons or the play of outside forces. What brings happiness is to have the right attitude, to choose the right actions, to aim correctly at the mark" [11].
Therefore, to "try" is within a man's power, and good intentions are indeed enough according to Stoic philosophy. The ability to succeed is not necessarily within man's grasp, but his attitude is his own doing. If a person does his or her best and has no self-reproach for this effort, then he or she is one with Nature, or God.
There are difficulties with accepting Stoic philosophy, which will be reviewed later. Nonetheless, the Stoic view of a physician's approach to the care of the medical outlier will be advocated as one that may allow a physician to be successful in outlier care. Success does not refer to monetary remuneration, but an approach that will allow a physician mental comfort (a good attitude) in such an advocacy.
Physicians must understand the concepts of good, the preferable, the morally indifferent, living consistently, appropriate actions, the cardinal virtues, and the passions to successfully apply the Stoic view.
Discussion
The good, the preferable, and the morally indifferent (in "right actions")
What is good? Any physician pondering this question may come up with answers such as a lucrative medical practice, minimal on-call days, or passing medical board examinations. These things are good, especially when compared to poverty, sickness, or unemployment. However, the Stoics believed what was "good", was also morally perfect (virtue, virtuous acts and virtuous people). Virtue and virtuous things belonged in a league of their own. If you were virtuous, according to the Stoics, you were "good", therefore happy, and this was moral perfection. If you were virtuous you always did what was morally right. Things that are "bad" are morally imperfect (not virtuous); here we speak of evil and wickedness, not poverty, illness, or death.
The Stoics' view of good and bad were extremes of perfection and imperfection. Beauty, wealth, a good job, and a good marriage were things that were preferable, but not morally "good". Illness, poverty, and death were less preferable, but not morally "bad". However, neither the preferable, nor the less preferable were considered "good" or "evil"; they were considered morally indifferent,
"Goodness, however, and knowledge, although they had value of a unique kind, could not be the only things to have value. Right action (author's italics) is a matter of choice concerned with morally indifferent things – will you look for wealth or accept poverty, marry or remain a bachelor, live or die? – and choice between absolutely different alternatives would not involve knowledge or reason.... Virtue then can consist in the effort to obtain these things that have value and avoid their contraries, and knowledge can be knowledge of what is to be preferred. But since things of this sort are not "good" or "bad", it is of no importance whether one has them or does not have them, as far as goodness is concerned. The good intention is enough; achievement may be impeded by forces outside a man's control" [11].
It can generally be agreed upon that caring for medical outliers, ideally, is a "right action" because we are dealing with sickness and economics, i.e., morally indifferent things that are not preferable. Many times caring for medical outliers is something we must do to keep our position, fulfill our contract, or to avoid a lawsuit. However, if we speak of doing what is right in the virtuous, or morally perfect sense, caring for medical outliers must be more than moral indifference. It must be an act of virtue, but can only be so if the physician is doing it out of the deepest sense of duty. Virtue in this case means having the right mental attitude toward outlier care and understanding that it is more than an ordinary good, it is an action that stands morally in a class of its own.
Living consistently
Philosophers in ancient Greece would inquire, "What is the goal of a perfect life" [11]. The Stoic would answer, "living consistently". It means to live harmoniously because those who live in conflict are unhappy.
Zeno explained that "the single plan by which life should be lived must be a plan formed by correct reason, and this would be one that is natural in the sense that it accords both with man's nature and with universal nature" [11]. In advocating the position of the Stoics it is very important to understand that they believed that man inherently considered the interests of his fellow humans important and accepted whatever difficulties divine providence placed upon him so that the wider plan of nature, or God, could be implemented.
To be happy or content in the care of medical outliers a physician must have this consistency of life and an absence of conflict in this regard. To be able to understand this concept and accept it, the physician must be aware that he or she is a part of the whole. For example, let us say a flower flourishes in a garden; we understand what "to flourish" means. Also, let us say we observe a group of birds, some are healthy and some are not. Therefore, we know what the natural condition, or norm, for a flower and a bird should be. This norm is good. So each thing, flower, or bird has a good that is a universal. There is a universal condition or nature that is appropriate for each thing,
"Eating hay is natural to horses, but not to men. It accords with universal nature that horses should eat hay and that men should speak a language. But the former is inappropriate to men and the latter to horses. Universal Nature sanctions a norm for particular things – the nature of plants, animals and men – by reference to which they can be said to attain or not to attain their individual ends" [12].
Thus we can understand what is meant by "the part". Man is born and man will die. Man fights the Universal Nature of death, a particular "norm" of humans. In struggling against his role, or his "part" in nature, man may do extraordinary things to keep himself healthy or alive. A man may decide to preserve his home or his child's education by not having (paying for) health insurance. A physician, for example, may decide that a patient needs an organ transplant. The physician will engage this struggle whether or not the patient has health insurance, and the end result may make the patient a medical outlier, physiologically and/or economically. Universal Nature may sanction a norm for particular things, but humans, and especially physicians, often struggle against their role as a "part" and come into conflict with nature.
Each flower, bird, person, or physician is a part of Nature's whole. Contrary events may happen to an individual "part". Birds are hunted and eaten, flowers are cut and put into a vase, and severely ill humans may end up on long-term ventilatory support. The Stoics believed that such events are part of nature's order. Such a view may be contrary to a human observing the whole of nature. However, if the human perspective is removed from an event then, "From the perspective of the whole even such conditions are not unnatural, because all natural events contribute to the universal well-being" [11].
The Stoics combine their views of the part and the whole. They view the whole as perfect and the nature of perfect requires inequities and incompatibilities; nothing that happens to a human is disadvantageous to him or her, nor is it a disadvantage to nature. Nature is perfect, so according to Stoicism, suffering does not occur for its own sake, but "it is necessary to the economy of the whole" [12].
For physicians to live consistently they must understand their place in nature. Stoics would explain to today's physician that their contentment in dealing with medical outliers depends on this understanding of nature. If this is understood and accepted there will be no conflict and "living consistently" will be attainable.
Appropriate actions
Most physicians hope to make the right, or moral, decisions in regard to their actions. Caring for medical outliers is a choice that has to be made by many physicians. Stoics believed it was always appropriate to act virtuously, but acting virtuously can only occur when a man is "perfectly good", and since men are not "perfectly good" then physicians cannot act virtuously, in other words, be morally perfect. However, Stoicism has made a niche for appropriate actions, "to act virtuously is always morally good, and to act faultily is always bad, to act appropriately is not in itself either good or bad in the sense of being morally "good" or "bad" [11]. Even though physicians as humans are not perfect, and thus cannot act virtuously, nonetheless they can make appropriate decisions, take appropriate actions, and "do the right thing".
A physician may care for a medical outlier, but it is not necessarily a morally good action if he or she does it without the complete understanding of why it is the right thing to do. In other words, taking care of a medical outlier is a just action, and thus an appropriately good action, but only if the physician is doing it without duress or not being mandated to attend to the patient (such as being on-call, by contract, or even being shamed into doing it).
Cardinal virtues
Irrational forces plague a man's mind according to Plato, and these forces have to be controlled before a man could have the needed knowledge to act virtuously. The Stoics, however, did not feel this was true. They felt that if a man could be trained to think correctly, then he could learn to act virtuously.
Zeno went on to define four cardinal virtues that were necessary for a man to acquire to be successfully trained to think correctly so that he could act virtuously. He used Plato's work as a basis and he defined the four virtues in terms of the fourth virtue, wisdom. Justice was wisdom concerned with distribution. Temperance (self-control) was wisdom concerned with acquisition. Bravery was wisdom concerned with endurance. Wisdom was defined as "knowledge of what should and should not be done, or knowledge of what is good or bad or neither" [11].
Medical outliers require justice. Physicians must be able to distribute their actions (medical practice) fairly and equitably to all those who are in need of such services. Temperance, or self-control, must be learned or acquired. Physicians have to control their emotions when assigned to a medical outlier, having their patients turn into medical outliers, or seeing other parties' responses to medical outliers. Physicians should allow neither anger, frustration, anxiety, nor fear to overtake them. Enduring the endless days, weeks, or months of caring for a medical outlier certainly requires bravery and stamina. Fielding the unending phone calls and the constant re-tuning of a patient's hemodynamic status can be of marathon proportions. Wisdom is what should or should not be done and what is good or bad must not only apply to the type and amount of medical care, but it should also encompass the virtue of the acts of justice, compassion, and care given to the medical outlier.
The passions
Physicians need emotion. The Stoics did not disagree, but they wished to eliminate passion (pathos) or what many of them called a mental disturbance. The Stoic "passion" is defined as an excessive uncontrollable drive due to an overestimation of the worth of the "indifferent" things (or events) mentioned previously. Nonetheless Stoics taught that to have great affection was indeed desirable, but at the same time one should remain passionless. Animals are driven to an action because of a stimulus, but in man such a stimulus (or impulse) requires the mind to accede to the stimulus. The Stoics found this to be important because they felt it to be a point of distinction between humans and animals. To the Stoics all living animals were compelled to respond to stimuli by their psyche, a mix of fire and air that was responsible for the functions of living animals (they held that the psyche was not immaterial and could be physically damaged). There are times, however, when a man's mind is out of control and his passions become excessive [11].
There were four kinds of passion the Stoics recognized: fear, lust, mental pain, and mental pleasure (as opposed to physical pleasure). The passions were explained by F.H. Sanbach,
"Fear is a contraction of the psyche caused by the belief that something bad is impending. It causes paleness, shivering, and thumping of the heart. But the belief is false: what is feared is not what a Stoic calls "bad", but one of the morally indifferent things, e.g. death, pain, ill-repute. Fear is the result of exaggerating their importance, of believing they will bring real harm, whereas they do not affect man's essential moral being and if they come are to be accepted as part of the great plan of nature. Lust is a longing for something believed to be good, but again is falsely so believed, since the supposed good is morally of the psyche. Mental pain is a contraction of the psyche resulting from the belief, again erroneous, that something bad is present.... Pleasure was defined as an irrational expansion of the psyche caused by the supposed presence of something good.... What is thought to be good is not in fact good, but at the best acceptable" [11].
In many instances the passions do come into play when a physician cares for a medical outlier. Fear affects the physician from several perspectives. Physicians fear for adverse outlier outcomes, not only for the patient's sake and that of the family, but also out of concern for potential litigation, non-reimbursement for services rendered, and long hours incurred in the care of the patient.
In regard to lust (desire), something believed to be good, but falsely so, several points can be made. The Stoics spoke of many types, or species, of lust, anger being foremost among them. This species of lust is very appropriate to discuss in regard to care of the medical outlier. Physicians do get angry and occasionally act out when challenged. Outliers involve a large investment of emotion, time, and a potential loss of income on the part of the physician (he or she could be caring for patients who are less involved and whose medical insurance has expired). Also, hostility toward staff for small deviations from the plan of care may occur more frequently than the staff would like. As the patient's course of illness drags on physicians may have anger for the patient and the family (even though it may be well concealed). Such anger occurs because the patient does not improve or improves too slowly. There may also be anger towards the family for asking too many questions or questioning the plan of care the physician is following. In addition, the family may also want other physicians consulted or more time from their current physician. What is, in fact, happening is that the family is merely trying to exert control over what little they can yet control.
Mental pain or anguish is self evident in medical outlier care. The previous passions of fear and lust (the species of anger) contribute to the burden of mental pain. The species of mental pain that Stoics address include grief and pity, two potentially powerful distorters of judgment for physicians.
Mental pleasure in caring for those that are seriously and/or chronically ill is not as self-evident. Again we do not speak here of physical pleasure. The species of mental pleasure include "pleasure at unexpected 'benefits', pleasure at other people's misfortunes, pleasures caused by deceit and magic" [11]. Physicians do not take pleasure in the misfortunes of their patients, but there may be interplay of this element when dealing with their colleagues in regard to medical outliers. There are times when various physicians have differing views as what should be done in the course of a patient's care. When one physician is "wrong" and another is found to be "right" concerning a particular decision, procedure, diagnosis, or course of therapy, there are instances of gloating, or taking pleasure in a colleague's fall, error, or misperception.
There is no doubt that medical outlier care evokes passions. There is little for a physician to do but his or her best in regard to giving health care. Much is out of the physician's control, and therefore much of Stoic thought is applicable; control the passions, keep a good attitude, have an open mind about plans of care, have an open mind as to who can participate in decisions (patient, family, other health care providers), understand what is a reasonable outcome, and remember that is does not matter who gets credit for good outcomes [13].
Flaws in Stoic thought
If all human events and actions are predetermined how are human freedoms and free will to be addressed? Universal causation is the bedrock of Stoic philosophy. If human attitudes and beliefs are within an individual's power or sphere of influence, is this truly congruent with Stoic determinism?
Robert L. Arrington illustrates the human attitude towards sickness as a foible in Stoic thought [14]. Illness can be a misfortune or an " indifference". The Stoics seem to hint that we should see illness as an "indifference" and a misfortune and then choose. If we apply universal causation in this matter there must be a cause for us to view illness one way or another. Arrington's interpretation of this dilemma in Stoic philosophy is illuminating,
"And if the causes that exist prior to our forming the attitude lead us to perceive the illness as misfortune, it is not possible for us to perceive it as a matter of indifference. If, on the contrary, the causes lead us to assume the attitude of indifference, then it becomes impossible for us to see the illness as misfortune. Either one of the sets of courses or the other must exist, from which it follows that it is either impossible for us to feel misfortune or impossible for us to feel indifference. If one of these options is impossible, the attitude we take is necessary in which case we really didn't have any options at all. And without options or choices, there is no thing as freedom or voluntary behavior. And, so it seems, our attitudes and beliefs are not in our power" [14].
This argument regarding whether universal causation and determinism is consistent with a free will has been debated for over 20 centuries. Today there are philosophers on both sides of the issue.
Another flaw is the Stoic approach to evil. Stoics simply tell us it does not exist; events may seem evil, but they are not. Stoics teach that only the human perspective allows the interpretation that evil exists. Religions of the world, many philosophers, and people who have viewed and/or endured suffering cannot agree with the Stoics.
A further distortion in Stoic thought involves the idea that the life of virtue is the only "good" life. What about the "preferred" things that we as humans know make our lives better? What is wrong with "attaining the goals of impulse" [14]? There was a gradual progression in the evolution of later Stoic philosophy to allow the acceptance of the "preferable" things and this erosion of principle led to many attacks on Stoicism from other philosophical quarters.
And, finally, the Stoics felt the universe was rational and in unity. A divine thread ran through the cosmos connecting everything and everybody. Many philosophers cannot accept this concept. However, as we see the progression of this line of reasoning as it regards the study of the "string" theory in physics and the further work and modification of Einstein's views of relativity, we realize that there may be a mathematical basis to existence. The Stoics may be criticized about their "thread" through the cosmos, but when we discuss how time "bends" and describe gravity as "curved space" the critics of Stoicism may be tightrope-walking this same thread.
The health care rendered to medical outliers can be for a substantial length of time and cost to the medical practitioner. The intensive interaction with the patient, his or her family, consultants, nurses, other ancillary staff, and the institution can sap the performance of involved physicians.
Placing the argument for the predetermination of events aside, if one is a physician that is involved in a hospital-based specialty, critical care medicine, or as surgeon, the fact of the matter is that medical outliers will come to your door. By fate or by choice physicians in the above-mentioned areas will be engaged with outliers. As the Stoics point out, the future is coming at you and there is nothing you can do about it, except adjust your attitude.
While engaged in such an endeavor physicians will "try" to do their best, hopefully without self-reproach as to their efforts. Physicians, as humans, cannot be "good", or morally perfect. They become tired, hungry, worry about the bills, their children, their practice, hospital policies, etc. Nonetheless, most physicians realize and understand what appropriate actions are necessary in regard to the most ill and poorest of patients. Also, though physicians may not be morally perfect (virtuous in the Stoic sense), they know what is "preferable" for their patients, i.e., to get well, go home and be with their families. These are morally "indifferent" things, but as Stoics point out, "virtue then can consist in the effort to obtain these things that have value and avoid their contraries, and knowledge can be knowledge of what is to be preferred" [11].
As mentioned previously, physicians frequently struggle against their role as a "part". By the mere fact that physicians acknowledge struggles with or against insurers, patients, families, colleagues, ancillary staff, and institutions there is realization that they are part of a "whole", but at the same to "live consistently" requires an understanding of one's role in nature and the need for absence of conflict. In the struggle to help the critically ill, the chronically ill, or the incredibly poor, avoiding conflict is challenging.
The cardinal virtues of justice, temperance, bravery, and wisdom come with upbringing, education, and life experiences. Lacking the proper intellectual or nurturing environment will not allow the flourishing of these virtues in an individual. In schools of medicine the faculty attempt to inculcate these virtues in the students, although they are not always successful.
The Accreditation Council for Graduate Medical Education (ACGME) has recently mandated six core competencies for resident education [15]. One of these competencies is Professionalism (and ethics). This, in effect, formalizes ethics education at the graduate level (residency). Such a graduate level discourse should be preceded by a problem-based learning format at the medical school level, preferably before the students begin their clinical work. Thus, the Stoic view, or any other philosophical view or ethical concept, could be taught pre-clinically and then reinforced in an ACGME residency format.
Justice is a far-reaching concept that more physicians need to embrace. It is difficult to teach and is best acquired through experience of its antithesis. Temperance, or self-control, can be mandated by medical staff guidelines and licensing boards, but this virtue actually needs to have been ingrained before medical school. Bravery is something that residency training seeds in a physician through the frequent facing of dying or hostile patients that come through a medical practice, especially at an academic institution that cares for the disenfranchised. Wisdom will come with time and maturity. Both are needed to acquire wisdom. Physicians, medical students, and residents can, indeed, be taught to think correctly, as the Stoics emphasized. However, individuals will have varying degrees of success depending their rearing/nurturing and educational environment.
Emotion is necessary to physicians, but the passions, those uncontrollable mental disturbances due to overestimation of the value of "indifferent" things, may blind them in their judgment and decision-making. It is important that when making important decisions in regard to medical outliers that the passions be "checked".
To avoid frustration, disappointment and unhappiness in the practice of medicine as it regards medical outliers, physicians must do two things: (1) control things that are within their power (attitudes, desires, beliefs), and (2) be indifferent to the things that they cannot control (things external to themselves) [16].
Even though Stoicism has evoked controversy for over twenty centuries it is relevant to a physician who must juggle patients, procedures, therapies, and colleagues in the care of a patient who has maximally taxed medical insurers, institutions, other practitioners, and their own families.
Summary
Insurers and institutions may have financial burdens, but those providing patient care, especially physicians, bear a disproportionate slice of the mental anguish associated with the care of medical outliers. Here an argument has been made that applying the philosophical tenets of Stoicism to the physician's intellectual pursuit of how to deal mentally with the care of medical outliers is appropriate. Physicians that are hospital based and those practicing critical care medicine may well be the providers most emotionally challenged by outliers. A Stoic approach to such a social and psychological burden may be helpful.
Competing interests
The author(s) declare that they have no competing interests.
Author's contributions
TJP is responsible for the manuscript in its entirety.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
none.
==== Refs
Webster Dictionary
Meadow A Wrobel MV Goldber HB Who are outliers? Characteristics of outlier episodes under Medicare's home health prospective payment system [abstract]
US Census
Ludmerer KM Time to Heal: American Medical Education from the Turn of the Century to the Era of Managed Care 1999 New York: Oxford University Press
Risse GB Mending Bodies, Saving Souls: A History of Hospitals 1999 New York: Oxford University Press
California Nurses Association. US Hospitals: Utilizing or Exploiting Medical Outliers
Ashdown DA Williams D Davenport K Kirby RM The impact of medical outliers on elective surgery lists RCS Bulletin 2003 85 46 47
Skowronski GA Bed rationing and allocation in the intensive care unit Curr Opin Crit Care 2001 7 480 484 11805556 10.1097/00075198-200112000-00020
Cassell EJ The Nature of Suffering and the Goals of Medicine 2004 New York: Oxford University Press
Papadimos TJ Marco AP The physician's obligation to medical outliers: A Kantian and Hegelian synthesis BMC Med Ethics 5 E3 2004 Jun 03 15176981 10.1186/1472-6939-5-3
Sanbach FH The Stoics 1989 Indianapolis: Hackett Publishing Company
Long AA Hellenistic Philosophy 1986 London: Gerald Duckworth & Company
Arrington RL Hellenisitic Ethics: Epicurus and the Stoics In Western Ethics 1998 Oxford: Blackwell Publishers Ltd 93 121
Stoicism
Accreditation Council on Graduate Medical Education. Outcome Project
Ferrand E Lemaire F Regnier B Kuteifan K Badet M Asfar P Jabir S Chagnon J Renault A Robert R Pochard F Herve C Brun-Buisson C Duvaldestin P Discrepancies between perceptions by physicians and nursing staff of intensive care unit end-of-life decisions Am J Resp Crit Care Med 2003 167 1310 1315 12738597 10.1164/rccm.200207-752OC
| 15588293 | PMC539259 | CC BY | 2021-01-04 16:31:54 | no | BMC Med Ethics. 2004 Dec 9; 5:8 | utf-8 | BMC Med Ethics | 2,004 | 10.1186/1472-6939-5-8 | oa_comm |
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-4-211558831110.1186/1472-6947-4-21Research ArticleQuantitative evaluation of recall and precision of CAT Crawler, a search engine specialized on retrieval of Critically Appraised Topics Dong Peng [email protected] Ling Ling [email protected] Sarah [email protected] Marie [email protected] Adrian [email protected] Medical and Clinical Informatics Group, Bioinformatics Institute, BMRC, A*STAR, Singapore2004 10 12 2004 4 21 21 20 8 2004 10 12 2004 Copyright © 2004 Dong 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
Critically Appraised Topics (CATs) are a useful tool that helps physicians to make clinical decisions as the healthcare moves towards the practice of Evidence-Based Medicine (EBM). The fast growing World Wide Web has provided a place for physicians to share their appraised topics online, but an increasing amount of time is needed to find a particular topic within such a rich repository.
Methods
A web-based application, namely the CAT Crawler, was developed by Singapore's Bioinformatics Institute to allow physicians to adequately access available appraised topics on the Internet. A meta-search engine, as the core component of the application, finds relevant topics following keyword input. The primary objective of the work presented here is to evaluate the quantity and quality of search results obtained from the meta-search engine of the CAT Crawler by comparing them with those obtained from two individual CAT search engines. From the CAT libraries at these two sites, all possible keywords were extracted using a keyword extractor. Of those common to both libraries, ten were randomly chosen for evaluation. All ten were submitted to the two search engines individually, and through the meta-search engine of the CAT Crawler. Search results were evaluated for relevance both by medical amateurs and professionals, and the respective recall and precision were calculated.
Results
While achieving an identical recall, the meta-search engine showed a precision of 77.26% (±14.45) compared to the individual search engines' 52.65% (±12.0) (p < 0.001).
Conclusion
The results demonstrate the validity of the CAT Crawler meta-search engine approach. The improved precision due to inherent filters underlines the practical usefulness of this tool for clinicians.
==== Body
Background
Healthcare has been steadily moving towards Evidence-Based Medicine (EBM) since the term was formally introduced in 1992 by a group led by Gordon Guyatt at McMaster University, Canada [1-3]. EBM promotes systematic literature review, critical appraisal skills and integrates scientific evidence with clinical expertise in the daily management of patients. The first three steps involved in the practice of EBM can comprehensively be summarized as a one-page written paper on a particular clinical topic, which is most commonly called a 'Critically Appraised Topic' (CAT) [4]. Different acronyms have emerged in various specialties, such as Best Evidence Topics (BET) [5] in emergency medicine and Evidence-Based Journal Club Reviews (EBJCR) [6] in pediatric critical care medicine. All these essentially provide physicians with a systematic method of formulating a clinical question and then critically evaluating the literature to answer the question posed.
With the use of resources on the World Wide Web becoming common practice, several academic and healthcare organizations have built online CAT libraries for knowledge sharing with peer physicians. The repository of CATs has been growing steadily since the setup of the first accessible CATBank developed by the Centre for Evidence Based Medicine, Oxford in 1992 [7]. Among those, BestBETs developed by the Emergency Department, Manchester Royal Infirmary [8] and UMHS by the Department of Pediatric, University of Michigan Health System, Ann Arbor [9] hold hundreds of distinct topics. They are furnished with individual search engines for fast and direct access to a particular topic. Given the wealth of such medical information scattered in cyberspace, the effectiveness of locating the correct information has become an important issue [10].
The CAT Crawler application
It is believed that more CATs will be added into the repositories as more people participate in EBM practice. However, the non-standardized electronic format of CATs has created much difficulty for physicians to access a particular topic. Accordingly, the CAT Crawler was developed at the Bioinformatics Institute, Singapore [11,12] to provide a one-stop search and download site for physicians by setting up a common platform to access eight popular online CAT libraries. CAT Crawler is freely accessible online [12].
The core component of the CAT Crawler is a meta-search engine. Its search is currently based on CAT resources from eight public online libraries [11]. Once the user chooses the libraries he intends to use in the search, information tailored to his needs can be produced. The matched results are sorted according to their origins.
Following the user input of a query keyword, a partial search is done through information extracted during an off-line process from six websites that do not hold search engines.
The remaining search is carried out by querying the two individual search engines at BestBETs and UMHS. Use of the CAT Crawler is expected to have a quantitative and qualitative improvement of the retrieved results by post-processing obtained raw results from both libraries.
Motivation of the evaluation
The work presented here aims to evaluate the quantity and quality of the obtained results from the CAT Crawler meta-search engine, and thus to evaluate the validity and the usefulness of the application. Recall and precision were estimated to measure the performance of this meta-search engine versus the two individual search engines at BestBETs and UMHS.
Methods
The workflow of this study is demonstrated in Figure 1.
Selection of ten query keywords
To find a viable sample of keywords for a test search, the titles of all CATs stored in the two CAT libraries, namely BestBETs and UMHS were submitted to AnalogX Keyword Extractor, which is freely available online [13]. This led to a list of around 2000 keywords, of which approximately 500 were present in both libraries, of which ten were randomly chosen. In a second step, that list was curated so that only medically relevant keywords remained, excluding words such as and and day.
Search for technically relevant documents in the dataset
In order to be able to calculate recall as detailed below, the technical relevance of all documents in the dataset must be assessed. In this study, a document is called technically relevant for a given search term if it contains this term in the full-text. Perl scripts were developed to examine all CATs in the two libraries BestBETs and UMHS and the total number of relevant documents as per the above definition in each library was collected for further calculation. This was done for each selected keyword and the process was independent from the search using the three search engines: the CAT Crawler, BestBETs and UMHS.
Relevance evaluation of the retrieval results
In the next step, those ten keywords were submitted to the search engines at BestBETs and UMHS, and to the CAT Crawler meta-search engine. The retrieved links were evaluated for their relevance by 13 volunteers, who are categorized into three groups. Among them, one physician in Group I represents medical professionals, six persons in Group II represent people who were trained in biology or medicine, and six persons in Group III represent people who do not have any medical background.
Calculation of recall and precision
Recall and precision are two accepted measurements to determine the utility of an information retrieval system or search strategy [14]. They are defined as:
Despite the relevance evaluation from 13 volunteers, it is necessary to know the total number of the relevant documents in a database for each query keyword in order to estimate the recall. In the present study, a particular CAT in a database was defined as technically relevant if the keyword could be found in its full-text article.
The CAT Crawler is designed not to hold permanently any full-text CATs [11]. When a query is done choosing the option to search only BestBETs and UMHS, the total number of relevant document in its acute database is equivalent to the sum of the number of relevant documents in the two libraries BestBETs and UMHS. Accordingly, the recall and precision of the CAT Crawler meta-search engine are revised as:
Similarly, the recall and precision of the search engines at BestBETs and UMHS are estimated based on the combined repository of the two individual sites. The revised formula are shown below:
Performance evaluation of the CAT Crawler versus BestBETs and UMHS
The averaged precision and recall over all evaluators are used to evaluate the performance of the CAT Crawler meta-search engine. These values are compared to the estimate based on the search results from the two individual search engines at BestBETs and UMHS.
Results
Ten keywords for the search engine evaluation
According to the predefined selection criteria, the ten keywords listed in Table 1 were selected as the seed for a test search. The number of retrieved results from each search engine was gathered with respect to each keyword query. For the selected ten medically relevant keywords, the total number of matched results are 116, 65 and132 corresponding to the three search engines at BestBETs, UMHS and CAT Crawler. The difference of 49 retrievals between the CAT Crawler and the sum of BestBETs and UMHS reflects the meta-search engine's inherent filter function which is described previously [11].
Performance evaluation of the CAT Crawler versus BestBETs and UMHS
To compare the performance of the CAT Crawler meta-search engine to that of the two individual search engines, recall and precision were computed and averaged over the evaluation of all 13 participators. The data recorded are shown in Table 2. As the CAT Crawler meta-search engine is built upon the two individual search engines, the document collection for evaluation is the combined repository of BestBETs and UMHS. The retrieved relevant documents from the CAT Crawler are the same as that from the individual search engines. This leads to the identical recall for both cases (Table 2). The average precision is increased from the individual search engines' 52.65% (±12.0) to the CAT Crawler's 77.26% (±14.45). Figure 2 provides a more intuitive comparison corresponding to each keyword.
Discussion
The performance evaluation clearly places the CAT Crawler meta-search engine on par with the individual search engines at BestBETs and UMHS as far as recall is concerned, and well above them for precision (see Table 2 and Figure 2). According to these results, the application can be called successful: by using the CAT Crawler to look for relevant information at specific sites, the medical professional will obtain as much information as by going to the sites directly, but the precision of the obtained results will be higher.
Benoit [15] has analyzed various methods of information retrieval and their impact on user behavior. He finds that users wish for greater interactive opportunities to determine for themselves the potential relevance of documents, and that a parts-of-document approach is preferable for many information retrieval situations. At present, the CAT Crawler allows a number of interactive opportunities [11], but their implementation would have no impact on the calculation of recall and precision under the condition of the present study. Benoit's reasoning should be kept in mind, however, for improving the user friendliness in the sense that some further useful filter functions can be included in future versions of the application. While such advanced search functions will be profitable when large datasets are studied, the currently still manageable information in the online CAT libraries [11] will serve the user better if initially displayed in a broader way. For example, some of the information displayed here may be older than 18 months, which makes it undesirable according to the strict rules for CAT updating as defined by Sackett et al [3]. Formally outdated information, however, may in a given situation still be "best evidence" and positively influence the decision-making. Use of filters to block aged information will certainly influence this process.
Despite the encouraging results, some fundamental questions regarding the evaluation of this meta-search engine in particular, and also meta-search engines in general remain unsolved.
With regard to recall, there is the theoretical possibility that manually searching all documents at a given repository will yield a higher recall for a given search term. In view of hundreds of CAT documents per repository, however, it seems unlikely that a human evaluator's attention will not wander, leading to less than optimal scrutiny of the documents and introducing a non-quantifiable error to the evaluation. This is a general problem of knowledge databases, especially when indexing is done by humans, whose decisions are not consistent. In a study of 700 Medline references indexed in duplicate, the consistency of main subject-heading indexing was only 68% and that for heading-subheading combinations was significantly less [16]. Also, in two studies [17,18] on Medline searching, there was considerable disagreement by those judging relevance of the retrieved documents regarding which documents were relevant to a given query.
In order to overcome this problem, the number of documents that contained a given keyword as found by the keyword extractor was used as the basis for calculating the technical recall. This may (or may not) lead to numerical results for recall that differ from the absolute true value as determined above. As the same numbers are used throughout, however, the comparison of search results obtained by the individual search engines and the CAT Crawler meta-search engine remains valid.
Critics have pointed out the over-reliance of researchers on the use of recall and precision in evaluation studies [18] and the difficulty to design an experiment that allows both laboratory-style control and operational realism [19]. For instance, recall may be of only little consequence once the user has found a useful document. Rhodes and Maes [20] evaluated both with a traditional field user test and then asked for relevance feedback. In their experiment, users gave a score 1–5 to each document that was delivered to calculate an overall average value for perceived precision. While a document can get a high score for precision, it may at the same time get a low score for practical usefulness. This was often due to the fact that the documents were already known to the users, in some cases had even been written by them. Accordingly, Rhodes and Maes [20] added features to the system that weeded out relevant documents that by some predefined criteria would not be useful. As a result, the measurable precision could be worse, but the overall usefulness could be better. In the study presented here, a similar approach was chosen in the instructions to the evaluators in the sense that they could make the distinction between 'irrelevant' (e.g. the retrieved document was only a web hosted clinical question) and 'medically irrelevant' (e.g. the word Appendicitis appeared only in the reference section of a document dealing with questions of abdominal pain relief). Due to the relatively small number, no difference could be detected between the various grades of relevance, and results were pooled to relevant/irrelevant and used for calculating recall and precision as described above. If a larger number of volunteers could be recruited, repetition of this evaluation might yield interesting results.
Other approaches have been spawned to evaluating system effectiveness in order to minimize these problems with recall and precision. One example are task-oriented methods that measure how well the user can perform certain tasks [21-24]. These different approaches were not chosen in this study for a reason: the primary aim was to compare the search engines. Under the present restrictions, recall and precision allow to answer this question.
Conclusions
In summary, the data obtained from the analysis of search results obtained from identical queries submitted to the two CAT libraries at BestBETs and UMHS, using either their respective search engines or the CAT Crawler meta-search engine, showed a competitive recall, and superior precision of the meta-search engine compared to the individual search engines.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
PD participated in the design of the study, data analysis and drafting of the manuscript. LLW and SN generated raw data for the study. ML was involved in drafting the manuscript. AM designed the study and participated in the drafting of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to thank the staff and students of the Bioinformatics Institute for volunteering to evaluate the performance of search.
Figures and Tables
Figure 1 Workflow for evaluation of the CAT Crawler meta-search engine
Figure 2 Precision plot of the CAT Crawler meta-search engine and two individual search engines
Table 1 Ten random keywords and corresponding number of retrieved results from search engine at BestBETs, UMHS and CAT Crawler
Keyword Search Engine
BestBETs UMHS CAT Crawler
Appendicitis 7 3 8
Colic 15 2 9
Intubation 26 5 22
Ketoacidosis 2 2 2
Octreotide 3 2 3
Palsy 6 5 10
Prophylaxis 18 19 30
Sleep 5 13 16
Tape 4 2 3
Ultrasound 30 12 29
116 65 132
Table 2 Numerical recall and precision for the CAT Crawler meta-search engine and two individual search engines at BestBETs and UMHS
Recall (%) Precision (%)
BestBETs & UMHS CAT Crawler BestBETs & UMHS CAT Crawler p-value
Appendicitis 96.15 96.15 76.92 (±4.80) 96.15 (±6.00) 0.000
Colic 54.81 54.81 51.58 (±2.58) 97.44 (±4.87) 0.000
Intubation 44.12 44.12 48.39 (±13.56) 68.18 (±19.10) 0.130
Ketoacidosis 48.72 48.72 36.54 (±12.97) 73.08 (±25.94) 0.001
Octreotide 59.62 59.62 47.69 (±10.13) 79.49 (±16.88) 0.000
Palsy 70.77 70.77 64.34 (±16.37) 70.77 (±18.01) 0.002
Prophylaxis 67.03 67.03 63.41 (±11.60) 78.21 (±14.31) 0.074
Sleep 57.95 57.95 48.29 (±19.82) 54.33 (±22.30) 0.038
Tape 46.15 46.15 46.15 (±7.31) 92.31 (±14.62) 0.000
Ultrasound 42.22 42.22 43.22 (±8.06) 62.60 (±11.68) 0.017
Average 58.75 (±16.25) 58.75 (±16.25) 52.65 (±12.0) 77.26 (±14.45) 0.000
==== Refs
Group EBMW Evidence-Based Medicine. A new approach to teaching the practice of medicine. JAMA 1992 268 2420 2425 1404801 10.1001/jama.268.17.2420
Sackett DL Rosenberg WM Gray JA Haynes RB Richardson WS Evidence based medicine: what it is and what it isn't BMJ 1996 312 71 72 8555924
Sackett DL Straus SE Richardson WS Rosenberg W Haynes RB Evidence-Based Medicine: How to practice and teach EBM 2000 London, Churchill Livingstone
Sauve S Lee HN Meade MO Lang JD Farkouh M Cook DJ Sackett DL The critically appraised topic: a practical approach to learning critical appraisal. Ann Roy Soc Phys Surg Canada 1995 28 396 398
BETs and CATs, Emergency Department, Manchester Royal Infirmary [http://www.bestbets.org/background/betscats.html]
Pediatric Critical Care Medicine, Evidence-Based Journal Club Review [http://pedsccm.wustl.edu/ebjournal_club.html]
CAT Library, Oxford-Centre for Evidence Based Medicine [http://www.minervation.com/cebm2/cats/allcats.html]
BET Database Search Engine, Emergency Department, Manchester Royal Infirmary [http://www.bestbets.org/database/search.html]
CAT Library Search Engine, University of Michigan [http://www.med.umich.edu/pediatrics/ebm/Search.htm]
Bin L Lun KC The retrieval effectiveness of medical information on the web. Int J Med Inf 2001 62 155 163 10.1016/S1386-5056(01)00159-9
Dong P Mondry A Enhanced quality and quantity of retrieval of Critically Appraised Topics using the CAT Crawler Med Inform Internet Med 2004 29 43 55 15204609 10.1080/14639230310001655849
CAT Crawler - an online resource for Critically Appraised Topics (CATs)
AnalogX Keyword Extractor
Hersh WR Detmer WM Frisse ME H SE and E PL Information-Retrieval Systems Medical Informatics 2001 New York, Springer 539 572
Benoit G Properties-based retrieval and user decision states: User control and behavior modeling. JASIST 2004 55 488 497 10.1002/asi.10399
Funk ME Reid CA Indexing consistency in MEDLINE Bull Med Libr Assoc 1983 71 176 183 6344946
Haynes RB McKibbon KA Walker CJ Ryan N Fitzgerald D Ramsden MF Online access to MEDLINE in clinical settings: A study of use and usefulness Ann Intern Med 1990 112 78 84 2403476
Hersh WR Relevance and retrieval evaluation: perspectives from medicine J Am Soc Inform Sci 1994 45 201 206 10.1002/(SICI)1097-4571(199404)45:3<201::AID-ASI9>3.0.CO;2-W
Robertson SE Agosti M, Crestani F and Pasi G Evaluation in information retrieval ESSIR LNCS 2000 , Springer-Verlag 81 92
Rhodes BJ Maes P Just-in-time information retrieval agents IBM Systems Journal 2000 39 685 704
Egan DE Remde JR Gomez LM Landauer TK Eberhardt J Lochbaum CC Formative design-evaluation of Superbook ACM Trans Inf Syst 1989 7 30 57 10.1145/64789.64790
Hersh WR Elliot DL Hickam DH Wolf SL Molnar A Leichtenstein C Towards new measures of information retrieval evaluation: Jul 9-13; Seattle, Washington, USA. 1995 New York: ACM Press 164 170
Whitely WP Rennie D Hafner AW The scientific community's response to evidence of fraudulent publication. The Robert Slutsky case. JAMA 1994 272 170 173 8015137 10.1001/jama.272.2.170
Hersh WR Pentecost J Hickham DH A task-oriented approach to information retrieval evaluation J Am Soc Inform Sci 1996 47 50 56
| 15588311 | PMC539260 | CC BY | 2021-01-04 16:03:41 | no | BMC Med Inform Decis Mak. 2004 Dec 10; 4:21 | utf-8 | BMC Med Inform Decis Mak | 2,004 | 10.1186/1472-6947-4-21 | oa_comm |
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-4-341558829910.1186/1472-6963-4-34Research ArticleDifferences in access to coronary care unit among patients with acute myocardial infarction in Rome: old, ill, and poor people hold the burden of inefficiency Ancona Carla [email protected]à Massimo [email protected] Carlo [email protected] Nera [email protected] Danilo [email protected] Valeria [email protected] Carlo A [email protected] Department of Epidemiology, Local Health Authority RME, Rome, Italy2 Regional Public Health Agency, Friuli Venezia Giulia, Italy3 Agency for Public Health, Lazio Region, Italy2004 9 12 2004 4 34 34 13 7 2004 9 12 2004 Copyright © 2004 Ancona 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
Direct admission to Coronary Care Unit (CCU) on hospital arrival can be considered as a good proxy for adequate management in patients with acute myocardial infarction (AMI), as it has been associated with better prognosis. We analyzed a cohort of patients with AMI hospitalized in Rome (Italy) in 1997–2000 to assess the proportion directly admitted to CCU and to investigate the effect of patient characteristics such as gender, age, illness severity on admission, and socio-economic status (SES) on CCU admission practices.
Methods
Using discharge data, we analyzed a cohort of 9127 AMI patients. Illness severity on admission was determined using the Deyo's adaptation of the Charlson's comorbidity index, and each patient was assigned to one to four SES groups (level I referring to the highest SES) defined by a socioeconomic index, derived by the characteristics of the census tract of residence. The effect of gender, age, illness severity and SES, on risk of non-admission to CCU was investigated using a logistic regression model (OR, CI 95%).
Results
Only 53.9% of patients were directly admitted to CCU, and access to optimal care was more frequently offered to younger patients (OR = 0.35; 95%CI = 0.25–0.48 when comparing 85+ to >=50 years), those with less severe illness (OR = 0.48; 95%CI = 0.37–0.61 when comparing Charlson index 3+ to 0) and the socially advantaged (OR = 0.81; 95%CI = 0.66–0.99 when comparing low to high SES).
Conclusion
In Rome, Italy, standard optimal coronary care is underprovided. It seems to be granted preferentially to the better off, even after controversial clinical criteria, such as age and severity of illness, are taken into account.
==== Body
Background
The Italian National Health Service is supposed to provide universal coverage of standard care to all citizens, no social or economic selection bias should limit access to high technology resources as far as they are available.
Direct admission to Coronary Care Unit (CCU) on hospital arrival can be considered as a good proxy for adequate management in patients with acute myocardial infarction (AMI) because it has been associated with better prognosis and shorter hospital stay [1,2].
Timely access to advanced diagnostic and therapeutic options, thorough cardiovascular monitoring, provision of primary angioplasty or thrombolytic therapy when indicated, and prompt defibrillation when necessary all contribute to favorable outcomes.
With regard to reperfusion therapies, they have been shown to be effective in reducing short and mid-term mortality in patients with ST-segment elevation AMI [3-8]. The treatment is beneficial regardless of gender, and although relative mortality reduction is greater in younger than in older patients, absolute mortality reduction progressively increases in patients up to 75 years of age. After age 75, the benefits of treatment are less certain [9-11].
Intensive coronary care should be administered without delay, ideally within 60 minutes of the onset of symptoms [12-14].
Previous studies have investigated the role of demographic characteristics (such as age, gender and ethnicity), clinical characteristics (such as time from symptoms, presence of ST elevation and Killip class at presentation), and hospital characteristics (such as location, teaching status and level of invasive capability) on the probability of being admitted to CCU and the probability of receiving initial thrombolysis [11,15-18], but no information is available on factors affecting direct admission to CCU in Italy in an ordinary clinical setting. We analyzed a cohort of patients with AMI hospitalized in Rome (Italy) in 1997–2000 to assess the proportion directly admitted to CCU and to investigate the effect of patient characteristics such as gender, age, illness severity on admission, and socio-economic status on CCU admission practices.
Methods
Cohort selection criteria
We used discharge abstract data, routinely collected by the regional Hospital Information System (HIS), to identify a cohort of patients with AMI (ICD-9 code of principal diagnosis at discharge = 410), aged 18 years of age or more, residing in Rome, admitted to one of the 11 city hospitals equipped with an emergency department and a CCU, and surviving the Emergency Room, from 1 July 1997 to 31 December 2000.
From the above defined cohort, we then excluded:
• patients who had been hospitalized for AMI in the previous six months, identified through record linkage within the HIS file,
• patients transferred from other acute care facilities,
• ruled-out AMI (those discharged alive or discharged against medical advice with a length of stay less than five days),
• episodes of care with a diagnosis of trauma, with an important surgical operation, or with DRG non compatible with an AMI diagnosis.
Exposure and outcome
We categorized patients into five age-intervals: (less than 50, 50–64, 65–74, 75–84 and over 85 years). The Deyo's adaptation of the Charlson's comorbidity index [19] was calculated in order to describe illness severity on admission: based on ICD-9 codes we identified four severity groups (Charlson's adapted index 0, 1, 2, 3+). A socio-economic status (SES) index had been derived for each of the 5736 census tracts (CT) in Rome (average population = 480 inhabitants) using selected census variables including level of education, occupation, dwelling ownership, family size, and people/room density [20]. Based on CT of residence, AMI patients in the cohort were classified into four levels of SES (level I referring to the highest SES).
For each patient, vital status 30 days after hospital admission was obtained through an automatic record linkage with the Municipal Registry of Rome.
The main outcome measure was the indication of CCU as admission ward in the discharge abstract. For the purposes of this study we considered admission to Intensive Care Unit (ICU) in hospitals equipped with CCU equivalent to direct admission to CCU, because both represent an intensive care for AMI patients.
Statistical analysis
As a first step, a logistic regression analysis was performed in order to confirm the association between direct admission to CCU and 30 days mortality. after adjusting for patient characteristics (gender, age, SES, Charlson's index) and admitting hospital (ORs and 95% CI). Moreover, we compared the average length of stay among patients admitted to CCU and among those admitted to other wards, after excluding deceased and transferred patients, using the Student's t test.
We then assessed the extent of differences among Rome hospitals with respect to the number of AMI patients admitted, the size of CCU, and the proportion of AMI patients directly admitted to CCU. For each hospital the ratio between AMI patients admitted and number of CCU beds was calculated as a proxy of the pressure on CCU resources.
The effect of personal characteristics, i.e. gender, age, illness severity and socio-economic status, on risk of non-admission to CCU was then investigated using a logistic regression model. Since important differences among hospital rates of CCU admission had been found, and we wanted to take into account the possible effect of patients clustering by hospital, we used a random effect model with admitting hospital as clustering variable.
The area under the receiver operator characteristics (ROC) curve was estimate as a measure of the overall predictive ability of each model, while the Hosmer and Lemenshow (H-L) statistics was used to assess models' calibration.
The possible effect modification of gender on the other variables was tested by forcing interaction terms in the multivariate models.
All statistical analyses were performed using STATA version 7.0 [21].
Results
Of 9127 patients hospitalized with an incident AMI diagnosis (32% females, mean age 68.5 years, SD 12.6 years), 53.9% were directly admitted to CCU. The crude 30-day mortality was 13.2% in patients admitted to CCU and 25.0% in those admitted to other wards. The protective effect of CCU admission on 30-day mortality remained strong after adjusting for potential confounders (OR 0.53, I.C. 95% 0.47–0.60; area under the ROC = 0.729; H-L = 7.86, p = 0.448). Moreover, among 3765 AMI patients admitted to CCU or ICU, and discharged to home, the length of stay was significantly shorter than among those admitted to other wards (12.2 versus 14.4 days, p < 0.0001).
The CCU admission proportion varied widely among hospitals (Table 1), from 27.7% in hospital a (a large, public, teaching hospital which admitted 962 AMI patients during the study period) to 87.5% in hospitals l (a medium sized, no-profit, catholic hospital which admitted 393 AMI patients during the study period). The ratio between AMI patients admitted and number of CCU beds varied from 37 to 166. This two indicators were not slightly correlated (r = 0.52). Patients characteristics with respect to CCU admission status are presented in table 2.
In the bivariate analysis it emerged that lower odds of being directly admitted to CCU were associated to female gender, older age, and higher Charlson's index, while we found no difference with respect to SES level (Table 3, first three columns).
In multiple logistic regression analysis (Table 3, last two columns), when the potential confounding was adjusted for and the clustering by hospital was taken into account, the independent roles of gender (OR = 0.72; 95%CI = 0.64-0.84 when comparing females to males), age (a strong trend over the whole age range, with OR = 0.35; 95%CI = 0.25-0.48 when comparing 85+ to up to 50 yrs.) and illness severity (OR = 0.48; 95%CI = 0.37-0.61 when comparing Charlson index 3+ to 0) were confirmed, while low SES level emerged as an independent determinant of non-admission to CCU (OR = 0.79; 95%CI = 0.65-0.95 and OR = 0.81; 95%CI = 0.66-0.99 when comparing levels III and IV, respectively, to level I, area under the ROC = 0.704; H-L = 15.1, p = 0.06). No effect modification by gender was observed.
Discussion
Coronary care units have now been in use for 40 years, and it is generally acknowledged that they have helped to improve prognosis and reduce hospital stay among patients with acute myocardial infarction. This was confirmed in our AMI cohort where we observed a strong protective effect of CCU admission on 30 days mortality and a significantly shorter hospital stay for patient admitted to CCU.
We observed that in most Rome hospitals the proportion of AMI patients directly admitted to CCU is lower than it should be according to international recommendations, [12] and lower than that observed in other developed countries [22,23]. Moreover, we found wide differences in rates of CCU admission among hospitals. It is beyond the scope of this paper to investigate which structural and organizational characteristics at the hospital level are associated to high proportion of non-admission to CCU, however admission rates do not increase in hospitals where the number of AMI patients is low in comparison to available CCU beds. While available data, and the results of a previous study [24] suggest a less than optimal use of CCU resources. In fact, we found that, among the 11243 patients who passed through the 112 CCU beds available in the 11 Rome hospitals in the year 2000, for an overall length of stay of 62622 days, only 40% had a diagnosis of AMI, while 46% had principal diagnosis of other acute cardiac disease and 14% had other diagnoses. In summary, variable, and incongruous admission and discharge policies as well as actual shortage of beds could have affected the CCU admission rate of AMI patients, whatever the reasons CCU is apparently a scarce resource in Lazio hospital which should be used unbiasedly.
On the contrary, our results showed that age, severity of illness, and SES are important determinants of the probability that a patient with AMI who reaches qualified Rome hospital is directly admitted to CCU. Previous studies have documented restricted access to CCU and invasive procedures, and under use of well-established therapies such as aspirin, reperfusion and beta-blockers among elderly [25], female [17], and poorer AMI patients [26]. A recent systematic review suggests that patients who are perceived not to benefit from critical care are more often refused intensive care unit admission [27].
The age-related admission policy to CCU we observed has been documented previously [27,28], as well as the lower probability of being accepted in CCU for patients with higher severity [30]. The factors influencing admission decisions are likely to exclude large numbers of patients who could benefit from advanced diagnostic and therapeutic options [31].
We used discharge abstract data, coded according to the International Classification of Diseases IX revision, so it was impossible for us to distinguish between ST-segment elevation and non ST-segment elevation MI. Even though ST-elevation may (and should) influence the physician referral decision, we think that, if the percentages of non-ST segment MI in the groups under study are the same, our results should not strongly be biased.
We used a small area-based SES index, because direct individual data on social class were not available. This index has been shown to be a strong predictor of differences in mortality, [20] and associated to inequalities in access to important health interventions [32,33] and medical management [34] in Rome. Small-area data have been widely used to impute individual socio-economic status, and despite some criticism [35] inferences based on this method appear to be valid [36,37].
Age and admitting hospitals were the variables responsible for a negative confounding effect on the association of socio-economic status with direct CCU admission. Patients with low SES levels are younger than patients with high SES levels and tend to be admitted to hospital with higher provision of CCU care.
Conclusions
In Rome, where high-technology resources are available, they do not seem to be efficiently and fairly used. Our results suggest that access to optimal care is offered selectively to the socially advantaged, as well as to younger patients (even well under the 75 years threshold) and to those with less severe illness. The National Health Service, its policy of unrestricted access notwithstanding, falls short of providing equal opportunities to all citizens. Delivery of effective services to the underprivileged should be actively promoted.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CA conceived of the study, participated in its design, planned the statistical analysis, drafted the manuscript
MA participated in the study design and coordination, drafted the manuscript
CS participated in the study design, drafted the manuscript
NA participated in the study design
DF performed the statistical analysis
VT assisted in data management
CAP conceived of the study
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors thank Ms Margaret Becker for taking care of their English and revising repeatedly their manuscript.
This work has been partly funded by the Italian National Health Service – Progetto "Efficacia ed equità dell'assistenza ospedaliera: pubblicizzazione e informazione ai cittadini" (Ministero della Sanità – Dipartimento della Programmazione "Programmi speciali" – Art.12, comma 2, lett. b), del d. lgs. 502/92).
Figures and Tables
Table 1 Number of CCU beds, AMI patients admitted and directly admitted to CCU, and ratio between AMI admitted and CCU beds by hospital
N° of CCU beds AMI patients admitted AMI patients directly admitted to CCU % directly admitted to CCU Ratio between AMI patients admitted and CCU beds
Admitting Hospital
a 17 962 295 30.7 56.6
b 8 1331 585 43.9 166.4
c 4 639 282 44.1 159.8
d 20 738 361 48.9 36.9
e 9 1414 718 50.8 157.1
f 24 1408 784 55.7 58.7
g 6 678 451 66.5 113.0
h 7 815 552 67.7 116.4
i 7 532 379 71.2 76.0
l 2 217 173 79.7 108.5
m 8 393 344 87.5 49.1
Total 112 9127 4924 53.9 81.5
Table 2 Characteristics of AMI patients according to CCU status
% not admitted to CCU % directly admitted to CCU
(4203) (4924)
Gender
male 64.26 71.93
female 35.74 28.07
Age (years)
<50 5.14 9.28
50–64 23.65 33.94
65–74 27.84 29.16
75–84 29.31 20.35
85+ 14.06 7.27
Charlson's comorbidity index
0 42.76 55.22
1 31.33 30.26
2 15.58 9.38
3+ 10.33 5.14
SES level
I 17.74 16.76
II 32.42 31.81
III 28.91 29.85
IV 20.93 21.58
Table 3 Effect of patients personal characteristics on risk of non admission to Coronary Care Unit (CCU)
Patients % directly admitted to CCU OR 95% C.I. OR* 95% C.I.
Gender
Male 6243 56.7 1.00 1.00
Female 2884 47.9 0.70 0.64 – 0.77 0.73 0.64 – 0.84
Age (years)
<50 673 67.9 1.00 1.00
50–64 2665 62.7 0.79 0.67 – 0.95 0.79 0.60 – 1.04
65–74 2606 55.1 0.58 0.48 – 0.69 0.72 0.55 – 0.95
75–84 2234 44.8 0.35 0.32 – 0.46 0.49 0.37 – 0.65
85+ 949 37.7 0.29 0.23 – 0.35 0.35 0.25 – 0.48
Charlson's comorbidity index
0 4516 60.2 1.00 1.00
1 2807 53.1 0.75 0.68 – 0.82 0.69 0.60 – 0.80
2 1117 41.4 0.47 0.41 – 0.53 0.45 0.36 – 0.55
3+ 687 36.8 0.38 0.33 – 0.45 0.48 0.37 – 0.61
SES level
I 1523 52.6 1.00 1.00
II 2840 53.5 1.04 0.91 – 1.17 0.90 0.75 – 1.09
III 2603 54.8 1.09 0.96 – 1.24 0.79 0.65 – 0.95
IV 1883 54.7 1.09 0.95 – 1.25 0.81 0.66 – 0.99
OR: Crude Odds Ratio
OR*: Odds Ratio adjusted for age, gender, severity of illness, and SES. Random effect model with admitting hospital as clustering variable
==== Refs
Cairns JA Coronary thrombolysis – overwhelming evidence for mortality reduction in acute myocardial infarction Can J Cardiol 1988 4 211 3 3136899
Walker DM Wicks M Hubbard WN Thomas RD Increased mortality from inadequate provision of coronary care unit facilities J R Soc Med 1994 87 211 2 8182676
Gruppo Italiano per lo Studio della Streptochinasi nell'Infarto Miocardico (GISSI) Effectiveness of intravenous thrombolytic treatment in acute myocardial infarction Lancet 1986 1 397 402 2868337
ISIS-2 (Second International Study of Infarct Survival) Collaborative Group Randomised trial of intravenous streptokinase, oral aspirin, both, or neither among 17,187 cases of suspected acute myocardial infarction: ISIS-2 Lancet 1988 2 349 60 2899772
Gruppo Italiano per lo Studio della Sopravvivenza nell'Infarto Miocardico GISSI-2: a factorial randomised trial of alteplase versus streptokinase and heparin versus no heparin among 12,490 patients with acute myocardial infarction Lancet 1990 336 65 71 1975321
Chareonthaitawee P Gibbons RJ Roberts RS Christian TF Burns R Yusuf S The impact of time to thrombolytic treatment on outcome in patients with acute myocardial infarction. For the CORE investigators (Collaborative Organisation for RheothRx Evaluation) Heart 2000 84 142 8 10908248 10.1136/heart.84.2.142
Rotstein Z Mandelzweig L Lavi B Eldar M Gottlieb S Hod H Does the coronary care unit improve prognosis of patients with acute myocardial infarction? A thrombolytic era study Eur Heart J 1999 20 813 8 10329079 10.1053/euhj.1998.1452
Hu JS The impact of coronary care unit on mortality from acute myocardial infarction in Hong Kong. A prospective randomized controlled study Chin Med J (Engl) 1990 103 101 6 2118025
Fibrinolytic Therapy Trialists' (FTT) Collaborative Group Indications for fibrinolytic therapy in suspected acute myocardial infarction: collaborative overview of early mortality and major morbidity results from all randomised trials of more than 1000 patients Lancet 1994 343 311 22 7905143
Berger AK Radford MJ Wang Y Krumholz HM Thrombolytic therapy in older patients J Am Coll Cardiol 2000 36 366 74 10933344 10.1016/S0735-1097(00)00723-3
Boucher JM Racine N Thanh TH Rahme E Brophy J LeLorier J Theroux P Quebec Acute Coronary Care Working Group Quebec Acute Coronary Care Working Group Age-related differences in in-hospital mortality and the use of thrombolytic therapy for acute myocardial infarction CMAJ 2001 164 1285 90 11341137
ACC/AHA guidelines for the management of patients with Acute Myocardial Infarction. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines: 1999 updated
Department of Health National Service Framework for Coronary Heart Disease (Emerging Findings). HSC 1998/218
National Health Research and Development Program Canadian Consensus Conference on Coronary Thrombolysis – 1994 recommendations Can J Cardiol 1994 10 522 9 8012880
Clarke KW Gray D Keating NA Hampton JR Do women with acute myocardial infarction receive the same treatment as men? BMJ 1994 309 563 6 7916228
Rogers WJ Canto JG Barron HV Boscarino JA Shoultz DA Every NR Treatment and outcome of myocardial infarction in hospitals with and without invasive capability. Investigators in the National Registry of Myocardial Infarction J Am Coll Cardiol 2000 35 371 9 10676683 10.1016/S0735-1097(99)00505-7
Hanratty B Lawlor DA Robinson MB Sapsford RJ Greenwood D Hall A Sex differences in risk factors, treatment and mortality after acute myocardial infarction: an observational study J Epidemiol Community Health 2000 54 912 6 11076987 10.1136/jech.54.12.912
Maynard C Litwin PE Martin JS Cerqueira M Kudenchuk PJ Ho MT Kennedy JW Cobb LA Schaeffer SM Hallstrom AP Characteristics of black patients admitted to coronary care units in metropolitan Seattle: results from the Myocardial Infarction Triage and Intervention Registry (MITI) Am J Cardiol 1991 67 18 23 1986498 10.1016/0002-9149(91)90092-Y
Deyo RA Cherkin DC Ciol MA Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases J Clin Epidemiol 1992 45 613 619 1607900 10.1016/0895-4356(92)90133-8
Michelozzi P Perucci CA Forastiere F Fusco D Ancona C Dell'Orco V Inequality in health: increasing socioeconomic differentials in mortality in Rome, 1990–95 J Epidemiol Community Health 1999 53 687 693 10656097
STATA for Windows release 7 2001 Stata Press. College station, Texas
Mahon NG O'rorke C Codd MB McCann HA McGarry K Sugrue DD Hospital mortality of acute myocardial infarction in the thrombolytic era Heart 1999 81 478 82 10212164
Cairns JA Singer J Gent M Holder DA Rogers D Sackett DL Sealey B Tanser P Vandervoort M Coronary care unit utilization in Hamilton, Ontario, a city of 375,000 people Can J Cardiol 1988 4 25 32 3359352
Ancona Carla Saitto Carlo Marino Claudia Fusco Danilo Perucci Carlo A A discharge abstract review suggests sub-optimal use of coronary care unit resources in Lazio, Italy In proceedings of the 1st HTAi Meeting, Krakow, May 30–June 2, 2004 2004 24
Fleming C D'Agostino RB Selker HP Is coronary-care-unit admission restricted for elderly patients? A multicenter study Am J Public Health 1991 81 1121 6 1951821
Shen JJ Wan TT Perlin JB An exploration of the complex relationship of socioecologic factors in the treatment and outcomes of acute myocardial infarction in disadvantaged populations Health Serv Res 2001 36 711 32 11508636
Sinuff T Kahnamoui K Cook DJ Luce JM Levy MM Values Ethics and Rationing in Critical Care Task Force. Rationing critical care beds: a systematic review Crit Care Med 2004 32 1588 97 15241106 10.1097/01.CCM.0000130175.38521.9F
Amit G Goldman S Ore L Low M Kark JD The association between hospital department, medical treatment and outcome in acute myocardial infarction Isr Med Assoc J 2003 5 255 9 14509129
Pabon Osuna P Aros Borau F San Jose Garagarza JM Bermejo Garcia J Lopez Bescos L Monton Rodriguez AJ Thrombolysis in the elderly with acute myocardial infarction. The PRIAMHO study Rev Esp Cardiol 2000 53 1443 52 11084002
Sprung CL Geber D Eidelman LA Baras M Pizov R Nimrod A Oppenheim A Epstein L Cotev S Evaluation of triage decisions for intensive care admission Crit Care Med 1999 27 1073 9 10397207 10.1097/00003246-199906000-00021
Rotstein Z Mandelzweig L Lavi B Eldar M Gottlieb S Hod H Does the coronary care unit improve prognosis of patients with acute myocardial infarction? A thrombolytic era study Eur Heart J 1999 20 813 8 10329079 10.1053/euhj.1998.1452
Miceli M Di Lallo D Perucci CA Absence of economic barriers does not reduce disparities in the access to renal transplantation: a population based study in a region of central Italy. Dialysis Register of Lazio Region J Epidemiol Community Health 2000 54 157 8 10715750 10.1136/jech.54.2.157
Ancona C Agabiti N Forastiere F Arca M Fusco D Ferro S Perucci CA Coronary artery bypass graft surgery: socioeconomic inequalities in access and in 30 day mortality. A population-based study in Rome, Italy J Epidemiol Community Health 2000 54 930 5 11076990 10.1136/jech.54.12.930
Rapiti E Porta D Forastiere F Fusco D Perucci CA Socioeconomic status and survival of persons with AIDS before and after the introduction of highly active antiretroviral therapy. Lazio AIDS Surveillance Collaborative Group Epidemiology 2000 11 496 501 10955400 10.1097/00001648-200009000-00003
Soobader M LeClere FB Hadden W Maury B Using aggregate geographic data to proxy individual socioeconomic status: does size matter? Am J Public Health 2001 91 632 6 11291379
Smith GD Hart C Watt G Hole D Hawthorne V Individual social class, area-based deprivation, cardiovascular disease risk factors, and mortality: the Renfrew and Paisley Study J Epidemiol Community Health 1998 52 399 405 9764262
Krieger N Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology Am J Public Health 1992 82 703 10 1566949
| 15588299 | PMC539261 | CC BY | 2021-01-04 16:03:29 | no | BMC Health Serv Res. 2004 Dec 9; 4:34 | utf-8 | BMC Health Serv Res | 2,004 | 10.1186/1472-6963-4-34 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-5-531557921210.1186/1471-2202-5-53Research ArticleChloride equilibrium potential in salamander cones Thoreson Wallace B [email protected] Eric J [email protected] Departments of Ophthalmology and Pharmacology, University of Nebraska Medical Center, Omaha, NE, USA2004 5 12 2004 5 53 53 16 9 2004 5 12 2004 Copyright © 2004 Thoreson and Bryson; 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
GABAergic inhibition and effects of intracellular chloride ions on calcium channel activity have been proposed to regulate neurotransmission from photoreceptors. To assess the impact of these and other chloride-dependent mechanisms on release from cones, the chloride equilibrium potential (ECl) was determined in red-sensitive, large single cones from the tiger salamander retinal slice.
Results
Whole cell recordings were done using gramicidin perforated patch techniques to maintain endogenous Cl- levels. Membrane potentials were corrected for liquid junction potentials. Cone resting potentials were found to average -46 mV. To measure ECl, we applied long depolarizing steps to activate the calcium-activated chloride current (ICl(Ca)) and then determined the reversal potential for the current component that was inhibited by the Cl- channel blocker, niflumic acid. With this method, ECl was found to average -46 mV. In a complementary approach, we used a Cl-sensitive dye, MEQ, to measure the Cl- flux produced by depolarization with elevated concentrations of K+. The membrane potentials produced by the various high K+ solutions were measured in separate current clamp experiments. Consistent with electrophysiological experiments, MEQ fluorescence measurements indicated that ECl was below -36 mV.
Conclusions
The results of this study indicate that ECl is close to the dark resting potential. This will minimize the impact of chloride-dependent presynaptic mechanisms in cone terminals involving GABAa receptors, glutamate transporters and ICl(Ca).
==== Body
Background
Regulation of intracellular chloride levels results in a chloride equilibrium potential (ECl) that is hyperpolarized with respect to the resting potential in many nerve cells, but depolarized in others [1-5]. For example, ECl in salamander rod photoreceptors is 25 mV more positive than the dark resting potential [6]. The resting potential of cone photoreceptors in darkness is around -42 to -47 mV and estimates of ECl in cones have ranged from -65 mV to -36 mV [7-11]. Cone photoreceptors possess a number of Cl- conductances that help to shape their responses and synaptic output. As discussed below, the value of ECl in cones is an important parameter for determining the strength and polarity of these effects.
It has been suggested GABAa receptors in the terminals of cones may mediate inhibitory synaptic feedback from horizontal cells to cones [8]. Under this hypothesis, the light-evoked hyperpolarization of horizontal cells causes a cessation of GABA release and this disinhibition leads to a "feedback depolarization" in cones. There is evidence both for [e.g., [8]] and against [e.g., [12,13]; see review in ref. [14]]) this hypothesis. However, one prediction of the hypothesis is that the Cl- equilibrium potential (ECl) must be negative to the resting potential in order for GABA disinhibition to depolarize a cone.
Cones possess prominent Ca2+-activated Cl- currents (ICl(Ca)) [15-17] activated by the influx of Ca2+ through voltage-gated Ca2+ channels as well as by release of Ca2+ from intracellular stores [16]. Cl- flux through ICl(Ca) can be substantial: during a 1.4 sec depolarizing step, the charge movement accompanying activation of ICl(Ca) is estimated to be 8.5 times that produced by activation of ICa alone [16]. These large membrane currents can strongly influence photoreceptor responses, but the nature of these effects depends on the value of ECl. If ECl is positive to the resting potential, activation of ICl(Ca) can boost depolarizing feedback responses from horizontal cells onto cones and produce prolonged, regenerative depolarizing responses lasting many seconds [9,18,19]. On the other hand, if ECl is negative to the resting potential, activation of ICl(Ca) can operate as a negative feedback mechanism to limit regenerative activation of Ca2+ channels [15,17]. In addition to altering membrane potential, depletion of intracellular Cl- can directly inhibit the open channel probability of single Ca2+ channels, presumably by modifying an anion binding site on the intracellular surface of the channel [11]. In rods, where ECl is positive to the resting potential, there is evidence for a negative feedback pathway between ICa and ICl(Ca) in which activation of ICa stimulates ICl(Ca) leading to a Cl- efflux that in turn inhibits Ca2+ channel activation [6,20]. If, however, ECl in cones is negative to the membrane potential, then activation of ICl(Ca) would stimulate an influx of Cl- that would be expected to enhance Ca2+ channel open probability [11].
Cone photoreceptors have presynaptic glutamate transporters that are coupled to Cl- channels [21-23]. The transporters in cones have been shown to respond to glutamate released from their own terminals [24]. Whether synaptically released glutamate causes cones to hyperpolarize or depolarize depends on ECl. Furthermore, analogous to the negative feedback from ICl(Ca) onto ICa described above, the chloride current produced by activation of glutamate transporters in rods can cause a Cl- efflux that inhibits ICa [25]. As with the feedback between ICl(Ca) and ICa, the strength and polarity of this potential interaction in cones depends on ECl.
Given the importance of ECl in determining the impact of various feedback mechanisms in the photoreceptor terminal, we determined ECl in cone photoreceptors of the salamander retina using a combination of imaging with a chloride-sensitive dye and electrophysiological approaches.
Results
In control superfusate, dark resting potentials of cones from slices prepared under visible light averaged -46.0 ± 2.00 mV (n = 9) after correcting for the liquid junction potential. This is nearly identical to the dark resting potentials of salamander cones prepared under infrared illumination (-46.8 ± 2.03 mV, n = 18).
To measure ECl, ICl(Ca) was recorded using gramicidin perforated patch whole cell recordings and activated by applying a 500 ms step from -78 to -8 mV. This depolarizing step typically evoked a sustained inward tail current arising largely from activation of ICl(Ca) [19]. Only cells that exhibited an inward tail current were used for analysis. As shown in the example of Fig. 1A, the current/voltage relationship of a cone cell was assessed during the tail current by using a ramp voltage protocol (1 mV/ms from -98 to +52 mV) begun 25 ms after the end of the depolarizing step. The same protocol was then repeated after applying niflumic acid (0.1 mM; Fig. 1B). At this concentration, niflumic acid is a selective inhibitor of ICl(Ca) in cones [[19]; niflumic acid may not be as selective in rods: [20,26]]. Subtracting the control ramp-evoked current from that obtained in the presence of niflumic acid yields the current/voltage profile for ICl(Ca) (Fig. 1C). In the example shown in Fig. 1C, the difference current reversed around -46 mV. The reversal potential of the niflumic acid-sensitive difference current determined from 8 cones averaged -45.5 ± 2.5 mV. As a control for the possible perturbation of intracellular Cl- by possible patch rupture, we repeated the same experiment using a pipette solution with only 3.5 mM Cl-. ECl was not significantly different when measured using the low Cl- pipette solution (-50.4 mV ± 3.4 mV; n = 7; p = 0.49, unpaired t-test). If patch rupture had occurred, ECl would be expected to attain -89 mV with the low Cl- pipette solution and -20 mV with the original pipette solution.
Bath application of GABA evoked small reversible inward currents that averaged -3.4 ± 0.3 pA at the holding potential of -78 mV (not shown). The small size of these currents may be due to receptor desensitization [27]. Consistent with results obtained from measurements of ICl(Ca), difference currents calculated from ramps applied before and during GABA application indicate that the GABA-evoked current reversed at -46.4 ± 2.7 mV (n = 9).
In a complementary approach for measuring ECl, we used a Cl-sensitive dye, MEQ, to examine the Cl- flux that accompanied cone depolarization evoked by bath application of various high K+ solutions (12, 22, 31, 41, 50 and 70 mM K+). In a separate set of experiments, we used gramicidin-perforated patch recording methods to measure the membrane potentials produced in cones by application of the different high K+ solutions. Slices used for MEQ experiments and for measurement of membrane potentials in different solutions were prepared using similar techniques under visible illumination; control experiments showed that the fluorescent illumination used during MEQ experiments did not produce any further changes in the cone resting membrane potential (n = 3). An example of a retinal slice loaded with MEQ is shown in Fig. 2A. Measurements of MEQ fluorescence were made from the cone soma (circle, Fig. 2A). For a single wavelength dye such as MEQ, the change in fluorescence relative to basal fluorescence (ΔF/F) can be used as a measure of the change in ion concentration [28]. In the cone in Fig. 2B, bath application of 12 mM K+, which depolarized cones to -36 mV, produced a 1.2% decrease in MEQ fluorescence. Since MEQ fluorescence is quenched by Cl- ions this indicates that depolarization to -36 mV stimulated an influx of Cl- ions. Application of a solution with 70 mM K+, which depolarizes cones to -7 mV, produced a greater influx of Cl- as evidenced by the 10% decrease in MEQ fluorescence seen in a different cone (Fig. 2C). Fig. 2D shows the average change in ΔF/F (x100) plotted as a function of the membrane potential evoked by the different high K+ solutions. The finding that 12 mM K+ consistently stimulated an influx of Cl- indicates that the reversal potential must be below -36 mV.
Discussion
The main finding of this study is that ECl in salamander cones is close to the dark resting potential (~-46 mV). ECl was found to be -46 mV from block of ICl(Ca) by niflumic acid; small GABA-evoked currents reversed around the same potential. MEQ fluorescence changes produced by depolarization support these electrophysiological measurements by indicating that ECl is below -36 mV.
There can be local variations of ECl within cells [4]. Large single cones in the salamander retina do not have a distinct axon and terminal; synaptic proteins are instead located at the base of the soma [29]. MEQ measurements were made in the cell soma from a region adjacent to the synaptic ending (see Fig. 2). ICl(Ca) is localized to the terminal region in rods (30) and these channels are probably also localized to the terminals of cones. Thus, the measurements in the present study are likely to provide estimates of ECl in the synaptic terminal and adjacent regions of the cone cell. Measurements of intracellular Cl- levels suggest that ECl in the inner segment is not significantly different from that measured in the soma [11].
The finding that the Cl- equilibrium potential is close to the resting potential does not necessarily mean that Cl- is passively distributed. Electrophysiological experiments required that cells be voltage clamped at -70 mV for many minutes. Nonetheless, the value of ECl determined from these electrophysiological experiments in which cells were voltage clamped at -70 mV was similar to the value estimated from MEQ studies in which cells were not voltage clamped and thus at their resting membrane potential. Results from experiments on the prolonged depolarization in cones also suggest that ECl can be maintained indefinitely at a value above the membrane potential. The plateau phase of the prolonged depolarization, which largely reflects ICl(Ca) activation [9,19], could remain above the membrane potential established by an adapting background for hours [9]. The ability of cones to maintain ECl above the membrane potential may arise from activity of the Na/KCl cotransporter as shown in rods (20) as well as from other mechanisms (e.g., CLC-2) [2,31].
Comparisons with other studies
ECl in cones has been estimated in a number of previous studies. The most positive value for ECl of -36 mV comes from calibration of MEQ fluorescence levels to determine the resting intracellular Cl- concentrations in cones isolated from the salamander retina (11). However, these measurements showed a large variability (range of S.E.M.: -26.5 to -46.6 mV). The most negative estimate of ECl comes from a study by Attwell et al [7] showing that the sign-reversing pathway from rods to cones reversed around -65 mV. Based on the presumption that this pathway involved disinhibition of GABAergic inputs into cones, this study has been interpreted as suggesting that ECl is around -65 mV. However, more recent evidence questions whether the horizontal cell to cone feedback pathway thought to underlie this sign-reversing pathway from rods to cones is truly GABAergic [9,13,14]. Other studies have arrived at values for ECl similar to those found in the present study. 1) By examining the polarity of GABA-evoked currents after patch rupture with either 12 or 24 mM Cl- in the recording pipette, Kaneko and Tachibana [8] estimated ECl to be around -47 mV in isolated turtle cones. 2) Based on the membrane potential attained by the plateau phase of the prolonged depolarization in turtle cones from the eyecup slice preparation, ECl was estimated to be at or slightly above the dark resting potential of -42 mV [after correction for a liquid junction potential of -2 mV; ref. [9]]. 3) In a single recording from a salamander cone obtained with a Cl-sensitive electrode, Miller and Dacheux [32] found that ECl was 2 mV more positive than the dark resting potential. 4) A slightly more negative value for ECl was found in ruptured patch recordings from goldfish cones by examining the voltage dependence of the ICl(Ca) tail current [10]. By extrapolating measurements back to the time of patch rupture, Kraaij et al [10] concluded that ECl was ~-55 mV.
Functional implications
ICa in cones, like that of rods, can be inhibited by lowering extracellular Cl- [33]. The inhibition of ICa produced by lowering extracellular Cl- appears to result from a reduction in intracellular Cl- which in turn causes a reduction in the open probability of single Ca2+ channels [11]. In rods, where ECl is positive to the resting potential, activation of Cl- channels leads to a Cl- efflux thereby producing an inhibition of Ca2+ channels [6,11,20]. The present results indicate that activation of Cl- channels when the cell is at its resting potential would produce minimal changes in intracellular Cl- in cones. Therefore, the feedback between ICa and ICl(Ca) postulated for rod photoreceptors [6,20] would be expected to be minimal in cones in darkness.
Another implication of the finding that ECl is close to the dark resting potential is that the stimulation of Cl- channels associated with glutamate transporters by glutamate released from cone terminals [24] would tend to stabilize the cell membrane potential near the dark potential. In rods, the Cl- efflux accompanying activation of glutamate transporters appears to contribute to a glutamate-mediated inhibition of ICa [25]. As with the feedback between ICa and ICl(Ca) considered in the previous paragraph, the finding that ECl is near the resting potential leads to the prediction that in darkness there would be no Cl- efflux accompanying glutamate transporter activation and therefore glutamate would not be expected to inhibit ICa.
Cones hyperpolarize to light, although with prolonged illumination the membrane potential recovers to near the dark resting potential. The impact of chloride-dependent negative feedback between ICl(Ca) and ICa or the glutamate transporter chloride current and ICa would be expected to increase as a cone hyperpolarizes in response to light. By reducing glutamate release, these chloride-dependent negative feedback mechanisms might thus contribute to making post-synaptic responses more transient.
The finding that ECl is near the resting potential of cones indicates that GABAergic disinhibition near the dark potential should produce little membrane potential change. This result is inconsistent with the postulated role for GABA in generating the feedback depolarization [8] and supports other studies suggesting that GABA is not directly responsible for horizontal to cone feedback [9,13,14].
Conclusions
Electrophysiological measurements, supported by experiments using chloride-sensitive dyes, indicate that ECl in salamander cones is close to the dark resting membrane potential. By minimizing the trans-membrane flux of chloride, this will minimize the presynaptic impact of GABAa receptors, ICl(Ca), and glutamate transporter chloride channels.
Methods
Tissue preparation
ECl is positive to the resting potential of many neurons in the immature brain [5]. Based on their size, the neotenous tiger salamanders (Ambystoma tigrinum, 15–25 cm) used in these experiments are thought to be 2–7 years old out of a life span of ~12 years (34).
Salamanders were handled humanely in accordance with protocols approved by the Institutional Animal Care and Use Committee at the University of Nebraska Medical Center. Chilled salamanders were rapidly decapitated, an eye was enucleated, and the front of the eye was removed. The resulting eyecup was cut into three or four pieces and a single piece was placed vitreal surface down onto a piece of filter paper (2 × 5 mm, Millipore type AAWP, 0.8 μm pores). After adhering to the filter paper, the retina was isolated under chilled amphibian superfusate and cut into 125 μm slices using a razor blade tissue chopper (Stoelting Co., Wood Dale, IL). The slices were rotated 90° to view the retinal layers when placed under a water immersion objective (60X, 1.0 NA) on an upright fixed stage microscope (EF 600, Nikon Inc., USA). Slices were prepared under visible light but recordings were performed in darkness. All experiments were done using red-sensitive large single cones selected by anatomical criteria [35].
Solutions and perfusion
Solutions were applied with a single-pass, gravity-feed perfusion system (1 ml/min). The normal amphibian superfusate contained (in mM): 111 NaCl, 2.5 KCl, 1.8 CaCl2, 0.5 MgCl2, 10 N-2-hydroxyethylpiperazine-N' 2-ethanesulfonic acid (HEPES), and 5 glucose (pH 7.8). The osmolarity was measured with a vapor pressure osmometer (Wescor, Logan, UT) and adjusted, if necessary, to 242 ± 5 mOsm. For high K+ solutions, various quantities of NaCl were replaced with equimolar KCl. Niflumic acid was diluted (1:10,000) from DMSO stock solutions. Unless otherwise specified, chemicals were obtained from Sigma/Aldrich/RBI (St. Louis, MO).
Electrophysiology
Patch pipettes were pulled on a PP-830 vertical puller (Narishige USA, New York) from borosilicate glass pipettes (1.2 mm O.D., 0.95 mm I.D., with internal filament) and had tips of ~1 μm outer diameter with resistances of 10 to 15 MΩ. To maintain endogenous levels of intracellular Cl-, we obtained perforated patch whole cell recordings using the cation channel, gramicidin [36]. Gramicidin was dissolved in ethanol (5 mg/ml) and then added to the pipette electrolyte solution to achieve a final concentration of 5 μg/ml. For current clamp measurements of membrane potentials, the pipette electrolyte solution contained (in mM): 54 KCl, 61.5 KCH3SO4 (Pfaltz and Bauer, Waterbury, CT), 3.5 NaCH3SO4, 10 HEPES. The pH was adjusted to 7.2 with KOH. The liquid junction potential (LJP) of this solution was estimated to be -7 mV using the junction potential calculator of PClamp (Axon Instruments). Membrane potential values reported throughout this manuscript were corrected for the LJP. For experiments with niflumic acid or GABA, pipettes were typically filled with a solution containing (in mM): 54 CsCl, 61.5 CsCH3SO3, 3.5 NaCH3SO4, 10 HEPES (LJP = -8 mV). In some experiments, a low Cl- pipette solution was used containing: 115.5 mM CsCH3SO3, 3.5 NaCl, 10 HEPES (LJP = -10 mV). The pH of both solutions was adjusted to 7.2 with CsOH. The osmolarity of pipette solutions were also adjusted, if necessary, to 242 ± 5 mOsm. Recordings were made using an Axopatch 200B amplifier (Axon Instruments Inc., Union City, CA) and PClamp 8 software (Axon Instruments). Cell input resistance calculated using a step from -70 to -90 mV averaged 695 ± 111 MΩ. Access resistance estimated from the peak of the capacitative transient averaged 30.7 ± 4.8 MΩ (n = 24).
Imaging experiments
Digital fluorescent images were obtained with a cooled CCD camera (SensiCam, Cooke Corp., Auburn Hills, MI). Axon Imaging Workbench (AIW 2.2, Axon Instruments Inc., Union City, CA) was used to control the camera, filter wheel, and image acquisition. Pixel binning (2 × 2) of the images was used to decrease acquisition time to ≤1 s. Images were acquired at 5 to 10 s intervals during experimental trials.
For measurements of [Cl-]i we used the dye, 6-methoxy-N-ethylquinolinium iodide (MEQ, Molecular Probes, Eugene, OR) [37]. MEQ was loaded into cells after reducing it to DiH-MEQ by adding 30 μM sodium borohydride (100 μl) to MEQ (5 mg) under a continuous stream of nitrogen gas [38]. DiHMEQ enters cells during the incubation period (15 min) where it is oxidized and retained in the form of MEQ. Fluorescence emission decreases as Cl- quenches MEQ. The slow exponential decay in MEQ fluorescence due to dye leakage and bleaching was determined from a 3 min. series of control measurements prior to drug application and subtracted before analysis [11,20].
Variance is reported as ± S.E.M.
Authors' contributions
WT conceived the study and drafted the manuscript. WT and EB participated in all aspects of the experiments but most recordings were performed by EB. Both authors have read and approved the manuscript.
Acknowledgments
Supported by the National Eye Institute (EY10542), Research to Prevent Blindness, Inc., the Gifford Foundation, and the Nebraska Lions Foundation. The authors thank Dr. Dwight Burkhardt for his helpful comments on the manuscript.
Figures and Tables
Figure 1 Cone ECl estimated from the reversal of ICl(Ca). A) ICl(Ca) tail current was activated by applying a 500 ms step from -78 to -18 mV during a gramicidin perforated patch whole cell recording from a rod. A ramp voltage protocol (-98 to +52 mV, 1 mV/ms) was applied during the tail current and begun 25 ms after termination of the step. B. The same protocol was then repeated in the presence of niflumic acid (0.1 mM) to inhibit ICl(Ca). C. The ramp current/voltage relationship obtained in control medium (A) was subtracted from that obtained in the presence of niflumic acid (B) to yield a niflumic acid-sensitive difference current that reversed in this cell around -46 mV (after correcting for a liquid junction potential of -8 mV).
Figure 2 Intracellular chloride measurements. A. Example of a retinal slice fluorescently stained with the Cl- sensitive dye MEQ. The circle indicates the somatic region of a cone in which MEQ measurements were made. B. An example of the reduction in MEQ fluorescence, indicating an increase in intracellular [Cl-], produced by application of 12.1 mM K+ to a cone. C. A greater decrease in MEQ fluorescence was produced by application of 69.9 mM K+. Different cone from panel B. D. MEQ fluorescence changes produced by bath application of superfusate containing 12.1, 21.6, 31.2, 40.7, 50.3, or 69.9 mM [K+]. After correcting for liquid junction potentials, these high K+ solutions depolarized cones to -35.3 ± 2.90, -26.3 ± 2.82, -20.6 ± 2.82, -15.4 ± 2.84, -11.6 ± 3.05, and -7.4 ± 3.25 mV (n = 9), respectively, determined in current clamp recordings using gramicidin-perforated patch recording techniques. The change in MEQ fluorescence relative to basal fluorescence (ΔF/F*100) is plotted against the cone membrane potential determined with each high K+ solution.
==== Refs
Alvarez-Leefmans FJ Gamino SM Giraldez F Nogueron I Intracellular chloride regulation in amphibian dorsal root ganglion neurones studied with ion-selective microelectrodes J Physiol 1988 406 225 246 3254412
Staley K Smith R Schaack J Wilcox C Jentsch TJ Alteration of GABAA receptor function following gene transfer of the CLC-2 chloride channel Neuron 1996 17 543 551 8816717 10.1016/S0896-6273(00)80186-5
Reuter D Zierold K Schroder WH Frings S A depolarizing chloride current contributes to chemoelectrical transduction in olfactory sensory neurons in situ J Neurosci 1998 18 6623 6630 9712634
Vardi N Zhang LL Payne JA Sterling P Evidence that different cation chloride cotransporters in retinal neurons allow opposite responses to GABA J Neurosci 2000 20 7657 7663 11027226
Ben-Ari Y Excitatory actions of gaba during development: the nature of the nurture Nature Rev Neurosci 2002 3 728 739 12209121 10.1038/nrn920
Thoreson WB Stella SL JrBryson EJ Clements J Witkovsky P D2-like dopamine receptors promote interactions between calcium and chloride channels that diminish rod synaptic transfer in the salamander retina Vis Neurosci 2002 19 235 247 12392173 10.1017/S0952523802192017
Attwell D Werblin FS Wilson M Wu SM A sign-reversing pathway from rods to double and single cones in the retina of the tiger salamander J Physiol 1983 336 313 333 6308225
Kaneko A Tachibana M Effects of gamma-aminobutyric acid on isolated cone photoreceptors of the turtle retina J Physiol 1986 373 443 461 3746679
Thoreson WB Burkhardt DA Ionic influences on the prolonged depolarization of turtle cones in situ J Neurophysiol 1991 65 96 110 1900325
Kraaij DA Spekreijse H Kamermans M The nature of surround-induced depolarizing responses in goldfish cones J Gen Physiol 2000 115 3 16 10613914 10.1085/jgp.115.1.3
Thoreson WB Nitzan R Miller RF Chloride efflux inhibits single calcium channel open probability in vertebrate photoreceptors: chloride imaging and cell-attached patch-clamp recordings Vis Neurosci 2000 17 197 206 10824674 10.1017/S0952523800172025
Thoreson WB Burkhardt DA Effects of synaptic blocking agents on the depolarizing responses of turtle cones evoked by surround illumination Vis Neurosci 1990 5 571 583 2085473
Verweij J Kamermans M Spekreijse H Horizontal cells feed back to cones by shifting the cone calcium-current activation range Vision Res 1996 36 3943 5393 9068848 10.1016/S0042-6989(96)00142-3
Kamermans M Spekreijse H The feedback pathway from horizontal cells to cones. A mini review with a look ahead Vision Res 1999 39 2449 2468 10396615 10.1016/S0042-6989(99)00043-7
Maricq AV Korenbrot JI Calcium and calcium-dependent chloride currents generate action potentials in solitary cone photoreceptors Neuron 1988 1 503 515 2483100 10.1016/0896-6273(88)90181-X
Barnes S Hille B Ionic channels of the inner segment of tiger salamander cone photoreceptors J Gen Physiol 1989 94 719 743 2482325 10.1085/jgp.94.4.719
Yagi T Macleish PR Ionic conductances of monkey solitary cone inner segments J Neurophysiol 1994 71 656 665 7513752
Burkhardt DA Gottesman J Thoreson WB Prolonged depolarization in turtle cones evoked by current injection and stimulation of the receptive field surround J Physiol 1988 407 329 348 3256619
Barnes S Deschenes MC Contribution of Ca and Ca-activated Cl channels to regenerative depolarization and membrane bistability of cone photoreceptors J Neurophysiol 1992 68 745 755 1331354
Thoreson WB Bryson EJ Rabl K Reciprocal interactions between calcium and chloride in rod photoreceptors J Neurophysiol 2003 90 1747 1753 12724369
Eliasof S Arriza JL Leighton BH Kavanaugh MP Amara SG Excitatory amino acid transporters of the salamander retina: identification, localization, and function J Neurosci 1998 18 698 712 9425012
Eliasof S Werblin F Characterization of the glutamate transporter in retinal cones of the tiger salamander J Neurosci 1993 13 402 411 8093715
Picaud SA Larsson HP Grant GB Lecar H Werblin F Glutamate-gated chloride channel with glutamate-transporter-like properties in cone photoreceptors of the tiger salamander J Neurophysiol 1995 74 1760 1771 8989410
Picaud S Larsson HP Wellis DP Lecar H Werblin F Cone photoreceptors respond to their own glutamate release in the tiger salamander Proc Natl Acad Sci USA 1995 92 9417 9421 7568144
Rábl K Bryson EJ Thoreson WB Activation of glutamate transporters in rods inhibits presynaptic calcium currents Vis Neurosci 2003 20 557 566 14977334 10.1017/S0952523803205095
Satoh TO Yamada M Niflumic acid reduces the hyperpolarization-activated current (I(h)) in rod photoreceptor cells Neurosci Res 2001 40 375 381 11463484 10.1016/S0168-0102(01)00252-8
Epstein R Grundfest H Desensitization of gamma aminobutyric acid (GABA) receptors in muscle fibers of the crab Cancer borealis J Gen Physiol 1970 56 33 45 4327880 10.1085/jgp.56.1.33
Helmchen F Yuste R, Lanni F, Konnerth A Calibration of fluorescent calcium indicators Imaging Neurons: A Laboratory Manual 2000 New York: Cold Spring Harbor Laboratory Press 32.1 32.9
Sherry DM Yang H Standifer KM Vesicle-associated membrane protein isoforms in the tiger salamander retina J Comp Neurol 2001 431 424 436 11223812 10.1002/1096-9861(20010319)431:4<424::AID-CNE1080>3.0.CO;2-Y
Macleish PR Nurse CA Ion channel compartments in photoreceptors [abstract] Invest Ophthalm Vis Sci 2000 41 S494
Bösl MR Stein V Hubner C Zdebik AA Jordt SE Mukhopadhyay AK Davidoff MS Holstein AF Jentsch TJ Male germ cells and photoreceptors, both dependent on close cell-cell interactions, degenerate upon ClC-2 Cl- channel disruption EMBO J 2001 20 1289 1299 11250895 10.1093/emboj/20.6.1289
Miller RF Dacheux RF Intracellular chloride in retinal neurons: measurement and meaning Vision Res 1983 23 399 411 6880038 10.1016/0042-6989(83)90087-1
Thoreson WB Nitzan R Miller RF Reducing extracellular Cl- suppresses dihydropyridine-sensitive Ca2+ currents and synaptic transmission in amphibian photoreceptors J Neurophysiol 1997 77 2175 2190 9114264
Townes-Anderson E Colantonio A St Jules RS Age-related changes in the tiger salamander retina Exp Eye Res 1998 66 653 667 9631666 10.1006/exer.1998.0472
Stella SL JrThoreson WB Differential modulation of rod and cone calcium currents in tiger salamander retina by D2 dopamine receptors and cAMP Eur J Neurosci 2000 12 3537 3548 11029623 10.1046/j.1460-9568.2000.00235.x
Kyrozis A Reichling DB Perforated-patch recording with gramicidin avoids artifactual changes in intracellular chloride concentration J Neurosci Methods 1995 57 27 35 7540702 10.1016/0165-0270(94)00116-X
Biwersi J Verkman AS Cell-permeable fluorescent indicator for cytosolic chloride Biochemistry 1991 30 7879 7883 1868062
Woll E Gschwentner M Furst J Hofer S Buemberger G Jungwirth A Frick J Deetjen P Paulmichl M Fluorescence-optical measurements of chloride movements in cells using the membrane-permeable dye diH-MEQ Pflugers Arch 1996 432 486 493 8766008 10.1007/s004240050160
| 15579212 | PMC539262 | CC BY | 2021-01-04 16:03:46 | no | BMC Neurosci. 2004 Dec 5; 5:53 | utf-8 | BMC Neurosci | 2,004 | 10.1186/1471-2202-5-53 | oa_comm |
==== Front
BMC NeurolBMC Neurology1471-2377BioMed Central London 1471-2377-4-221557920110.1186/1471-2377-4-22Case ReportAndrogen-induced cerebral venous sinus thrombosis in a young body builder: case report Sahraian Mohammad Ali [email protected] Mahmood [email protected] Amir Reza [email protected] Babak [email protected] Department of Neurology, Sina Hospital, Tehran University of Medical Sciences, Hassan Abad Square, Imam Khomini Street, Tehran, Iran2004 3 12 2004 4 22 22 21 9 2004 3 12 2004 Copyright © 2004 Sahraian et al; licensee BioMed Central Ltd.2004Sahraian 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
Cerebral venous sinus thrombosis is an infrequent disease with a variety of causes. Pregnancy, puerperium, contraceptive pills and intracranial infections are the most common causes. The patient may present with headache, focal neurological deficits and seizures.
The clinical outcome is highly variable and treatment with heparin is advised.
Case presentation
The patient is a 22 year old male who presented with headache, repeated vomiting and papilledema.
He was a bodybuilder doing exercise since 5 years ago, who had used nandrolone decaonoate 25 milligrams intramuscularly during the previous 5 months. Brain MRI and MRV showed superior sagital and transverse sinus thrombosis and extensive investigations did not reveal any known cause.
Conclusions
We suggested that androgen was the predisposing factor in our patient. Androgens may increase coagulation factors or platelet activity and cause arterial or venous thrombosis.
As athletes may hide using androgens it should be considered as a predisposing factor for thrombotic events in such patients.
==== Body
Background
Cerebral venous sinus thrombosis (CVST) is a disease with a wide spectrum of non specific clinical signs and symptoms, including headache, focal neurological deficits, seizures and coma.
The clinical outcome is highly variable; patients may recover completely or may develop severe and lasting neurological deficits [1].
There are many causes for this disease but the most common predisposing factors are pregnancy, puerperium, contraceptive pills, coagulopathies and intracranial infections [2].
There are few reports of patients with CVST after androgen therapy. We present a young bodybuilder man who developed CVST with abusing androgens for increasing muscle mass.
Case presentation
In May 2004 a 22 year old male was admitted to our department with chief complaints of headache and vomiting.
The patient was well till 10 days prior to admission that developed progressive, intense bitemporal headache exacerbated with bending.
The patient also had history of malaise, nausea and several episodes of vomiting from 3 days before admission. The only objective finding on physical examination was bilateral papilledema. The patient was a body builder doing exercise from 5 years ago who had used nandrolone decaonoate 25 mg once or twice a week during the last 5 months. He had injected 20 ampoules in this period.
Brain computed tomography without contrast was done for the patient which showed cord sign, emergency MR imaging including T1 – T2, weighted and MRV showed prominent superior sagital and transverse sinus thrombosis. The C.S.F opening pressure was 480 mm/H2o without any other abnormality. Heparin 80 IU/kg started as loading dose then continued 1000 IU/ hr for 10 days. On the 5th day of treatment headache resolved and warfarin added to heparin.
Laboratory tests including antithrombin III activity, protein C, S factor V leiden, Plasma hemocystein and anticardiolipin were all within normal limits.
The patient was discharged in a good condition and was maintained on 6 months warfarin protcol.
Discussion
There are few reports of CVST following androgen therapy [3], but there is just one reported case of CVST in androgen using young body builder [4]. The anabolic activity of testosterone and its derivatives is primarily manifested in its myotrophic actions which result in greater muscle mass and strength.
This has led to widespread use of androgenic anabolic steroids by athletes at all levels. Nandrolone decaonoate is a synthetic anabolic steroid. In focus on homeostasis system the most important factors under testosterone regulation are fibrinogen, Plasminogen activator inhibitor-1 (PAI – 1) and platelet aggregability.
The current data indicate that testosterone lowers fibrinogen and PAI – 1, however these anticoagulatory and profibrinolytic may be opposed by proaggregatory effects on platelets because high dosages of androgens were found to decrease cycloxygenase activity and thereby increase platelet functions [5].
Proaggregatory effect of testosterone and other synthetic androgens become more reliable theory for CVST, according to recent publication [6].
Conclusions
This case report presents a patient with CVST following exogenous androgen usage with a mechanism which is not completely understood, but it may be related to platelet activation or an increase in coagulation factors. As androgen use may be frequent and hidden in athletes, it may be an underestimated cause of cerebral venous thrombosis in young adults and careful history should be taken in these groups of patients.
Lists of abbreviations
CVST: cerebral venous sinus thrombosis
PAI-1: Plasminogen activator inhibitor-1
MRI: magnetic resonance imaging
MRV: magnetic resonance venography
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
M.A.S: Admitting and treating the case, preparing the article.
M.M : Revising the article.
A.R.A : searching previous articles, preparing case presentation part.
B.M : searching previous articles, preparing background.
Figure 1 T1 weighted horizontal nonGd Images show transverse sinus thrombosis.
Figure 2 T1 weighted sagital Images with Gd show thrombosis in sagital sinus.
Figure 3 2D MR venogram shows sagital sinus thrombosis.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are grateful to Mr. Hejrani for secretarial help and Mr. Solimani for preparing the photographs. It should be noted that written consent was obtained from the patient for publication of study.
==== Refs
Ameri A Bousser MG Cerebral venous sinus thrombosis Neurol clin 1992 10 87 111 1557011
Shelley Renowden Cerebral Venous sinus thrombosis Eur Radiol 2004 14 215 26 14530999 10.1007/s00330-003-2021-6
Shiozawa Z Yamada H Mabuchic Hotta T saito M Sobue I Huang YP Superior sagital sinus thrombosis associated with androgen therapy for hypoplastic anemia [abstract] Ann Neurol 1982 12 578 80 7159062
Jaillard AS Hommel M Mallaret M Venous sinus thrombosis associated with androgens in a healthy young man [abstract] Stroke 1994 25 212 13 8266370
Shahidi NT A review of the chemistry, biological action and clinical applications of anabolic androgenic steroids Clin therap 2001 23 1355 90 11589254 10.1016/S0149-2918(01)80114-4
Wu FC Von Eckardstein A Androgens and coronary artery disease Endocr Rev 2003 24 183 217 12700179 10.1210/er.2001-0025
| 15579201 | PMC539263 | CC BY | 2021-01-04 16:28:50 | no | BMC Neurol. 2004 Dec 3; 4:22 | utf-8 | BMC Neurol | 2,004 | 10.1186/1471-2377-4-22 | oa_comm |
==== Front
BMC NeurolBMC Neurology1471-2377BioMed Central London 1471-2377-4-221557920110.1186/1471-2377-4-22Case ReportAndrogen-induced cerebral venous sinus thrombosis in a young body builder: case report Sahraian Mohammad Ali [email protected] Mahmood [email protected] Amir Reza [email protected] Babak [email protected] Department of Neurology, Sina Hospital, Tehran University of Medical Sciences, Hassan Abad Square, Imam Khomini Street, Tehran, Iran2004 3 12 2004 4 22 22 21 9 2004 3 12 2004 Copyright © 2004 Sahraian et al; licensee BioMed Central Ltd.2004Sahraian 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
Cerebral venous sinus thrombosis is an infrequent disease with a variety of causes. Pregnancy, puerperium, contraceptive pills and intracranial infections are the most common causes. The patient may present with headache, focal neurological deficits and seizures.
The clinical outcome is highly variable and treatment with heparin is advised.
Case presentation
The patient is a 22 year old male who presented with headache, repeated vomiting and papilledema.
He was a bodybuilder doing exercise since 5 years ago, who had used nandrolone decaonoate 25 milligrams intramuscularly during the previous 5 months. Brain MRI and MRV showed superior sagital and transverse sinus thrombosis and extensive investigations did not reveal any known cause.
Conclusions
We suggested that androgen was the predisposing factor in our patient. Androgens may increase coagulation factors or platelet activity and cause arterial or venous thrombosis.
As athletes may hide using androgens it should be considered as a predisposing factor for thrombotic events in such patients.
==== Body
Background
Cerebral venous sinus thrombosis (CVST) is a disease with a wide spectrum of non specific clinical signs and symptoms, including headache, focal neurological deficits, seizures and coma.
The clinical outcome is highly variable; patients may recover completely or may develop severe and lasting neurological deficits [1].
There are many causes for this disease but the most common predisposing factors are pregnancy, puerperium, contraceptive pills, coagulopathies and intracranial infections [2].
There are few reports of patients with CVST after androgen therapy. We present a young bodybuilder man who developed CVST with abusing androgens for increasing muscle mass.
Case presentation
In May 2004 a 22 year old male was admitted to our department with chief complaints of headache and vomiting.
The patient was well till 10 days prior to admission that developed progressive, intense bitemporal headache exacerbated with bending.
The patient also had history of malaise, nausea and several episodes of vomiting from 3 days before admission. The only objective finding on physical examination was bilateral papilledema. The patient was a body builder doing exercise from 5 years ago who had used nandrolone decaonoate 25 mg once or twice a week during the last 5 months. He had injected 20 ampoules in this period.
Brain computed tomography without contrast was done for the patient which showed cord sign, emergency MR imaging including T1 – T2, weighted and MRV showed prominent superior sagital and transverse sinus thrombosis. The C.S.F opening pressure was 480 mm/H2o without any other abnormality. Heparin 80 IU/kg started as loading dose then continued 1000 IU/ hr for 10 days. On the 5th day of treatment headache resolved and warfarin added to heparin.
Laboratory tests including antithrombin III activity, protein C, S factor V leiden, Plasma hemocystein and anticardiolipin were all within normal limits.
The patient was discharged in a good condition and was maintained on 6 months warfarin protcol.
Discussion
There are few reports of CVST following androgen therapy [3], but there is just one reported case of CVST in androgen using young body builder [4]. The anabolic activity of testosterone and its derivatives is primarily manifested in its myotrophic actions which result in greater muscle mass and strength.
This has led to widespread use of androgenic anabolic steroids by athletes at all levels. Nandrolone decaonoate is a synthetic anabolic steroid. In focus on homeostasis system the most important factors under testosterone regulation are fibrinogen, Plasminogen activator inhibitor-1 (PAI – 1) and platelet aggregability.
The current data indicate that testosterone lowers fibrinogen and PAI – 1, however these anticoagulatory and profibrinolytic may be opposed by proaggregatory effects on platelets because high dosages of androgens were found to decrease cycloxygenase activity and thereby increase platelet functions [5].
Proaggregatory effect of testosterone and other synthetic androgens become more reliable theory for CVST, according to recent publication [6].
Conclusions
This case report presents a patient with CVST following exogenous androgen usage with a mechanism which is not completely understood, but it may be related to platelet activation or an increase in coagulation factors. As androgen use may be frequent and hidden in athletes, it may be an underestimated cause of cerebral venous thrombosis in young adults and careful history should be taken in these groups of patients.
Lists of abbreviations
CVST: cerebral venous sinus thrombosis
PAI-1: Plasminogen activator inhibitor-1
MRI: magnetic resonance imaging
MRV: magnetic resonance venography
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
M.A.S: Admitting and treating the case, preparing the article.
M.M : Revising the article.
A.R.A : searching previous articles, preparing case presentation part.
B.M : searching previous articles, preparing background.
Figure 1 T1 weighted horizontal nonGd Images show transverse sinus thrombosis.
Figure 2 T1 weighted sagital Images with Gd show thrombosis in sagital sinus.
Figure 3 2D MR venogram shows sagital sinus thrombosis.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are grateful to Mr. Hejrani for secretarial help and Mr. Solimani for preparing the photographs. It should be noted that written consent was obtained from the patient for publication of study.
==== Refs
Ameri A Bousser MG Cerebral venous sinus thrombosis Neurol clin 1992 10 87 111 1557011
Shelley Renowden Cerebral Venous sinus thrombosis Eur Radiol 2004 14 215 26 14530999 10.1007/s00330-003-2021-6
Shiozawa Z Yamada H Mabuchic Hotta T saito M Sobue I Huang YP Superior sagital sinus thrombosis associated with androgen therapy for hypoplastic anemia [abstract] Ann Neurol 1982 12 578 80 7159062
Jaillard AS Hommel M Mallaret M Venous sinus thrombosis associated with androgens in a healthy young man [abstract] Stroke 1994 25 212 13 8266370
Shahidi NT A review of the chemistry, biological action and clinical applications of anabolic androgenic steroids Clin therap 2001 23 1355 90 11589254 10.1016/S0149-2918(01)80114-4
Wu FC Von Eckardstein A Androgens and coronary artery disease Endocr Rev 2003 24 183 217 12700179 10.1210/er.2001-0025
| 15571626 | PMC539264 | CC BY | 2021-01-04 16:29:00 | no | BMC Fam Pract. 2004 Nov 30; 5:27 | latin-1 | BMC Fam Pract | 2,004 | 10.1186/1471-2296-5-27 | oa_comm |
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-5-471558827510.1186/1471-2474-5-47Research ArticleHypothalamic-pituitary-gonadal axis hormones and cortisol in both menstrual phases of women with chronic fatigue syndrome and effect of depressive mood on these hormones Cevik Remzi [email protected] Ali [email protected] Suat [email protected] Kemal [email protected] Ayşegül Jale [email protected] Physical Medicine and Rehabilitation, School of Medicine, Dicle University, Diyarbakir, Turkey2004 8 12 2004 5 47 47 16 6 2004 8 12 2004 Copyright © 2004 Cevik 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
Chronic fatigue syndrome (CFS) is a disease which defined as medically unexplained, disabling fatigue of 6 months or more duration and often accompanied by several of a long list of physical complaints. We aimed to investigate abnormalities of hypothalamic-pituitary-gonadal (HPG) axis hormones and cortisol concentrations in premenopausal women with CFS and find out effects of depression rate on these hormones.
Methods
We examined follicle stimulating hormone (FSH), luteinizing hormone (LH), estradiol, progesterone and cortisol concentrations in 43 premenopausal women (mean age: 32.86 ± 7.11) with CFS and compared matched 35 healthy controls (mean age: 31.14 ± 6.19). Patients were divided according to menstrual cycle phases (follicular and luteal) and compared with matched phase controls. Depression rate was assessed by Beck Depression Inventory (BDI), and patients with high BDI scores were compared to patients with low BDI scores.
Results
There were no significant differences in FSH, LH, estradiol and progesterone levels in both of menstrual phases of patients versus controls. Cortisol levels were significantly lower in patients compared to controls. There were no significant differences in all hormone levels in patients with high depression scores versus patients with low depression scores.
Conclusion
In spite of high depression rate, low cortisol concentration and normal HPG axis hormones of both menstrual phases are detected in premenopausal women with CFS. There is no differentiation between patients with high and low depression rate in all hormone levels. Depression condition of CFS may be different from classical depression and evaluation of HPG and HPA axis should be performed for understanding of pathophysiology of CFS and planning of treatment.
==== Body
Background
Chronic fatigue syndrome (CFS) is a clinical presentation that primarily affects women and characterized by severe disabling fatigue, and other symptoms; including musculoskeletal pain, sleep disturbance, impaired concentration, and headaches, in the absence of organic illness or severe psychiatric disorder that would explain the fatigue. For diagnosis of CFS operational criteria have been developed by Centers for Disease Control (CDC) [1]. Although the cause CFS is poorly understood, there are some causative theories for underlying conditions. Hypotheses about its etiology have included viral infections, immune dysregulation and abnormal endocrine function among others [2].
It has been reported that onset of CFS mostly following a significant stressor, most frequently a viral infection, and the course of the syndrome remits and relapses with occurrence of physical and psychological stressors [3]. Stress is known to interfere with the menstrual cycle and may lead to chronic anovulation and amenorrhea [4]. This generally thought to be caused by a decrease in the activity of hypothalamic gonadotropin releasing hormone (GnRH) pulse generator with subsequent inhibition of the pituitary-gonadal axis [5,6]. Stress-induced activation of hypothalamic-pituitary-adrenal (HPA) hormonal axis plays an important role in suppressing the HPG axis [7]. Infusion of corticotropin releasing hormone (CRH) into the cerebral ventricles leads to inhibition of LH secretion in primates [8]. CRH antagonism has also been shown to prevent the inhibitory effect of stress on the HPG axis in the rodent and in the monkey [9]. Women with hypothalamic amenorrhea have higher basal cortisol levels and blunted cortisol response to exogenous administration of CRH, suggesting that the increase in cortisol secretion may reflect increased endogenous CRH activity [10].
Pertuberations of HPA axis function have been described in CFS [11,12]. Studies of the HPA axis in CFS show a mild hypocortisolism of central origin, in contrast to the hypercortisolism of major depression [13,14]. There are similarites between onset, course, and clinical syndromes of CFS and glucocorticoid defficiency states. Clinical syndromes of CFS and Addison's disease share many common features: one of the principal clinical features of Addison's is fatigue, the core feature of CFS. The other common symptoms of CFS include arthralgia, myalgia, adenopathy, exacerbation of allergic responses, intermittent fever, postexertional fatigue, and depressed mood. These symptoms can also be experienced by those withdrawing from hypercortisolaemic states [15].
CFS occurs more commonly in women [16]. It was suggested that alterations in reproductive hormone levels might be involved in the pathoetiology of CFS [17]. There has been reported that this condition may be due to estrogen deficiency and reflect underactivity of the HPG hormonal axis [18]. The GnRH secretion from hypothalamus drives secretion of LH and FSH from pituitary gonadotropes [19]. FSH and LH govern the cyclical secretion of estradiol and progesterone over the course of menstrual cycle. The pulsatile pattern of GnRH secretion is critical for the control of serum LH, FSH, and ovulation.
Interaction between HPA and HPG axes in stress-induced amenorrhea suggests that there may be perturbation of these axes in CFS. One important confound is co-morbid depressive illness, present in approximately 50% of CFS patients [20]. Relation between depressive mood and these axes has been more investigated yet. This relationship may contribute to clarification of pathoetiology of CFS and to describe treatment strategy of this complex syndrome.
In this study, we aimed to investigate main hormones of HPG and HPA axes; FSH, LH, estradiol, progesterone and cortisol levels in premenopausal women with CFS. Furthermore to find out relationship between these hormones and depressive mood with comparing patients with high and low depression scores.
Methods
Subjects
A total of 43 premenopausal women diagnosed as CFS, according to the international CFS definition criteria [1], were recruited from outpatient clinic of Physical Medicine and Rehabilitation Department of Dicle University (Diyarbakýr, Turkey) for this study. Fatigue assessment was done according to CDC criteria [1]. Fatigue characteristics are persistent or recurrent lasting at least 6 months; recent and/or well defined onset; not secondary to excessive physical activity, or any organic or psychiatric disorders; not resolved by rest; and inducing important reduction of previous levels of physical and mental activities. Thirty five age matched demographically similar healthy premonopausal women were also selected as controls. Ethic committee of Dicle University hospital approved the study, and all subjects voluntarily agreed to participate. All patients underwent medical screening that included physical examination and relevant investigation, with a minimum of urine analysis, full blood count, measurement of urea, electrolytes, and erythrocyte sedimentation rate, and test for thyroid and liver function. All patients and controls were evaluated by structered psychiatric interview to exclude any additional psychiatric disorder prior to inclusion in the study. Depression rate was assessed by Beck Depression Inventory (BDI) in all patients and controls. Patients with CFS were divided to two groups according to the BDI higher and lower than 17 scores. All prescription medications, included psychoactive and non-prescription medications, vitamins, and herbal remedies were tapered and then stopped at least 2 weeks prior to study [17,21]. All subjects and controls had no frank hypocortisolism on endocrine assessment. No patients and controls had received any oral or intraarticular corticosteroid therapy during the three months preceding the study, or had received exogenous estrogens, progesterone, or any other drugs affecting sex or cortisol hormones metabolism. The other exclusion critetria [1] were: active, unresolved, or suspected disease likely to cause fatigue; alcohol or other substance abuse within 2 years prior to onset of the chronic fatigue and any time afterward; and a body mass index ≥45. All subjects and controls were undergoing normal menstrual cycles and were not taking contraceptive pill. Subjects and controls were all studied during the follicular (n: 28 with CFS and n: 23 controls) and luteal (n: 15 with CFS and n: 12 controls) phases of menstrual cycle.
Procedures and Hormone Assays
Blood samples were collected in the early morning (8.30–10.30 AM) after an all night fast and plasma was separated immediately by centrifugation; then sera obtained were stored at -20°C until hormonal assaying. All hormone values assayed by "Electro Chemil Luminescence Immunassay (ECLIA)" (Roche, 1010/1020 Elecsys Systems Immunoassay) method. Serum concentrations of follicle stimulating hormone (FSH, normal values 3.3 to 11.3 mIU/mL for follicular phase and 1.8 to 8.2 mIU/mL for lueal phase), luteinizing hormone (LH, normal values 2.4 to 12.6 mIU/mL for follicular phase and 1.0 to 11.4 mIU/mL for luteal phase), estradiol (E2, normal values 24.5 to 195 pg/mL for follicular phase and 40 to 261 pg/mL for luteal phase), progesterone (normal values 0–1.6 ng/mL for follicular phase and 1.1 to 21 ng/mL for luteal phase), and cortisol (normal values 6.2 to 19.4 μgr/dl) were evaluated in all patients and controls in both of menstrual periods.
Statistical analysis
were done by SPSS 8.0 PC program. Results were expressed as means ± SD (standard deviation). Statistical significances were tested using the independent student's t test for women with CFS and control group comparisons. Mann Whitney U test was used for comparison of hormonal data of women in both phases of menstrual cycle, and patients with high and low BDI scores groups comparison. The level of statistical significance was set at a two-tailed p-value of 0.05.
Results
All patients with CFS and healthy controls were reproductive premenopoausal women; and mean ages were 32.76 ± 7.07 and 31.14 ± 6.19 respectively for both groups. All of the patients had debilitating clinically evaluated and medically unexplained fatingue that does not resolve with bed rest and severe enough to significantly reduce daily activity for at least 6 months. The other clinical findings of patients with CFS were summarized in Table 1.
There wasn't significant differentiation between ages and BMI of groups (p > 0.05). There were no significant differences between the means of FSH, LH, progesterone and estradiol in all CFS patients compared to all controls (p > 0.05). Mean concentrations of cortisol were significantly lower in all CFS patients than all controls (p < 0.001) (Table 2).
There were no significant differences between the levels of FSH, LH, progesterone and estradiol in CFS patients compared to controls in follicular phase (p > 0.05). Mean concentrations of cortisol were significantly lower in CFS patients than controls in follicular phase (p = 0.001) (Table 3).
There were no significant differences between the levels of FSH, LH, progesterone and estradiol in CFS patients compared to controls in luteal phase (p > 0.05). Mean concentrations of cortisol were significantly lower in CFS patients than controls in luteal phase (p < 0.05) (Table 4).
Fifty patients (69.76%) with CFS had ≥17 BDI scores. Mean cortisol concentrations of patients with BDI scores ≥17 were higher than patients with BDI scores <17 but differentiation wasn't significant. There were no significant differences between these patients groups in HPG axis hormone levels (Table 5).
Discussion
In this study, reproductive HPG axis hormone levels demonstrated no significant differences in women with CFS from controls during follicular and luteal phases. These findings are in agreement with Korszun et al. [17] who reported data from 9 premenopausal women with fibromyalgia and 8 with CFS. They showed no significant differentiations of reproductive axis function in both of patients groups in estrogen and progesterone levels, as well as LH pulsatility during the follicular phase. However, Studd and Panay [18] reported data from 28 premenopausal women with CFS 87 and of these, 25% showed low plasma estradiol concentrations. The authors reported that CFS may represent an hypoestrogenic state and recommend the use of hormone replacement therapy for women with CFS. In addition, they claim that 80% of patients improved after treatment of estradiol patches and cyclical progestagens.
Chronic fatigue syndrome is generally accepted stress related disease and disfunction of HPA axis has been reported in this syndrome [22]. In this study, cortisol levels were lower in women with CFS compared to healthy controls. Some sudies of HPA in CFS show a mild hypocortisolism of central origin in contrast to hypercortisolism of major depression [13,14]. In an early study of the HPA axis in patients with CFS Demitract et al. [22] reported low 24-hour urine free cortisol compared with that of control subjects. Baseline evening plasma corticotropin levels were elevated and cortisol levels were depressed. Significantly lower baseline cortisol levels was reported in an earlier study [23]. Despite these findings, the majority of further studies have failed to replicate this. Differentiations of studies methodology and sample characteristics may explain the variety of results.
High circulating cortisol is well replicated finding in major depression [24] and so presence of depression makes the cortisol findings more difficult to interpret. Significantly raised baseline cortisol levels in subjects with CFS studied by Wood et al. [25] was explained by their high BDI scores [20]. In this study, high BDI scores (≥17) were detected in 69.76% of patients with CFS. There were no significant high level of cortisol and HPG axis hormone concentrations in patients with high BDI score compared to patients with low BDI scores (<17). Scott and Dinan [14] reported finding of low urine free cortisol in patients with CFS compared with healthy controls. In addition, there was no difference between depressed and non depressed patients with CFS. These findings are in agreement with our study. In another study [26] was reported blunted corticotropin and cortisol in response to administration of ovine CRH without differences in basal levels.
Studies in primates have demonstrated that intracerebroventricular infusion of CRH as well as proinflammatory cytokines such as interleukin-1 can decrease LH secretion [27,28]. Stress induced (hypothalamic) amenorrhea, as well as exercise-induced amenorrhea and anorexia nervosa, activate the HPA axis, increasing cortisol secretion and decreasing corticotropin or cortisol response to exogenous CRH [29-31]. These HPA axis abnormalities are similar to those seen in depression, suggesting that activation of the HPA axis may be linked to inhibition of the HPG axis. Young et al. [32] found %30 lower plasma estradiol level in women with depression than controls in the follicular phase. The half-life LH was significantly shorter in women with depression than controls during both of the follicular and luteal phases. The other reproductive hormones were normal in women with depression compared to control women in both the phases of menstrual cycle.
In this study was found significantly lower circulating cortisol levels in patients with CFS in contrast to high BDI scores. However, there was no significant differentiation in cortisol levels between the patients with low and high depression scores. This is in contradiction to those hypercortisolism of classical major depression and stress condition. More recent studies support this contradicton [21,33]. In recent years, however, it has become increasingly apparent that deppression is a heterogenous condition from both a psychological and a physiological perspective [34]. Moreover, decreased HPA axis activity was reported in some stress related states such as CFS, atypical and seasonal depression [35]. These results suggest that depression condition which is seen in CFS may be different from classical depression. There may be overlapping between symptoms of CFS and those depressive subtypes or reactive form of depression in CFS. This condition may explaine both hypocortisolism in patients with CFS and the lack of HPG axis hormone abnormalities in this study.
Determining single basal levels of HPA and HPG axes do not reflect activity of these axes entirely. Dynamic test indicates differences in function of these axes. We did not carry out dynamic characteristics of these axes. This point is limit of our study.
In conclusion, we detected low cortisol levels in patients with CFS in spite of their high depressive mood rate. However, in this study, we were unable to describe HPG axis hormone abnormalities in both menstrual phases. Hypocortisolism may be a biological factor that contributes fatigue chronicity and the reason of normality in HPG axis. Depressive mood of chronic fatigue syndrome may be different from classical depression. There are need to carry out future controlled and larger clinical trials to clarify these matters.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Table 1 Clinical findings of 43 patients with CFS
(N) (%)
Fatigue 43 100
Sore Throat 32 74.42
Arthralgıa 34 79.07
Myalgıa 39 90.7
Muscle Weaknes 7 16.27
Fever 24 55.8
Fharyngitis 32 74.42
Lenfadenopathy 31 72.1
Headache 23 53.5
Neuropsychologic findings 43 100
Sleep Disturbance 31 72.1
Table 2 HPG axis hormonal data and cortisol levels in all patients and controls
Patients (n = 43) Controls (n = 35) T p
Age 32.76 ± 7.07 31.14 ± 6.19 1.06 0.290
BMI 24.47 ± 4.37 24.63 ± 4.39 -0.147 0.883
BDI 21.83 ± 8.49 11.6 ± 3.45 6.68 0.000
FSH 6.57 ± 4.87 6.52 ± 7.45 0.034 0.973
LH 6.82 ± 5.72 7.09 ± 6.04 -0.202 0.841
Progesterone 2.23 ± 3.14 3.15 ± 4.21 -1.105 0.273
Estradiol 111.54 ± 176.83 89.08 ± 55.37 0.722 0.472
Cortisol 10.33 ± 4.55 15.11 ± 6.95 -3.64 0.000
Table 3 HPG axis hormonal data and cortisol levels in patients and controls in follicular phase
Patients (n = 28) Controls (n = 23) T P
Age 32.71 ± 7.43 30.74 ± 6.39 -1.17 0.240
BDI 22.85 ± 9.04 12.3 ± 3.26 -4.71 0.000
FSH 6.73 ± 5.32 7.91 ± 8.84 -0.142 0.88
LH 7.6 ± 6.37 8.34 ± 6.85 -0.58 0.55
Progesterone 1.96 ± 2.95 2.10 ± 3.56 -0.22 0.83
Estradiol 129.65 ± 217.05 89.66 ± 58.21 -0.21 0.83
Cortisol 10.12 ± 4.46 15.26 ± 7.17 -3.25 0.001
Table 4 HPG axis hormonal data and cortisol levels in patients and controls in luteal phase
Patients (n = 15) Controls (n = 12) T P
Age 32.86 ± 6.58 31.91 ± 5.99 -0.34 0.75
BDI 19.93 ± 7.26 10.25 ± 3.57 -3.30 0.000
FSH 6.26 ± 4.07 3.87 ± 1.99 -1.66 0.103
LH 5.36 ± 4.03 4.69 ± 3.05 -0.53 0.61
Progesterone 2.72 ± 3.5 5.19 ± 4.76 -1.415 0.167
Estradiol 77.73 ± 32.61 87.97 ± 51.94 -0.293 0.792
Cortisol 10.33 ± 4.3 14.81 ± 6.81 -2.001 0.047
Table 5 Cortisol levels in women with CFS according to BDI higher and lower than 17 scores
BDI ≥17 scores (n = 30) BDI scores <17 (n = 13) T P
Age 33.76 ± 6.89 30.46 ± 7.20 -1.52 0.130
BDI 25.93 ± 6.7 12.38 ± 2.29 -5.17 0.000
FSH 6.62 ± 5.36 6.46 ± 3.72 -0.212 0.84
LH 6.71 ± 6.25 7.08 ± 4.5 -0.80 0.38
Progesterone 2.08 ± 2.92 2.56 ± 3.69 -0.42 0.68
Estradiol 127.46 ± 206.32 74.79 ± 66.81 -1.45 0.151
Cortisol 10.37 ± 4.65 9.78 ± 3.8 -0.42 0.68
==== Refs
Fukuda K Straus S Hickie I Sharpe MC Dobbins JG Komaroff A The chronic fatigue syndrome: a comprehensive approach to its definition and study. International Chronic Fatigue Syndrome Study Group Ann Intern Med 1994 121 953 959 7978722
Kastin AJ Olson RD Martins JM Yehuda S, Mostofsky DI Chronic fatigue syndrome: Possible integration of hormonal and immunological observations in: Chronic fatigue syndrome 1997 New York, Plenum Press 161 192
Demitrack MA Engleberg NC Chronic fatigue syndrome Curr Ther Endocrinol Metab 1994 5 135 142 7704698
Lachelin GC Yen SS Hypothalamic chronic anovulation Am J Obstet Gynecol 1978 130 825 31 147631
Rivier C Rivest S Effects of stress on the activity of the hypothalamic-pituitary-gonadal axis: peripheral and central mechanisms Biol Reprod 1991 45 523 32 1661182
Ferin M Stress and reproductive cycle J Clin Endocrinol Metab 1999 84 1768 74 10372662 10.1210/jc.84.6.1768
Rivest S Rivier C The role of corticotropin-releasing factor and interleukin-1 in the regulation of neurons controlling reproductive functions Endocr Rev 1995 16 177 99 7781593 10.1210/er.16.2.177
Xiao E Luckhaus J Nieman W Ferin M Acute inhibition of gonadotropin secretion by corticotropin-releasing hormone in the primate: are the adrenal glands involved? Endocrinology 1989 124 1632 1637 2494035
Feng YJ Shalts E Xia LN Rivier J Rivier C Vale W Ferin M An inhibitory effect of interleukin-1α on basal gonadotropin release in the ovariectomized rhesus monkey: reversal by a corticotropin-releasing factor antagonist Endocrinology 1991 128 2077 82 1900786
Bilier BM Federoff HJ Koenig JI Klibanski A Abnormal cortisol secretion and responses to corticotropin-releasing hormone in women with hpothalamic amenorrhea J Clin Endocrinol Metab 1990 70 311 317 2153693
Neeck G Crofford LJ Neuroendocrine perturbations in fibromyalgia and CFS Rheumatic Disease Clinics of North America 2000 26 989 1002 11084955
Racciatti D Guagnano MT Vecchiet J De Remigis PL Pizzigallo E Della Vecchia R Di Sciascio T Merlitti D Sensi S Chronic fatigue syndrome: circadian rhythm and hypothalamic-pituitary-adrenal (HPA) axis impairment Int J Immunopathol Pharmacol 2001 14 11 15 12622884
Cleare AJ Bearn J Allain T McGregor A Wesseley S Murray RM O'Keane V Contrasting neuroendocrine responses in depression and chronic fatigue syndrome J Affective Disorders 1995 35 283 289 10.1016/0165-0327(95)00026-J
Scott LV Dinan TG Urinary free cortisol excretion in chronic fatigue syndrome, major depression, and in healthy volunteers J Affect Disord 1998 47 49 54 9476743 10.1016/S0165-0327(97)00101-8
Avgerinos PC Chrousos GP Nieman LK Oldfield EH Loriaux DL Cutler BG The corticotrophin-releasing hormone test in the postoperative evaluation of patients with cushing's syndrome Clin Endocrinol Metab 1987 65 906 13
Gunn WJ Connell DB Randall B Bock GR, Whelan J Epidemiology of chronic fatigue syndrome: The Centre for Disease Control study Chronic fatigue syndrome 1993 Chichester: John Wiley & Sons 83 101
Korszun A Young AE Engleberg NC Masterson L Dawson EC Spindler K McClure LA Brown MB Crofford LJ Follicular phase hypothalamic-pituitary-gonadal axis function in women with fibromiyalgia and chronic fatigue syndrome J Rheumatol 2000 27 1526 1530 10852283
Studd J Panay N Chronic fatigue syndrome Lancet 1996 348 1384 8918295
Knobil E The GnRH pulse generator Am J Obstet Gynecol 1990 163 1721 7 2122728
Parker AJR Wessely S Cleare AJ The Neuroendocrinology of CFS and fibromyalgia Psychological Medicine 2001 31 1331 45 11722149 10.1017/S0033291701004603
Gur Ali Cevik Remzi Nas Kemal Colpan Leyla Sarac Serdar Cortisol and hypothalamic-pituitary-gonadal axis hormones in follicular phase women with fibromyalgia and chronic fatigue syndrome and effect of depression on these hormones Arthritis& Res Ther 2004 6 232 8 10.1186/ar1163
Demitrack MA Dale JK Strauss SE Laue L Listwak SJ Kruesi MJ Chrousos GP Gold PW Evidence for impaired activation of the hypothalamic-pituitary-adrenal axis in patients with chronic fatigue syndrome J Clin Endocrinol Metab 1991 73 1224 34 1659582
Poteliakhoff A Adrenocortical activity and some clinical findings in chronic fatigue Journal of Psychosomatic Research 1981 25 91 95 6974238 10.1016/0022-3999(81)90095-7
Dinan TG Glucocorticoids and the genesis of depressive illness. A psychobiological model Br J Psychiatry 1994 164 365 71 7832833
Wood B Wessely S Papadopoulos A Poon L Checkley S Salivary cortisol profiles in CFS Neuropsychobiology 1998 37 1 4 9438265 10.1159/000026469
Scott LV Medback S Dinan G Blunted adrenocrticotropin and cortisol responses to corticotropin-releasing hormone stimulation in CFS Acta Psychiat Scand 1998 97 450 9669518
Olster DH Ferin M Corticotropin-releasing hormone inhibits gonadotropin secretion in the ovariectomized rhesus monkey J Clin Endocrinol Metab 1987 65 262 67 3110201
Xiao E Xia-Zhang L Barth A Zhu J Ferin M Stress and menstrual cycle quality in the short and long term response to a five day endotoxin challenge during the follicular phase in the rhesus monkey J Clin Endocrinol Metab 1998 83 2454 60 9661628 10.1210/jc.83.7.2454
Loucks AB Mortola JF Girton L Yen SS Alteration in the hypothalamic-pituitary-ovarian and the hypothalamic-pituitary-adrenal axes in athletic women J Clin Endocrinol Metab 1989 68 402 11 2537332
Hohtari H Salminen-Lappalainen K Laatikainen T Response of plasma endorphins, corticotropin, cortisol, and LH in the corticotropin-releasing hormone stimulation test in eumenorrheic and amenorrheic athletes Fertil Steril 1991 552 176 80
Gold PW Gwirtsman H Avgerinos P Nieman LK Galluci WT Kaye W Jimerson D Ebert M Rittmaster R Loriaux DL Abnormal hypothalamic-pituitary-adrenal function in anorexia nervosa: pathophysiological mecanisms in underwight and weight corrected patients N Eng J Med 1986 314 1335 37
Young EA Midgley AR Carlson NE Brown MB Alteration in the Hypothalamic-Pituitary-Ovarian Axis in Depressed Women Arch Gen Psychiatry 2000 57 1157 62 11115329 10.1001/archpsyc.57.12.1157
Gür Ali Çevik Remzi Saraç Jale Em Serda Sarac Serdar Hypothalamic-pituitary-gonadal axis and cortisol in young females with fibromyalgia: the potential roles of depression, fatigue and sleep disturbance in the occurrence of hypocortisolism Ann Rheum Dis 2004 63 1504 1506 15479904 10.1136/ard.2003.014969
Crofford LJ The hypothalamic-pituitary-adrenal stress axis in fibromyalgia and CFS Z Rheumatol 1998 57 67 71 10025087 10.1007/s003930050239
Tsigos C Chrousos GP Hypothalamic-pituitary-adrenal axis, neuroendocrine factors and stress Journal of Psychosomatic Research 2002 53 865 71 12377295 10.1016/S0022-3999(02)00429-4
| 15588275 | PMC539265 | CC BY | 2021-01-04 16:03:43 | no | BMC Musculoskelet Disord. 2004 Dec 8; 5:47 | utf-8 | BMC Musculoskelet Disord | 2,004 | 10.1186/1471-2474-5-47 | oa_comm |
==== Front
BMC GastroenterolBMC Gastroenterology1471-230XBioMed Central London 1471-230X-4-301557920010.1186/1471-230X-4-30Research ArticleTGF β1 and PDGF AA override Collagen type I inhibition of proliferation in human liver connective tissue cells Geremias Alvaro T [email protected] Marcelo A [email protected] Radovan [email protected] Alvaro NA [email protected] Departamento de Bioquímica, Instituto de Química, Universidade Federal do Rio de Janeiro, Rio de Janeiro 21949, Brazil2 Laboratório de Metabolismo Macromolecular Firmino Torres de Castro, Instituto de Biofísica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro 21941, Brazil3 Departamento de Histologia e Embriologia, Instituto de Ciências Biomédicas, Universidade Federal do Rio de Janeiro, Rio de Janeiro 21941, Brazil2004 3 12 2004 4 30 30 16 8 2004 3 12 2004 Copyright © 2004 Geremias et al; licensee BioMed Central Ltd.2004Geremias 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 marked expansion of the connective tissue population and an abnormal deposition of extracellular matrix proteins are hallmarks of chronic and acute injuries to liver tissue. Liver connective tissue cells, also called stellate cells, derived from fibrotic liver have been thoroughly characterized and correspond phenotypically to myofibroblasts. They are thought to derive from fat-storing Ito cells in the perisinusoidal space and acquire a contractile phenotype when activated by tissue injury. In the last few years it has become evident that several peptide growth factors such as PDGF AA and TGF-β are involved in the development of fibrosis by modulating myofibroblast proliferation and collagen secretion. The fact that during the development of chronic fibrosis there is concomitant deposition of collagen, a known inhibitory factor, and sustained cell proliferation, raises the possibility that stellate cells from chronic liver fibrosis patients fail to respond to normal physiologic controls.
Methods
In this study we address whether cells from fibrotic liver patients respond to normal controls of proliferation. We compared cell proliferation of primary human liver connective tissue cells (LCTC) from patients with liver fibrosis and skin fibroblasts (SF) in the presence of collagens type I and IV; TGF-β, PDGF AA and combinations of collagen type I and TGF-β or PDGF AA.
Results
Our results indicate that despite displaying normal contact and collagen-induced inhibition of proliferation LCTC respond more vigorously to lower concentrations of PDGF AA. In addition, we show that collagen type I synergizes with growth factors to promote mitogenesis of LCTC but not SF.
Conclusions
The synergistic interaction of growth factors and extracellular matrix proteins may underlie the development of chronic liver fibrosis.
==== Body
Background
In normal liver, connective tissue cells are rare and mostly restricted to the periportal and pericentrovenular spaces and to the Glisson's capsule. A minor population of connective tissue cells is present inside the hepatic lobule, in the perisinusoidal space. They are known as hepatic stellate cells (HSC) and are considered to be one of the major contributors for fibrogenesis in liver [1]. Chronic and acute injuries to liver tissue induce a marked expansion of the connective tissue cells and concomitantly an abnormal deposition of extracellular matrix proteins [2]. Extensive studies of experimental models and humans have shown that these cells are of the myofibroblast lineage, characterized by the expression of smooth muscle α-actin [3-5]. In fact, there is now increasing evidence that are several populations of myofibroblasts in the diseased liver in addition to those derived from HSC [6,7]. The origin of these cells is still debated. In experimental models, it is considered that an acute liver injury is followed by activation and increase of stellate cells, while chronic injuries elicit activation of similar cells that can be either of lobular or periportal origin [7]. In humans, extensive primary periportal fibrosis such as schistosomal fibrosis was related to expansion of myofibroblasts originating from the portal vein wall [8], while in cirrhosis resident periportal connective tissue cells, as well as lobular stellate cells and pericentrovenular cells were reported to convert into myofibroblasts [3,9,10]. Independently of their origin, connective tissue cells of human fibrotic and cirrhotic liver tissues were shown to have homogeneous morphologic characteristics, as well as a specific growth patterns [11-15].
Recent studies on the mechanisms controlling connective tissue cell expansion and abnormal collagen deposition during liver fibrosis have shown that several peptide growth factors are involved in the development of fibrosis. Evidence from in vivo and in vitro studies indicated that TGF-β has a relevant role in the development of fibrosis, being a potent regulator of proliferation and of extracellular matrix protein synthesis [16-18]. TGF-β is highly expressed in fibrotic regions and overexpression of TGF-β in animal models results in fibrosis [18,19]. Conversely, TGF-β antagonists can prevent experimentally induced fibrosis [20-22]. PDGF AA was shown to have a pivotal role in controls of normal and pathologic proliferation of two myofibroblastic cell lineages: mesangial cells in kidney and alveolar smooth muscle cells in lungs [23,24].
On the other hand, it has also been demonstrated in vitro that the extracellular matrix can in its turn modulate cellular responses to peptide growth factors [25]. In vitro, fibroblast proliferation can be regulated by the presence of collagen type I. It has been suggested that collagen induces a quiescent state by decreasing the responsiveness to different growth factors [26,27]. An increase in cell density is also known to inhibit fibroblast proliferation in vitro. During the development of chronic fibrosis there is concomitant deposition of collagen and sustained cell proliferation, raising the possibility that liver connective tissue cells (LCTC) from chronic liver fibrosis patients fail to respond to normal physiologic controls.
In this study we addressed the question of whether cells from fibrotic liver patients respond to normal physiologic controls of proliferation. We compared the behavior of primary human LCTC from patients with liver fibrosis with skin fibroblasts (SF). We compared cell proliferation in the presence of collagen types I and IV, of TGF-β and PDGF AA, and in combinations of collagen type I and TGF-β or PDGF AA. Our results indicate that LCTC respond more vigorously to lower concentrations of peptide growth factors than SF. Interestingly, collagen type I synergizes with the growth factors tested in promoting mitogenesis of LCTC but not of SF. We believe that this mechanism may underlie the development of chronic liver fibrosis.
Methods
Cell lines
Liver tissue fragments and skin biopsies were obtained through collaboration with the University Hospital, Department of Surgery, Federal University of Rio de Janeiro (O.M. Vieira, M.D.). An informed consent was obtained from every patient, and the study was conducted in agreement with the ethical guidelines of the Federal University of Rio de Janeiro. Diagnoses were given by the pathological anatomy service of the hospital. All surgical liver biopsies were obtained for diagnostic purposes during spleno-renal or porto-caval anastomosis. Skin fibroblasts were obtained from the mid-abdominal incision performed for the surgery. We studied six primary normal SF lines and ten primary LCTC lines derived from patients with schistosomal fibrosis or alcoholic cirrhosis.
Biopsies were collected and brought to the laboratory in chilled Dulbecco's modified Eagle's medium (DMEM) (Sigma, St. Lois, MO) with 10% fresh human serum and 60 μg/ml of gentamicin (Schering, Rio de Janeiro, Brazil). They were cut into pieces of approximately 1 mm3, washed in fresh medium and plated, 6 or 9 pieces per 25 cm2 flask. They were maintained in DMEM supplemented with 4 g/liter HEPES and 10% fresh human serum, at 37°C in a humid incubator with 5% CO2, 95% air. All the cultures used in this study were finite cell lines derived from the primary cultures described above, and were discarded before the 10th passage.
LCTC have been described in detail previously, and compared to skin fibroblasts and vascular smooth muscle cells [12,13,15]. They were characterized by electron microscopy, immunofluorescence, morphology and proliferation in culture and ability to contract a collagen matrix. These established primary cell cultures are homogeneous in terms of production of smooth muscle α-actin, fibronectin, collagen I and III, and elements of the basement membrane such as collagen IV and laminin [12,13,15]. Hence, they can be phenotypically characterized as myofibroblasts and different from fibroblasts or smooth muscle cells.
Serum and growth factors
Fresh human citrated plasma was obtained from the hemotherapy service of the Hospital dos Servidores do Estado (Rio de Janeiro, Brazil). Plasma was coagulated with calcium and sera from several patients were pooled. Human TGF-β1 and human recombinant PDGF-AA were purchased from Sigma Chemical Company, St Louis, MO. TGF-β1 was activated by incubation in bovine serum albumin 1 mg/ml and 4 mM HCl.
Preparation of collagen coated dishes
One hundred μl of solutions with varying concentrations of collagen type I (rat tail tendon collagen, prepared in our laboratory as previously described [13]) or collagen type IV (Sigma) were dispensed onto a 12-well or a 96-well plate and dried under a laminar flow for 24 hr at room temperature. This procedure allowed us to control for the exact amount of collagen added to each plate. The plates used were freshly prepared and were washed three times in serum free medium.
Proliferation
Cell proliferation was assayed by cell counting and [H3]-thymidine incorporation. These methods correlated well with autoradiography results in previous studies [15]. For the experiments to assay contact inhibition of proliferation, approximately 1 × 103 cells/cm2 were plated. Each day, during a period of 8 days, cells were trypsinized and counted in a hemocytometer and medium was changed every two days. To assay the proliferative response induced by growth factors and extracellular matrix proteins cells were trypsinized and plated at a concentration of 5 × 103 and allowed to adhere for 2 hr in DMEM with 10% serum. Then, cells were starved in serum-free medium for 18 hr to synchronize the cell population. When testing for growth factors effects, serum-free medium was replaced by medium containing 1% serum, supplemented with the growth factors and 0.5 μCi/ml [H3]-thymidine. When assessing the effect of extracellular matrix proteins, serum-free medium was replaced by medium containing 10% serum. Cells were collected 24 or 32 hr after the stimulus and lysed in 10 N NaOH. The trichloroacetic acid precipitable material was then spotted on a filter and counted on a scintillation counter.
Kinetics of thymidine incorporation
Cells were plated at a concentration of 2.5 × 103 per well (0.5 cm2) in 96 well plates in DMEM with 10% serum. Cells were allowed to adhere and spread for two hr. Medium was then replaced with serum-free medium for 24 hr. Subsequently, cells were incubated in different media containing 1% serum and growth factors. [H3]-thymidine (1 μCi/ml) was added at 6, 24, 32 and 40 hr and incubated for 2 hr. Cells were harvested and processed as described above.
Adhesion and recovery
To determine cell adhesion on different substrata plates were prepared with 0.3 mg/ml collagen type I or IV. Cells were plated on each substratum and the supernatant was removed at 0, 10, 20, 40 and 60 min and the cells remaining in suspension counted in a hemocytometer. To determine cell recovery, cells plated after 2 hr were trypsinized and counted.
Statistical analysis
Difference between the means of various subgroups was assessed using the Mann-Whitney U-test.
Results
Chronic liver fibrosis is characterized by abnormal proliferation of connective tissue cells. Therefore, we hypothesized that connective tissue cells from fibrotic lesions have lost normal proliferation controls. By using a series of primary connective tissue cell lines from skin and from fibrotic livers, we investigated several parameters of cell growth in vitro in order to identify potential mechanisms to explain the excessive proliferation of LCTC during fibrosis. We observed no differences among the cell lines obtained from patients with schistosomal fibrosis or alcoholic cirrhosis, and all the studied primary LCTC lines were included in a single experimental series. We established and characterized 16 primary human cell lines. The experiments performed were done at or before the tenth passage and every experiment was performed with at least two cell lines from each type (LCTC or SF) isolated from different individuals.
Contact inhibition of proliferation
To assess whether connective tissue cells from fibrotic livers (LCTC) and normal skin fibroblasts (SF) responded to contact inhibition of proliferation, we measured cell proliferation during eight days. Both LCTC and SF showed a comparable pattern of growth and reached a plateau at a similar cell density (Figure 1A,1B). Reproducibly, LCTC declined in number after reaching a maximal density (Figure 1B, day 8). To confirm these results, we measured [H3]-thymidine incorporation in various cell densities (Figure 1C). Both LCTC and SF showed maximum [H3]-thymidine incorporation at the same cell density (1 × 104/cm2) and a marked decline at higher densities. We also examined the response of both cell types to increasing concentrations of fresh human serum. Both SF and LCTC reached maximum proliferation, as measured by [H3]-thymidine incorporation at 20% serum (not shown). In conclusion, primary SF and LCTC were similarly responsive to normal contact inhibition of proliferation and serum concentration, although SF seemed to respond more efficiently to contact inhibition.
Figure 1 Contact inhibition of proliferation. A. Proliferation of normal human skin fibroblasts (SF) in vitro. B. Proliferation of human liver connective tissue cells (LCTC). Cells were plated at 1 × 103 cells/cm2. Each day, during a period of 8 days, cells were trypsinized and counted in a hemocytometer (n = 3). Medium containing 10% serum was replaced every two days. C. [H3]-Thymidine incorporation in SF and LCTC at increasing cell density (n = 4).
The effects of extracellular matrix
Despite the abnormal deposition of extracellular matrix in chronic fibrotic livers, LCTC continuously proliferate suggesting that these cells are not responsive to normal inhibition of proliferation by collagen I. To determine the influence of the extracellular matrix on LCTC, we analyzed DNA synthesis in the presence of collagen type I and type IV, two major components of extracellular matrix in fibrotic livers. Collagen type I inhibited DNA synthesis in a dose-dependent manner, as measured by [H3]-thymidine incorporation in LCTC and SF (Figure 2A,2B). In SF, [H3]-thymidine incorporation at the highest collagen concentration was approximately 30% of the control, suggesting that this cell type is more susceptible to collagen-mediated inhibition of proliferation. However, the level of the inhibition was variable among the studied patients, and a less marked reduction, (~50%) was observed in SF derived from some of the studied cases (not shown). We did not observe any change in cell morphology dependent upon the substrate.
Figure 2 Collagen-mediated inhibition of proliferation. A. Cell proliferation of LCTC plated on increasing collagen concentration as measured by [H3]-Thymidine incorporation (n = 4). Note inhibition of incorporation in a dose-dependent manner. B. Cell proliferation of SF plated on increasing collagen concentration as measured by [H3]-Thymidine incorporation (n = 4). Note inhibition of incorporation in a dose-dependent manner. (*) denotes significant difference from control; p ≤ 0.05. C. Left panel. Kinetics of DNA synthesis of LCTC comparing growth on plastic (empty circles) and on collagen (filled circles) (n = 4). Right panel, top. Adhesion of cells plated on plastic (filled circles), collagen IV (empty triangles) and collagen I (empty squares) (n = 3). Right panel, bottom. Recovery of cells from plastic or collagen I (n = 4).
The question of whether the difference observed between collagen-coated and plastic plates resulted from an alteration in kinetics of cell cycle progression or to an actual block in DNA synthesis was addressed in the following experiment. After serum-mediated activation of quiescent cells, the cultures were pulsed with [H3]-thymidine for 2 hr, at 6, 20, 26, 32 and 40 hr. [H3]-thymidine incorporation reached its peak at 32 hours after activation in control experiments (plated on plastic), and cells plated on collagen-coated plates did not show any delayed peak, demonstrating that the effect of collagen was not delaying the progression of cell cycle but blocking DNA synthesis (Figure 2C). These results indicate that LCTC are responsive to normal inhibition of proliferation mediated by collagen I.
Next, we asked whether collagen type IV could also influence the proliferation of the studied cells. Even at high concentrations (90 μg/cm2), collagen type IV did not induce any significant change on [H3]-thymidine incorporation in two LCTC cell lines, neither in two SF cell lines (not shown). Again, no difference in morphology was noted.
To rule out the possibility that the results were being masked by differential adhesion to collagen or differential cell recovery when measuring the [H3]-thymidine incorporation, we monitored cell adhesion on plastic, on collagen type I and on collagen type IV. Although by 10 min there were a significantly less cells attached on collagen IV, by 40 min the same number of cells had attached to the plates in all substrates (Figure 2C, right panel). These results rule out the possibility that lower [H3]-thymidine incorporation found in cells plated on collagen I was due to differential adhesion, since our experiments allowed 2 hr for attachment. We also counted cell numbers on the samples recovered for scintillation counting and found it to be comparable (Figure 2C, right panel), thus ruling out the possibility that the difference was due to differential recovery. Taken together these results indicate that LCTC are responsive to normal inhibition of proliferation mediated by collagen I in a manner similar to that of normal SF.
Effects of PDGF AA and TGF-β on proliferation
To investigate the effects of PDGF AA we starved cells for 18 hr and stimulated them with increasing concentrations of PDGF AA. As shown in Figure 3A, LCTC were more sensitive to lower concentrations of the growth factor (5, 10 and 20 ng/ml), reaching a peak at 10 ng/ml. Although reaching the same level of stimulation at 40 ng/ml, SF were less sensitive to PDGF AA at lower concentrations. PDGF AA did not affect timing of cell cycle progression, since cells in the presence and in the absence of PDGF AA displayed a peak of [H3]-thymidine incorporation at 26 hr after release in medium containing serum and factors (results not shown). In conclusion, LCTC appear to be more sensitive to lower concentrations of PDGF AA.
Figure 3 Mitogenic effects of PDGF AA and TGF-β on SF and LCTC. A. Effect of increasing PDGF AA concentrations on growth of LCTC (black bars) and SF (gray bars) (n = 4). B. Effect of increasing concentrations of TGF-β on growth of LCTC (black bars) and SF (gray bars) (n = 4). (*) denotes significant difference between SF and LCTC; p ≤ 0.05. C. Kinetics of DNA synthesis in LCTC as measured by [H3]-thymidine incorporation in the presence (filled circles) and absence (empty circles) of TGF-β (n = 4). Note second delayed peak of incorporation at 40 hr (black arrow). D. Kinetics of DNA synthesis in SF as measured by [H3]-thymidine incorporation in the presence (filled circles) and absence (empty circles) of TGF-β (n = 4). Note the absence of the second delayed peak of incorporation at 40 hr (black arrow).
To test the responsiveness of LCTC and SF to TGF-β, cells were starved overnight and stimulated with increasing concentrations of acid-activated TGF-β. In Figure 3B we note that at lower concentrations, stimulation of LCTC and SF is comparable. However, higher concentrations of TGF-β (0.1 ng/ml) seemed to be less effective in promoting mitogenesis of SF. Interestingly, at the highest concentration (1 ng/ml) TGF-β was inhibitory to SF but still highly stimulatory for LCTC. When we analyzed the kinetics of thymidine incorporation during TGF-β treatment (Figure 3C,3D) we observed a second delayed peak of DNA synthesis at 40 hr (Figure 3C and 3D, arrow). In smooth muscle cells this second delayed peak has been shown to be due to a PDGF AA autocrine loop [28]. While in LCTC the second peak was 85% of the first peak, in SF it was slightly under 50%. These results indicate that TGF-β can induce a prolonged and strong induction of DNA synthesis in LCTC even at concentrations that are inhibitory for SF. In this set of experiments we were able to compare matched LCTC and SF from the same patient. Our results indicate that the observed differences between LCTC and SF are not due to individual variation or genetic background.
Synergy of extracellular matrix and growth factors
Next, we assessed the interplay between growth factors and the extracellular matrix on the stimulation of proliferation. We compared proliferation of cells cultured on collagen film using the highest inhibitory concentration (42 μg/cm2; Figure 2) and cultured on plastic. Initially, we treated cells with 10 ng/ml of PDGF AA in the presence and absence of collagen after 24 hr (Figure 4A). Both LCTC and SF were efficiently inhibited by collagen, in accordance with the experiments in Figure 2A and 2B, performed with different cell lines. In the absence of collagen, PDGF AA stimulated proliferation comparable to that obtained in Figure 3A and 3B. These conditions provide an internal control to distinguish effects that are cell type-specific from individual variations. Surprisingly, PDGF AA was able to override the collagen-mediated inhibition only in LCTC and not in SF.
Figure 4 Interaction between peptide growth factors and extracellular matrix. A. Effects of PDGF AA on the growth of LCTC (black bars) and SF (gray bars) plated on collagen or plastic (n = 4). B. Effects of TGF-β on the growth of LCTC (black bars) and SF (gray bars) plated on collagen or plastic (n = 3). (*) denotes significant difference between SF and LCTC; p ≤ 0.05.
Next, cells were stimulated with TGF-β (Figure 4B) and proliferation measured. Since there was a possibility that the autocrine loop could induce a late response, we assessed [H3]-thymidine incorporation at 40 hr. (Figure 4B). Collagen inhibited these cell lines as in previous experiments with LCTC and SF cell lines. TGF-β alone had little effect on proliferation. However, similarly to the experiment with PDGF AA, TGF-β was not only able to override the collagen-mediated inhibition but showed a synergistic effect only in LCTC. SF were not able to escape inhibition mediated by collagen.
Discussion
Injuries to liver tissue usually involve transient or long-term development of fibrosis. In acute injuries, fibrotic tissue is frequently reabsorbed and normal tissue architecture is restored. In cases when the primary agent or the secondary pathogenic mechanisms are persistent, the fibrotic reaction can be perpetuated causing a severe impairment of organ function [1]. It is therefore important to identify the factors that are involved in this perpetuation in order to devise more efficient preventive measures and therapies. However, there is a dearth of knowledge about the behavior of the human LCTC isolated from fibrotic livers. In fact, most studies done to date used LCTC cells from experimental models such as rats and mice and the few studies dealing with myofibroblasts from human liver derived the cells from normal tissue.
In order to evaluate the hypothesis that LCTC from fibrotic livers had lost control of proliferation we investigated their growth in tissue culture. Contact inhibition of proliferation is a characteristic of normal cells and is lost in neoplastic cell lines. Our results showed that LCTC respond normally to contact inhibition of proliferation. Collagen type I, which is the predominant extracellular matrix protein deposited during hepatic fibrogenesis [2], has been shown to be a potent inhibitor of mesenchymal cell proliferation [26,29,30]. Since the perpetuated proliferation of LCTC is a hallmark of chronic liver diseases, it was conceivable that these cells had lost normal collagen-mediated inhibition. However, our results indicate that LCTC derived from fibrotic livers are responsive to collagen I to an extent comparable to normal skin fibroblasts. In addition, collagen IV a major component of basement membranes did not affect the growth of either cell type studied.
While recent evidence from rat vascular smooth muscle cells indicates that PDGF AA promotes only protein synthesis without activation of DNA synthesis [31], we demonstrate that PDGF AA is active as a mitogen for human LCTC. These results are in accordance with previous data using connective tissue cells derived from normal liver [32]. Interestingly, the same authors have found that in fibrotic livers there is marked increase in the expression of both PDGFα receptor and its ligand PDGF AA [33]. Taken together, these results strengthen the notion that PDGF-AA is an important mediator of connective tissue expansion during liver fibrosis.
Although generally agreed to contribute to the fibrogenic process by upregulating expression of genes encoding extracellular matrix proteins, a role for TGF-β in promoting mitogenesis has emerged in the last few years. TGF-β displayed a bi-phasic curve with low concentrations stimulating and higher doses inhibiting the proliferation of SF. On the other hand, TGF-β was a potent mitogenic stimulus for LCTC even at high concentrations. Interestingly, it has been shown that TGF-β also stimulates an additional delayed growth response mediated by an autocrine loop of PDGF AA [28], which we observed only in LCTC.
One surprising outcome of this study was that not only the peptide growth factors tested were able to override the collagen-mediated inhibition, but that they were also able to synergize with the extracellular matrix. Since the collagen we used was a crude preparation from rat tail tendon we cannot rule out the possibility that the observed effect may depend on other associated proteins that are known to be accessory factors in promoting more efficient ligand-binding to the receptor. These results reinforce the notion that connective tissue cells can be therapeutically controlled using a strategy to inhibit the action of peptide growth factors.
It is well known that continuous presence of the injury agent determines the evolution of the disease, favoring its progression and perpetuation (establishment of a permanent scar tissue). On the other hand, it is largely unclear which are the other physiologic determinants for disease evolution. Genetic background is known to influence the development of hypertrophic cheloid skin scars as well as the outcome of schistosomal liver fibrosis [34]. In experiments in which we were able to use matched LCTC and SF from the same individual our results suggest that tissue origin was consistently more important than the genetic background, since LCTC behaved differently from the same-patient's SF but similarly to LCTC from different individuals.
Conclusions
The results obtained in this study using human primary liver connective tissue cells suggest a plausible scenario for the development of liver fibrosis. Stellate quiescent cells are triggered to proliferate by high molecular weight serum factors [15] and may be subjected to a positive autocrine and paracrine feedback loop of growth factor stimulation. Whereas under normal conditions connective tissue cells are kept under strict check by the extracellular matrix, PDGF AA and TGF-β are capable of not only overriding the inhibitory effect caused by collagen type I on LCTC but also synergize to provide a stronger stimulus for proliferation. This interplay of extracellular signals may underlie the development of irreversible liver fibrosis.
List of abbreviations
HEPES: N-[2-Hydroxyethyl] piperazine-N'-[2-ethanesulfonic acid]; HSC: hepatic stellate cells; LCTC: liver connective tissue cells; SF: skin fibroblasts; PDGF: platelet-derived growth factor; TGF-β: transforming growth factor-β.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AG established primary cell line cultures, carried out the cell biological studies, participated in designing and interpreting the experiments. MC provided technical support for the cell cultures and participated in the writing of the manuscript. RB participated in the design of the study and in the interpretation of the results. AM established primary cultures, conceived of the study, and participated in its design and coordination. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Fundação de Amparo à Pesquisa do Rio de Janeiro (FAPERJ) and Financiadora de Estudos e Projetos (FINEP), Brazil.
==== Refs
Friedman SL Molecular regulation of hepatic fibrosis, an integrated cellular response to tissue injury J Biol Chem 2000 275 2247 2250 10644669 10.1074/jbc.275.4.2247
Gressner AM The cell biology of liver fibrogenesis Cell Tissue Res 1998 292 447 452 9582401 10.1007/s004410051073
Rudolph R McClure WJ Woodward M Contractile fibroblasts in chronic alcoholic cirrhosis Gastroenterology 1979 76 704 709 421998
Grimaud JA Borojevic R Chronic human schistosomiasis mansoni. Pathology of the Disse's space Lab Invest 1977 36 268 273 839738
Grimaud JA Borojevic R Myofibroblasts in hepatic schistosomal fibrosis Experientia 1977 33 890 892 891763
Wells RG Kruglov E Dranoff JA Autocrine release of TGF-beta by portal fibroblasts regulates cell growth FEBS Lett 2004 559 107 110 14960316 10.1016/S0014-5793(04)00037-7
Knittel T Kobold D Piscaglia F Saile B Neubauer K Mehde M Timpl R Ramadori G Localization of liver myofibroblasts and hepatic stellate cells in normal and diseased rat livers: distinct roles of (myo-)fibroblast subpopulations in hepatic tissue repair Histochem Cell Biol 1999 112 387 401 10603079 10.1007/s004180050421
Grimaud JA Borojevic R Portal fibrosis: intrahepatic portal vein pathology in chronic human schistosomiasis mansoni J Submicrosc Cytol 1986 18 783 793 3097335
Li D Friedman SL Liver fibrogenesis and the role of hepatic stellate cells: new insights and prospects for therapy J Gastroenterol Hepatol 1999 14 618 633 10440206 10.1046/j.1440-1746.1999.01928.x
Tang L Tanaka Y Marumo F Sato C Phenotypic change in portal fibroblasts in biliary fibrosis Liver 1994 14 76 82 8196513
Voss B Rauterberg J Pott G Brehmer U Allam S Lehmann R von Bassewitz DB Nonparenchymal cells cultivated from explants of fibrotic liver resemble endothelial and smooth muscle cells from blood vessel walls Hepatology 1982 2 19 28 7033099
Borojevic R Vinhas SA Monteiro AN Domont GB Zyngier FR Grimaud JA Liver connective tissue cells isolated from human schistosomal fibrosis or alcoholic cirrhosis represent a modified phenotype of smooth muscle cells Biol Cell 1985 53 231 238 3160417
Monteiro AN Borojevic R Interaction of human liver connective tissue cells, skin fibroblasts and smooth muscle cells with collagen gels Hepatology 1987 7 665 671 3301615
Monteiro AN Borojevic R In vitro formation of fibrous septa by liver connective tissue cells In Vitro Cell Dev Biol 1987 23 10 14 3100500
Monteiro AN Borojevic R Complement-dependent induction of DNA synthesis and cell proliferation in human liver connective tissue cells in vitro In Vitro Cell Dev Biol Anim 1995 31 149 155 7735566
Border WA Okuda S Languino LR Sporn MB Ruoslahti E Suppression of experimental glomerulonephritis by antiserum against transforming growth factor beta 1 Nature 1990 346 371 374 2374609 10.1038/346371a0
Czaja MJ Weiner FR Flanders KC Giambrone MA Wind R Biempica L Zern MA In vitro and in vivo association of transforming growth factor-beta 1 with hepatic fibrosis J Cell Biol 1989 108 2477 2482 2500447 10.1083/jcb.108.6.2477
Sanderson N Factor V Nagy P Kopp J Kondaiah P Wakefield L Roberts AB Sporn MB Thorgeirsson SS Hepatic expression of mature transforming growth factor beta 1 in transgenic mice results in multiple tissue lesions Proc Natl Acad Sci U S A 1995 92 2572 2576 7708687
Bissell DM Roulot D George J Transforming growth factor beta and the liver Hepatology 2001 34 859 867 11679955 10.1053/jhep.2001.28457
George J Roulot D Koteliansky VE Bissell DM In vivo inhibition of rat stellate cell activation by soluble transforming growth factor beta type II receptor: a potential new therapy for hepatic fibrosis Proc Natl Acad Sci U S A 1999 96 12719 12724 10535989 10.1073/pnas.96.22.12719
Ueno H Sakamoto T Nakamura T Qi Z Astuchi N Takeshita A Shimizu K Ohashi H A soluble transforming growth factor beta receptor expressed in muscle prevents liver fibrogenesis and dysfunction in rats Hum Gene Ther 2000 11 33 42 10646637 10.1089/10430340050016139
Dooley S Hamzavi J Breitkopf K Wiercinska E Said HM Lorenzen J Ten Dijke P Gressner AM Smad7 prevents activation of hepatic stellate cells and liver fibrosis in rats Gastroenterology 2003 125 178 191 12851882 10.1016/S0016-5085(03)00666-8
Betsholtz C Raines EW Platelet-derived growth factor: a key regulator of connective tissue cells in embryogenesis and pathogenesis Kidney Int 1997 51 1361 1369 9150445
Bostrom H Willetts K Pekny M Leveen P Lindahl P Hedstrand H Pekna M Hellstrom M Gebre-Medhin S Schalling M Nilsson M Kurland S Tornell J Heath JK Betsholtz C PDGF-A signaling is a critical event in lung alveolar myofibroblast development and alveogenesis Cell 1996 85 863 873 8681381 10.1016/S0092-8674(00)81270-2
Davis BH Transforming growth factor beta responsiveness is modulated by the extracellular collagen matrix during hepatic ito cell culture J Cell Physiol 1988 136 547 553 3170650
Rhudy RW McPherson JM Influence of the extracellular matrix on the proliferative response of human skin fibroblasts to serum and purified platelet-derived growth factor J Cell Physiol 1988 137 185 191 3170657
Lin YC Grinnell F Decreased level of PDGF-stimulated receptor autophosphorylation by fibroblasts in mechanically relaxed collagen matrices J Cell Biol 1993 122 663 672 8335692 10.1083/jcb.122.3.663
Battegay EJ Raines EW Seifert RA Bowen-Pope DF Ross R TGF-beta induces bimodal proliferation of connective tissue cells via complex control of an autocrine PDGF loop Cell 1990 63 515 524 2171777 10.1016/0092-8674(90)90448-N
Yoshizato K Taira T Shioya N Collagen-dependent growth suppression and changes in the shape of human dermal fibroblasts Ann Plast Surg 1984 13 9 14 6331269
Schor SL Cell proliferation and migration on collagen substrata in vitro J Cell Sci 1980 41 159 175 7364880
Inui H Kitami Y Tani M Kondo T Inagami T Differences in signal transduction between platelet-derived growth factor (PDGF) alpha and beta receptors in vascular smooth muscle cells. PDGF-BB is a potent mitogen, but PDGF-AA promotes only protein synthesis without activation of DNA synthesis [published erratum appears in J Biol Chem 1995 Apr 28;270(17):10358] J Biol Chem 1994 269 30546 30552 7982973
Pinzani M Gentilini A Caligiuri A De Franco R Pellegrini G Milani S Marra F Gentilini P Transforming growth factor-beta 1 regulates platelet-derived growth factor receptor beta subunit in human liver fat-storing cells Hepatology 1995 21 232 239 7806159 10.1016/0270-9139(95)90433-6
Pinzani M Milani S Herbst H DeFranco R Grappone C Gentilini A Caligiuri A Pellegrini G Ngo DV Romanelli RG Gentilini P Expression of platelet-derived growth factor and its receptors in normal human liver and during active hepatic fibrogenesis Am J Pathol 1996 148 785 800 8774134
Dessein AJ Hillaire D elWali NE Marquet S Mohamed-Ali Q Mirghani A Henri S Abdelhameed AA Saeed OK Magzoub MM Abel L Severe hepatic fibrosis in Schistosoma mansoni infection is controlled by a major locus that is closely linked to the interferon-gamma receptor gene Am J Hum Genet 1999 65 709 721 10441577 10.1086/302526
| 15579200 | PMC539266 | CC BY | 2021-01-04 16:29:54 | no | BMC Gastroenterol. 2004 Dec 3; 4:30 | utf-8 | BMC Gastroenterol | 2,004 | 10.1186/1471-230X-4-30 | oa_comm |
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-5-481558829610.1186/1471-2474-5-48Research ArticleCerebrospinal fluid levels of opioid peptides in fibromyalgia and chronic low back pain Baraniuk James N [email protected] Gail [email protected] Jill [email protected] Daniel J [email protected] Center for the Advancement of Clinical Research, The University of Michigan, Ann Arbor, MI USA2 Chronic Pain and Fatigue Research Center, Division of Rheumatology, Immunology and Allergy, Room B107, Lower Level Kober-Cogan Building, Georgetown University, 3800 Reservoir Road, N.W. Washington, D.C. 20007-2197, USA2004 9 12 2004 5 48 48 10 3 2004 9 12 2004 Copyright © 2004 Baraniuk 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 mechanism(s) of nociceptive dysfunction and potential roles of opioid neurotransmitters are unresolved in the chronic pain syndromes of fibromyalgia and chronic low back pain.
Methods
History and physical examinations, tender point examinations, and questionnaires were used to identify 14 fibromyalgia, 10 chronic low back pain and 6 normal control subjects. Lumbar punctures were performed. Met-enkephalin-Arg6-Phe7 (MEAP) and nociceptin immunoreactive materials were measured in the cerebrospinal fluid by radioimmunoassays.
Results
Fibromyalgia (117.6 pg/ml; 85.9 to 149.4; mean, 95% C.I.; p = 0.009) and low back pain (92.3 pg/ml; 56.9 to 127.7; p = 0.049) groups had significantly higher MEAP than the normal control group (35.7 pg/ml; 15.0 to 56.5). MEAP was inversely correlated to systemic pain thresholds. Nociceptin was not different between groups. Systemic Complaints questionnaire responses were significantly ranked as fibromyalgia > back pain > normal. SF-36 domains demonstrated severe disability for the low back pain group, intermediate results in fibromyalgia, and high function in the normal group.
Conclusions
Fibromyalgia was distinguished by higher cerebrospinal fluid MEAP, systemic complaints, and manual tender points; intermediate SF-36 scores; and lower pain thresholds compared to the low back pain and normal groups. MEAP and systemic pain thresholds were inversely correlated in low back pain subjects. Central nervous system opioid dysfunction may contribute to pain in fibromyalgia.
==== Body
Background
Fibromyalgia (FM) is an enigmatic condition characterized by increased complaints of widespread pain with tenderness to palpation [1]. The tenderness is traditional tested by manually pressing over so-called tender points, but more recent studies have shown that the tenderness is generalized phenomenon [2]. The mechanisms responsible for the increase in the perception of pain in FM, and the variation of pain sensitivity in the general population are unclear. A similar continuum is seen with heat – induced pain. However, when subjects who report pain to a minimal stimulus (low pain threshold) were compared to subjects reporting less pain with the same stimulus (high pain threshold), there was enhanced functional magnetic resonance imaging (fMRI) responses in the low pain threshold group [3]. The differences in activation were greatest in the primary somatosensory cortex, anterior cingulate cortex and prefrontal cortex. These fMRI patterns suggest there may be a continuum within the population for pain thresholds, central cortical activation and verbalized pain perception. These results may be applicable to FM since the same brain regions have been identified in response to painful stimuli [4].
The pain present in FM may induce antinociceptive neural mechanisms with the release of opioid peptides. This hypothesis was tested by measuring opioid peptides in cerebrospinal fluid, and comparing these levels to systemic pain thresholds, subjective complaints, and quality of life measures in 3 sets of volunteers. FM and Chronic Fatigue Syndrome often overlap [5]. Chronic Fatigue Syndrome is characterized by severe fatigue associated with exertional exhaustion, pain symptoms, neurocognitive and sleep dysfunction [6]. Therefore, opioid levels were compared for FM and FM/ Chronic Fatigue Syndrome subsets. The second group had chronic low back pain (LBP) without FM or Chronic Fatigue Syndrome. These subjects have a chronic regional pain syndrome [7] and served as a positive control group. The negative control group was formed by healthy persons with no pain or fatigue.
Two peptides were selected for measurements because they were involved in antinociceptive responses. Alternatively, dysfunction of their release could predispose to chronic pain. Preproenkephalin A is the precursor for leucine-enkephalin, methionine-enkephalin (Met-enk), Met-enk-Arg6-Gly7-Leu8, and Met-enk-Arg6-Phe7 (MEAP) [8]. MEAP was elevated in many brain regions in inflammatory models of arthritis and gluteal carregeenan injection in rats [9,10]. Nociceptin, also known as orphanin FQ, was increased in the cingulate gyrus in rat chronic pain models [9].
Methods
Subjects
FM, LBP and Normal control subjects between the ages of 18 and 70 years were recruited to this IRB-approved protocol from rheumatology and orthopedics clinics, advertisements, and word of mouth. Normal subjects were pain and chronic fatigue free, had no diabetic, neurologic, inflammatory, autoimmune, or other chronic disorder that could predispose to pain, alterations in sensation, or known variations in cerebrospinal fluid composition. LBP inclusion criteria were (i) dominant pain complaint of low back pain, and (ii) imaging studies within the past 6 months. Exclusion criteria were: (a) evidence of a lumbar fracture or tumor to explain the pain, (b) any chronic illness that may affect functional status such as diabetes, cancer, chronic obstructive pulmonary disease, chronic inflammatory diseases, renal insufficiency or similar debilitating disorders, (c) previous back or neck surgery, and (d) FM or Chronic Fatigue Syndrome. FM subjects had a prior clinical diagnosis of FM including widespread pain affecting all 4 quadrants and the axial skeleton lasting at least 3 months that was not explained by any other chronic illness, and the presence of at least 11 of 18 tender points when manual, digital pressure of ~4 kg was applied [1]. All subjects were medication-free for at least 4 days prior to study. Subjects participated in a 1/2 day protocol that involved confirmatory history and physical examination, questionnaires, tender point examinations, and lumbar puncture.
Questionnaires
Systemic complaints questionnaire
This self – report questionnaire containing 44 queries grouped into the following modules: (i) Fatigue; (ii) Musculoskeletal: morning stiffness, muscle pain, muscle spasms, dry eyes, dry mouth, fingers sensitive to the cold, fingers turn blue and/or white in the cold, swollen lymph nodes, swollen joints, fever; (iii) Chest: shortness of breath (SOB), SOB when hurrying on level ground or walking up a slight hill, SOB when walking with other people of own age on level ground, stop for breath when walking at own pace on level ground, SOB when washing or dressing, rapid heart rate, chest pain, irregular heart rate, palpitations; (iv) Headaches: migraine or tension type; (v) Neurological: numbness or tingling of hands or legs, inability to concentrate, problems with memory, dizziness; (v) Ear, Nose and Throat (ENT): problems with balance, hearing loss, ear pain, sensation of ear blockage or fullness, ringing in the ears, sinus pain; (vi) Bladder: urinary urgency, pelvic discomfort / pain / pressure, persistent bladder fullness after urination, dysuria; and (vii) Irritable Bowel Syndrome (Rome I criteria): abdominal pain relieved with bowel movement, abdominal pain with a change in frequency or consistency of stool, changes in stool consistency, changes in stool form (hard or loose/watery), changes in passing of stool, bloating or feeling of abdominal distention, passage of mucus, nausea or vomiting [11]. Subjects were asked to respond "Yes" if they had recurrent or chronic symptoms for more than 3 of the past 12 months. The sum of positive responses for each module and the total were determined.
Subjects completed the MOS SF-36[12]. The domains were Physical Functioning (PF), Social Functioning (SF), Role Limitation due to Physical Problems (PP), Role Limitation due to Emotional Problems (EP), Mental Health (MH), Energy / Fatigue (E/F), Pain (P), General Perception of Health (GP) and Change in Health (CH).
Pain threshold and tender point examinations
All subjects had pressure testing at 9 bilateral sites (18 total) using a hand held dolorimeter (algometer) with a 1 cm2 rubber stopper making contact with the skin (Chatillon, etc.) [1,2]. The degree of pressure required to cause pain (pain threshold) was recorded at each site, and the number with pain induced by < 4 kg / cm2 recorded. The average pressure causing pain was the Average Pain Threshold. FM and Normal subjects had manual digital pressure examinations of these points [1,2]. The number of points that were painful was recorded as Manual Tender Points.
Cerebrospinal fluid (CSF) radioimmunoassays
Lumbar punctures were performed using local anesthetic and 23G spinal catheters. Volumes of 4 to 8 ml were obtained, centrifuged, aliquoted, and immediately frozen at -80°C. Samples were shipped on dry ice to Dr. Lars Terenius for measurement of neuropeptides. Neuropeptides were extracted from 1 ml aliquots using C-18 SepPak cartridges, eluted, dried, and resuspended for validated radioimmunoassays for MEAP [13,14] and nociceptin [15] using the standard methodologies developed in their laboratory. Concentrations in samples were interpolated from parallel standard curves. There was insufficient CSF to use HPLC for precise peptide identification. Hence, immunoreactive materials (irm) were measured as MEAP-irm and nociceptin-irm.
Statistics
Geometric mean and 95% confidence intervals were determined for each neuropeptide, with arithmetic means and 95% CI's for all other variables. Differences between groups were assessed by ANOVA. Differences between means for each pair of groups were assessed by 2-tailed unpaired Student's t-tests with Bonferroni corrections for multiple comparisons. Significance was ascribed for p < 0.05.
Results
Demographics
Lumbar punctures were performed on 14 FM (1 male), 10 LBP (5 male) and 6 Normal (2 male) subjects. The averages and ranges of ages for these 3 groups were similar (overall average 42.7 yr, 38.8 to 46.6; 95% CI). There were 4 African-Americans in the FM group, and 1 each in the LBP and Normal groups, and 1 Asian in the FM group. The remainder was Caucasian.
Systemic complaints questionnaire
Significant differences were found between FM and Normal results by 2-tailed unpaired Student's t-tests with p < 0.05 after Bonferroni corrections for multiple comparisons of this data. All of the FM subjects complained of fatigue (figure 1). Other highly prevalent individual symptoms in FM were morning stiffness, muscle pain and spasms, and difficulties concentrating on cognitive tasks. FM scores were higher than Normal for ENT and IBS (p < 0.05), Chest, Headache, Neurological and Bladder (p < 0.01), Fatigue and Musculoskeletal (p < 0.001) complaints. When compared to LBP, FM scores were higher for Neurological (p < 0.05), Chest (p < 0.01) and Fatigue (p < 0.001) complaints. Normal and LBP scores were not different. Chronic Fatigue Syndrome co-existed with FM in 7 of the 14 subjects compared to none in the LBP and Normal groups. IBS was present in 58% of FM, 20% of LBP and 17% of Normal subjects.
SF-36
Domain Scores for the Normal group were all near the predicted values of 100 (figure 2). FM scores were significantly lower for Emotional Problems (p < 0.05), Social Functioning (p < 0.01), Physical Functioning, Role Limitation due to Physical Problems, Energy / Fatigue, Pain and General Perception of Health (p < 0.001), but not Mental Health or Change in Health. All Normal means were highly significantly greater (p < 0.001) than all LBP scores. FM scores were higher than LBP for Emotional Problems, General Perception of Health, and Change in Health (p < 0.05), Social Functioning (p < 0.01), and Mental Health and Pain (p < 0.001). FM and LBP were not different for Physical Functioning, Role Limitation due to Physical Problems or Energy / Fatigue.
Pain thresholds and tender point counts
Dolorimetry identified significantly lower pressure pain thresholds for FM (1.51 kg/cm2) compared to the LBP (2.48 kg/cm2; p < 0.01) and Normal (2.60 kg/cm2; p < 0.001) groups (figure 3). The numbers of tender points determined by dolorimetry were 9.07 for FM, 3.67 for LBP, and 1.33 (p < 0.05 vs. FM) for Normal subjects (figure 4). Digital pressure identified more manual tender points in FM (13.00; 12.27 to 14.73) than Normal (4.67; 1.00 to 8.34; p < 0.001) groups. An average of 3.8 (1.90 to 5.70) more tender points were detected by manual examinations than by dolorimetry (p < 0.01). This difference has been attributed to higher anxiety and other psychometric variables in FM [16]. (Manual tender point counts were not recorded for LBP subjects.)
Cerebrospinal fluid neuropeptide concentrations
MEAP – irm
The 3 groups had significantly different mean CSF concentrations (p = 0.0014, ANOVA). The normal volunteers had significantly lower geometric mean concentrations (26.3 pg/ml; 13.9 to 49.9) than the FM (101.7 pg/ml; 72.8 to 142.0; p < 0.01 vs. normal), and LBP (78.0 pg/ml; 51.1 to 119.0; p < 0.05 vs. normal) (figure 5). Co-morbid Chronic Fatigue Syndrome did not affect the MEAP – irm results, since the arithmetic means were 112 pg/ml for FM subjects with this syndrome (n = 7) and 124 pg/ml with FM alone (n = 6). There were no obvious relationships with age, gender or race. The small sample size precluded further statistical analysis of these variables.
Nociceptin – irm was not different between FM (4.27 pg/ml, 3.22 to 5.66, n = 14), LBP (4.52 pg/ml, 3.12 to 6.55, n = 10) and Normal (5.65 pg/ml, 2.65 to 12.04, n = 6) groups.
Systemic pain thresholds and MEAP – irm
These 2 variables were correlated (Pearson's correlation coefficient of -0.38, p < 0.05; explained variance 0.15) when all subjects were examined as a single group (figure 6). This correlation was not found when the Normal and FM groups were examined by themselves. Normal subjects had higher thresholds and lower MEAP – irm concentrations. FM subjects had pain thresholds below 2.3 kg/cm2, which was coincidentally the lower 95% CI for the Normal group. MEAP – irm concentrations had a wide range in the FM subjects, but the geometric mean was significantly higher than for Normal subjects. Pain threshold and MEAP – irm concentrations did not have linear correlations in either the FM or Normal group. These data suggested that FM subjects were fundamentally different from Normal. When the pain threshold was below 2 to 2.3 kg/cm2, MEAP – irm levels increased approximately 4-fold compared to Normal. The combination of the widespread pain, systemic complaints, low pain threshold and high MEAP – irm concentrations in CSF was distinct from the low level of symptomatic complaints, normal (high) pain thresholds, and lower MEAP – irm levels found in the Normal group.
The LBP group had a chronic regional pain syndrome, no fibromyalgia, normal pain thresholds and systemic complaints, but severe disability (SF-36 scores). Only the LBP group showed a linear correlation between pain threshold and MEAP – irm concentrations (figure 6). The parameters of this correlation were similar to that of the entire group. This was due to the overlap of some high pain threshold / low MEAP – irm LBP subjects with the Normal group, and low pain threshold / high MEAP – irm LBP subjects with the FM data. This continuum of pain threshold and MEAP – irm levels in LBP was different from the clustered FM and Normal datasets, and suggested a different mechanism of MEAP – irm regulation in LBP from FM.
Discussion
The Normal group had Systemic Complaints and SF-36 scores in the normal ranges, high pain thresholds, low numbers of manual and dolorimetry-derived tender points, and low CSF concentrations of MEAP – irm and nociceptin – irm.
The LBP group was a positive control for chronic regional pain. Their Systemic Complaints scores, systemic pain thresholds and dolorimetry defined tender point counts were not significantly different from Normal. However, most of their SF-36 results were near zero indicating the worst level of impairment of the 3 groups. They were the only group to show a correlation of decreasing pain thresholds with increasing MEAP – irm concentrations. The continuum of MEAP – irm levels in the LBP group led to borderline significance for the comparison to Normal levels. Inclusion of LBP subjects with higher or lower pain thresholds may have shifted the MEAP – irm concentration distribution towards or away, respectively, from the Normal group results. This is important when comparing these data to those of other studies. For example, chronic sciatica patients did not have elevated MEAP – irm compared to controls [33]. However, severity was not graded as extensively as in our study. Some subjects may have had less severe low back pain than in our group. If so, then the linear correlation noted in FIGURE 6 would predict no significant difference from normal subjects. Conversely, female LBP subjects with the lowest pain thresholds and highest MEAP – irm levels may have been making a transition from chronic low back pain to fibromyalgia [7]. Half of our LBP group was male, introducing gender as a potentially confounding factor.
The FM group's results were distinctly different from the Normal and LBP groups. FM had widespread pain complaints, the highest Systemic Complaints scores, the lowest pain thresholds, and highest numbers of tender points of the 3 groups. Their SF-36 scores were intermediate between LBP and Normal groups. Widespread pain, low pain thresholds, and high Systemic Complaints scores differentiated FM from Normal and LBP. CSF MEAP – irm concentrations were approximately 3-fold higher in FM than Normal (figure 5). This confirmed earlier findings [14] where a group of women meeting an older set of fibromyalgia criteria [17,18] had 34% higher CSF MEAP – irm concentrations than a group of 8 age-matched female control subjects [13]. None of the FM subjects in the earlier group required analgesics or other medications suggesting that their symptoms may have been milder than for our FM group. Our group contained 1 male and 13 females. In contrast, Lui et al. found MEAP concentrations (peptide identity confirmed by HPLC) that were 38% lower in FM than control subjects (p < 0.01) [19]. There was inadequate clinical data to compare the severity of complaints between these FM populations. These investigators also used a liquid-liquid peptide extraction method. The differences in FM severity, control groups, extraction procedures, and lack of sufficient CSF to identify precise peptides by HPLC [20] made it difficult to compare these sets of divergent results. Standardized measurements on CSF withdrawn from highly characterized and clearly defined subjects and controls will be required to resolve these inconsistencies.
This is the first investigation to examine the potential effect of co-existing CFS on MEAP – irm levels in FM. This suggested that the mechanism(s) of CFS were probably independent of those responsible for the elevated MEAP – irm levels in FM. Unfortunately, CFS subjects without FM could not be simultaneously tested to determine if their MEAP – irm levels were normal (as would be predicted).
Nociceptin – irm levels were the same in our three groups. The levels were about 10% of that found in women during labor [21]. It was unclear if the higher concentrations were due to pregnancy, neurohormonal adaptations during labor and delivery, or the effects of acute pain. Again, the absence of a control group comparable to our pain-free Normal group makes mechanistic comparisons difficult.
Conclusions
The Normal group had Systemic Complaints and SF-36 scores in the normal ranges, high pain thresholds, low numbers of manual and dolorimetry-derived tender points and low cerebrospinal fluid MEAP and nociceptin concentrations. The LBP chronic regional pain group had similar Systemic Complaints scores, systemic pain thresholds and dolorimetry-defined tender point counts. However, most of their SF-36 results were near zero indicating the worst level of impairment of the 3 groups. MEAP – irm was just significantly elevated compared to the Normal group, and was correlated to the systemic pain threshold. The FM group was distinct since they had widespread pain complaints, the highest Systemic Complaints scores, the lowest pain thresholds, and highest numbers of tender points of the 3 groups. Their SF-36 scores were intermediate between the LBP and Normal groups. MEAP – irm concentrations were significantly higher in the FM than Normal group. The co-existence of Chronic Fatigue Syndrome with FM did not alter the MEAP – irm concentrations. This suggested that Chronic Fatigue Syndrome mechanism(s) did not involve preproenkephalin dysfunction. Nociceptin – irm levels were not different between these groups, and were lower than previously reported results from pregnant women in labor. Significant differences in MEAP – irm concentrations from previous studies may be due to the highly controlled definition of patients in this study, selection of control groups, and differences in peptide extraction methods.
Abbreviations
CSF, cerebrospinal fluid; FM, fibromyalgia; fMRI, functional magnetic resonance imaging; irm, immunoreactive material; LBP, low back pain; MEAP, Met-enk-Arg6-Phe7; SF-36, short-form of 36 questions with the following domains: PF, physical functioning; SF social functioning; PP, role limitation due to physical problems; EP, role limitation due to emotional problems; MH, mental health; E/F, energy / fatigue; P, pain; GP, general perception of health; CH, change in health; Systemic Complaints domains: MS, musculoskeletal; HA, headache; Neuro, neurological; ENT, ear, nose & throat; IBS, irritable bowel syndrome (Rome I criteria); SOB, shortness of breath.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JNB compiled the results and wrote the manuscript. DJC oversaw the clinical investigations that were performed by JC. GW was responsible for the sample repository, shipping, and collection of data. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Special thanks to Dr. Lars Terenius, Karolinska Institute, for providing the antibodies and performing the radioimmunoassays for this study. The work was supported by Dept of Army Grant # DAMD17-00-2-0018 (GW, JC & DC), U.S. Public Health Service Award RO1 AI42403 (JNB), and 1 M01-RR13297-01A1 from the General Clinical Research Center Program of the National Center for Research Resources, National Institute of Health.
Figures and Tables
Figure 1 Systemic symptoms questionnaire results. FM (white bars), LBP (grey bars) and Normal (black bars) results are shown for each domain. The x-axis shows the systems domains with the total number of questions in parentheses. The y-axis shows the number of positive responses within each domain. The error bars are the 95% confidence intervals. FM scores were higher than Normal for ENT and IBS (p ≤ 0.05), Chest, Headache, Neurological and Bladder (p ≤ 0.01), Fatigue and Musculoskeletal (p ≤ 0.001). FM scores were higher than LBP for Neurological (p ≤ 0.05), Chest (p ≤ 0.01) and Fatigue (p ≤ 0.001). Normal and LBP scores were not different.
Figure 2 SF-36 domain results. Scores were ranked Normal (white bars with down-going 95%CI) > FM (black bars with up-going 95%CI) > LBP (grey bars with up-going 95%CI). The Normal group had significantly higher scores than FM for EP (p ≤ 0.05), SF (p ≤ 0.01), PF, PP, E/F, P and GP, but not MH or CH. All Normal means were highly significantly greater (p ≤ 0.001) than all LBP scores. FM scores were higher than LBP for EP, GP, and CH (p ≤ 0.05), SF (p ≤ 0.01), and MH and P (p ≤ 0.001). FM and LBP were not different for PF, PP or E/F.
Figure 3 Pressure pain thresholds (kg/cm2) determined by dolorimetry. Pain thresholds for FM (triangles) were significantly lower than LBP (squares; p = 0.009) and Normal (circles; p = 0.002) groups. The bars indicated means and 95% confidence intervals.
Figure 4 Tender points. The numbers of tender points detected by dolorimetry with ≤ 4 kg/cm2 pressure were higher for FM (triangles, n = 14), than LBP (squares, n = 9, p = 0.03) and Normal (circles, n = 6, p = 0.008) groups. The bars indicated means and 95%CI's.
Figure 5 Met-Enkephalin-Arg6-Phe7 (MEAP) concentrations (pg/ml) in cerebrospinal fluid (CSF) from normal (circles), low back pain (LBP, diamonds) and fibromyalgia (FM, triangles) subjects. The bars indicate geometric means and 95% confidence intervals. The groups were significantly different by ANOVA (p = 0.0014). MEAP in the FM (p < 0.01) and LBP (p < 0.05) groups were significantly higher than Normal (2-tailed unpaired Student's t-tests).
Figure 6 Semi-logarithmic relationships between pain thresholds and MEAP. FM subjects (black triangles) had pain thresholds below 2.3 kg/cm2 (heavy, black error bars). Normal subjects (grey circles, cabled grey error bars) had lower MEAP and higher pain thresholds compared to FM. LBP subjects (open diamonds) had results that covered the full ranges for both variables (upper right error bars). The data for all subjects were linearly correlated (Pearson's correlation coefficient R = -0.38, p < 0.05) with an explained variance (R2) of 0.15. The line appears curved on this semi-logarithmic plot.
==== Refs
Wolfe F Smythe HA Yunus MB Bennett RM Bombardier C Goldenberg DL Tugwell P Campbell SM Abeles M Clark P The American College of Rheumatology 1990 Criteria for the Classification of Fibromyalgia. Report of the Multicenter Criteria Committee Arthritis Rheum 1990 33 60 72
Wolfe F The relation between tender points and fibromyalgia symptom variables: evidence that fibromyalgia is not a discrete disorder in the clinic Ann Rheum Dis 1997 56 268 271 9166001
Coghill RC McHaffie JC Yen Y-F Neural correlates of interindividual differences in the subjective experience of pain Proc Natl Acad Sci USA 2003 100 8538 8542 12824463 10.1073/pnas.1430684100
Gracely RH Petzke F Wolf JM Clauw DJ Functional magnetic resonance imaging evidence of augmented pain processing in fibromyalgia Arthritis Rheum 2002 46 1333 1343 12115241 10.1002/art.10225
Bradley LA McKendree-Smith NL Alarcon GS Pain complaints in patients with fibromyalgia versus chronic fatigue syndrome Curr Rev Pain 2000 4 148 157 10998728
Fukuda K Straus SE Hickei I Sharpe MC Dobbins JC Komaroff A The chronic fatigue syndrome: a comprehensive approach to its definition and study Ann Intern Med 1994 121 953 959 7978722
Lapossy E Maleitzke R Hrycaj P Mennet W Muller W The frequency of transition of chronic low back pain to fibromyalgia Scand J Rheumatol 1995 24 29 33 7863275
Yoshimasa T Nakao K Sakamoto M Suda M Morii N Ikeda Y Ishihara T Manno M Hamada S Shimbo S Demonstration and characterization of immunoreactive methionine-enkephalin, leucine-enkephalin, methionine-enkephalin-Arg6-Gly7-Leu8 and methionine-enkephalin-Arg6-Phe7 in human phaeochromocytoma Acta Endocrinol (Copenh) 1984 107 261 267 6495993
Rosen A Lundeberg T Bytner B Nylander I Central changes in nociceptin, dynorphin B and Met-enkephalin-Arg-Phe in different models of nociception Brain Res 2000 857 212 218 10700570 10.1016/S0006-8993(99)02432-4
Spetea M Rydelius G Nylander I Ahmed M Bileviciute-Ljungar I Lundeberg T Svensson S Kreicbergs A Alteration in endogenous opioid systems due to chronic inflammatory pain conditions Eur J Pharmacol 2002 435 245 252 11821033 10.1016/S0014-2999(01)01554-0
Baraniuk JN Clauw JD Gaumond E Rhinitis symptoms in chronic fatigue syndrome Annals Allergy Asthma Immunol 1998 81 359 365
Ware JE Sherbourne CD The MOS 36-item short-form health survey (SF-36): I. Conceptual framework and item selection Med Care 1995 30 473 483
Lyrenaes S Nyberg F Folkesson R Lindberg B Terenius L Met-enkephalin-Arg6-Phe7 in human CSF. Increased levels in late pregnancy Progress in Opioid Research NIDA Res Monogr 1986 75 430 433
Vaeroy H Nyberg F Terenius L No evidence for endorphin deficiency in fibromyalgia following investigation of cerebrospinal fluid (CSF) dynorphin A and Met-enkephalin-Arg6-Phe7 Pain 1991 46 139 143 1684241 10.1016/0304-3959(91)90068-9
Terenius L Sandin J Sakurada T Nociceptin/orphanin FQ metabolism and bioactive metabolites Peptides 2000 21 919 922 10998525 10.1016/S0196-9781(00)00228-X
Giesecke T Williams DA Harris RE Cupps TR Tian X Tian TX Gracely RH Clauw DJ Subgrouping of fibromyalgia patients on the basis of pressure-pain thresholds and psychological factors Arthritis Rheum 2003 48 2916 2922 14558098 10.1002/art.11272
Yunus MB Masi AT Calabro JJ Miller KA Feigenbaum SL Primary fibromyalgia (fibrositis): a clinical study of 50 patients with matched normal controls Semin Arthritis Rheum 1981 11 151 171 6944796 10.1016/0049-0172(81)90096-2
Smythe HA Hollander E, McCarty DJ Non-articular rheumatism and the fibrositis syndromes In Arthritis and Allied Conditions 1972 11 8 Lea and Febviger, Philadelphia, PA 808 816
Liu Z Welin M Bragee B Nyberg F A high-recovery extraction procedure for quantitative analysis of substance P and opioid peptides in human cerebrospinal fluid Peptides 2000 21 853 860 10959008 10.1016/S0196-9781(00)00219-9
Lui DX Wood GW Desiderio DM Proenkephalin A and proopiomelanocortin peptides in human cerebrospinal fluid J Chromatogr 1990 530 235 252 1964161
Brooks H Elton CD Smart D Rowbotham DJ McKnight AT Lambert DG Identification of nociceptin in human cerebrospinal fluid: comparison of levels in pain and non-pain states Pain 1998 78 71 73 9822213 10.1016/S0304-3959(98)00130-4
| 15588296 | PMC539267 | CC BY | 2021-01-04 16:03:42 | no | BMC Musculoskelet Disord. 2004 Dec 9; 5:48 | utf-8 | BMC Musculoskelet Disord | 2,004 | 10.1186/1471-2474-5-48 | oa_comm |
==== Front
BMC EcolBMC Ecology1472-6785BioMed Central London 1472-6785-4-161553742610.1186/1472-6785-4-16Research ArticleLocomotor activity in common spiny mice (Acomys cahirinuse): The effect of light and environmental complexity Eilam David [email protected] Department of Zoology, Tel-Aviv University, Ramat-Aviv 69 978, Israel2004 10 11 2004 4 16 16 18 8 2004 10 11 2004 Copyright © 2004 Eilam; licensee BioMed Central Ltd.2004Eilam; 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
Rodents typically avoid illuminated and open areas, favoring dark or sheltered environments for activity. While previous studies focused on the effect of these environmental attributes on the level of activity, the present study tested whether the spatio-temporal structure of activity was also modified in illuminated compared with dark and complex compared with open arenas. For this, we tested common spiny mice (Acomys cahirinus) in empty or stone-containing arenas with lights on or lights off.
Results
In an illuminated or open arena, spiny mice moved in less frequent but longer trips with relatively long distances between consecutive stops. In contrast, in either a dark arena or an arena with stones, the animals took shorter and more frequent trips, with more stops per trip and shorter inter-stop distances. In illuminated arenas spiny mice remained mainly along the walls, whereas locomotion in the center was more prevalent in dark empty arenas, and was carried out along convoluted paths. Increasing environmental complexity by adding stones to either illuminated or dark arenas increased locomotion along straight trajectories and away from walls.
Conclusions
Earlier findings of reduced activity in illuminated or open areas have been extended in the present study by demonstrating changes in the spatio-temporal structure of locomotor behavior. In the more complex arenas (with stones) spiny mice traveled along short straight segments whereas in the open their trips were longer and took the shape of a zigzag path which is more effective against fast or nearby predators. Alternatively, the zigzag path may reflect a difficulty in navigation.
==== Body
Background
Rodents typically avoid illuminated and open areas, favoring dark or sheltered environments for activity. Indeed, higher activity was described in numerous field and laboratory studies of nocturnal species tested in the dark, compared with their activity when tested under light. For example, common spiny mice (Acomys cahirinus) decreased activity and foraging in open spaces under moonlit compared with dark nights [1]. When tested in the dark, laboratory rats increase their activity and display behaviors that indicate reduced habituation, fear and anxiety [2]. Deermice (Peromyscus maniculatus) were shown to reduce activity in the open during moonlit nights and were active only near objects such as rocks, grasses, and walls, where they could successfully evade a predator attack [3]. Thus, it appears that rodents perceive increased risk of predation in open spaces and/or during moonlit nights and in consequence shift their activity to more protected microhabitats [4-7]. While many of these studies used indirect measures of locomotor activity, such as footprints [8], the present laboratory study was aimed at direct observation of locomotor behavior under various light levels and arena complexities.
The 'open field' is a widely used apparatus in laboratory studies of rodents' locomotor activity [9]. This apparatus has been criticized for being "a poor and explicitly aversive environment with excess light and open spaces..."[10]. Nevertheless, it is a relatively simple testing environment for a variety of species, in which they display a typical behavioral structure [11] that withstands drastic environmental changes [12]. Studies in wild and laboratory rodents in an illuminated open field (e.g., [9,11-14]) have shown that their locomotor behavior is organized in reference to a key location – the home-base. At the home base, the rodent demonstrates typical behaviors (e.g. grooming and crouching), and sets out on round trips in the area. The building block of the round trip is a stop, with an upper limit of 8–10 stops per trip [15]. The limited number of stops/trip is preserved by scaling the distance between successive stops and adjusting the number of trips, even under large changes in arena size [12,16]. Accordingly, rodents in a larger area made fewer yet longer trips, whereas in a small area they made shorter but more frequent trips. Following these earlier studies, the present study tested open field behavior under varying light level and arena complexity.
The common spiny mouse (Acomys cahirinus) was selected for this study since it is a strictly nocturnal species that displaces other species to crepescular or diurnal activity [17,18]. Common spiny mice were thus expected to be sensitive to tests in illuminated compared with dark environments. Also, they live in rocky environments, nimbly foraging in crevices between and under rocks and boulders [18,19], where the complex habitat structure provides shelter and escape from predators. They were thus expected to be also sensitive to changes in environmental complexity. Three questions were posed in this study of common spiny mice: i) is their behavior in an illuminated arena similar to that seen in mice, rats, and voles? ii) does behavior change in dark arena and/or with increased environmental complexity? iii) what is the functional mechanism that may underlie behavioral changes in dark or complex environments? As shown below, activity increased in dark or in complex environment, and took a different form of short straight trajectories. In contrast, spiny mice traveled through the center in a convoluted path in either lit or dark empty arenas. This later form of progression may have a defensive advantage.
Results
Level of locomotor activity
In the illuminated arena, the distance traveled by spiny mice was significantly affected by the density of stones. In contrast, neither the number of stones nor arena size alone significantly increased activity. As shown for small arena in Fig. 1, traveled distance was significantly greater when four stones were present than when stones were absent. Traveled distance was not greater in comparing small with large empty arenas, or small with large 4 stone arenas. Therefore, changing arena size alone did not increase activity. However, stone density of four/m2 significantely increased activity, as shown for small arena with four stones or large arena with 16 stones (Fig 1). This trend of increased activity with increased stone-density was also echoed in traveling speed (Table 1).
Figure 1 Distance traveled (mean ± SEM) in the lit arenas (open bars) and dark arenas (dark bars). Arena size and number of stones in each arena are depicted along the x-axis. Significant comparisons, as revealed in Tukey HSD test, are depicted by lines at top left. As shown, traveled distance did not change with only arena size. Adding four stones to a small arena significantely increased traveled distance (compare small arena with 0 and with 4 stones); however, adding 4 stones to a large arena did not have a significant effect. In the dark arena, the number of stones did not have a significant effect on the traveld distance.
Table 1 Parameters of locomotion in an illuminated small arena (1 × 1 m) and large arena (2 × 2 m). For most variables, values increased when stones were added or when arena size was increased. Bonferroni adjustment of p-value was calculated by P = 0.05/10 = 0.005. Mean (± SEM) are followed by superscript numerals that indicate the significantely different test groups (as appeared in the top row) in Tukey Honest Significant Difference (HSD) test.
Small arena Large arena
Empty (1) 4 Stones (2) Empty (3) 4 Stones (4) 16 Stones (5) F45;p
Level of activity
Traveled distance (m.) 52.0 ± 9.7 2,4,5 86.7 ± 8.81 60.9 ± 9.55 90.4 ± 9.71 115.6 ± 7.31,3 8.68; <0.0001
Speed (m/sec) 0.10 ± 0.022,3,5 0.15 ± 0.041,5 0.12 ± 0.024,5 0.18 ± 0.021,3,5 0.36 ± 0.121–4 5.68; = 0.0012
Temporal Structure
Stops 87.0 ± 31.42,4,5 228.6 ± 55.71,3,4 51.2 ± 13.92,4,5 146.9 ± 23.9 1–5 235.8 ± 48.6 1,3,4 7.15; = 0.0002
# of trips 18.5 ± 5.52,4,5 41.4 ± 11.01,3 12.2 ± 5.02,4,5 38.5 ± 7.81,2,5 74.7 ± 25.21,3,4 5.81; = 0.001
Stops/trip 5.0 ± 0.6 5.6 ± 0.4 5.5 ± 1.2 4.4 ± 0.6 3.6 ± 0.3 0.95; ns
Trip length 4.9 ± 1.43,5 2.4 ± 0.43 10.2 ± 2.31,2,4,5 3.1 ± 0.63,5 2.0 ± 0.31,3,4 5.88; = 0.0009
Inter-stop distance (m.) 0.94 ± 0.22 0.43 ± 0.06 2.68 ± .1.14 0.70 ± 0.09 0.54 ± 0.06 2.51;ns
Spatial Distribution
Center stops (%) 9.3 ± 1.52–5 22.6 ± 4.91,5 55.5 ± 30.81 20.3 ± 1.91,5 41.7 ± 2.81,2,4 3.08; = 0.0277
Center time (%) 2.4 ± 1.12,4,5 19.2 ± 6.11,3 2.4 ± 1.12,4,5 12.8 ± 3.91,3,5 25.5 ± 7.11,3,4 8.99; <0.0001
Meander (deg/m) -1.87 ± 0.312 -0.99 ± 0.362 -1.52 ± 0.194,5 -0.60 ± 0.163,5 -0.46 ± 0.093,4 7.68; <0.0002
In a dark arena, the level of activity resembled the highest level that was measured in the illuminated arena, and did not vary significantly with the number of stones (Fig. 1). Thus, activity of spiny mice in a dark arena was steady and high, regardless of the number of stones or their density (Table 2).
Table 2 Parameters of locomotion in a dark large arena (2 × 2 m). As shown, level of activity was not affected, whereas the spatiotemporal structure underwent significant changes. Bonferroni adjustment of p-value was calculated by P = 0.05/10 = 0.005. Mean (± SEM) are followed by the results of Tukey HSD test, indicating the numbers of the significantely different test groups (as appeared in the top row).
Empty (1) 4 Stones (2) 16 Stones (3) F18;p
Level of activity
Traveled distance (m.) 116.51 ± 7.52 110.17 ± 11.58 99.09 ± 10.38 0.91; ns
Speed (m/sec) 0.21 ± 0.02 0.21 ± 0.02 0.20 ± 0.01 0.24; ns
Temporal Structure
Stops 159.1 ± 18.23 185.6 ± 9.43 322.7 ± 32.01,2 18.67; .0004
# of trips 40.14 ± 5.532,3 62.57 ± 3.261,3 134.14 ± 13.971,2 36.68; 0.000001
Stops/trip 4.04 ± 0.232,3 3.01 ± 0.231,3 2.41 ± 0.041,2 21.20; 0.00019
Trip length (m.) 3.08 ± 0.262,3 1.81 ± 0.251,3 0.74 ± 0.041,2 36.41;0.00000
Inter-stop distance (m.) 0.77 ± 0.072,3 0.59 ± 0.051,3 0.31 ± 0.011,2 26.75; <0.00001
Spatial Distribution
Center stops (%) 25.8 ± 1.32,3 49.8 ± 4.81,2 70.0 ± 1.01,2 66.09; 0.00000
Center time (%) 29.3 ± 4.22,3 53.4 ± 6.01 57.7 ± 3.61 11.58; 0.00058
Meander -0.44 ± 0.031,2 -0.33 ± 0.021,3 -0.29 ± 0.051,2 7.10; 0.0053
Temporal organization of locomotor activity
In illuminated arenas, increases in traveled distance were echoed in the number of stops, and it was not possible to distinguish whether the increase in stops was directly linked to increased traveled distance, or whether it was due to the increased number of stones. In the dark arena, however, traveled distance was not different in the three groups (Table 2; Fig. 1), but number of stops increased with number of stones, indicating that stops depended on the number of stones and not on the traveled distance.
The number of trips to the home base increased with the number of stops, which increased with the number of stones. However, the mean number of stops in a trip did not vary in the various groups tested in the illuminated arena (Table 1). As shown in Table 1, in the absence of stones, trip length significantly increased with arena size and spiny mice took fewer but longer trips in the large arena compared with more but shorter trips in the small arena. In addition, inter-stop distance was significantly higher in the large compared with the small illuminated arena. Consequently, the traveled distance was similar in both small and large arenas with same number of stones (Table 1 and Fig. 2).
Figure 2 Scaling of interstop distance according to arena size. In the small arena (left illustration) the spiny mouse takes two round trips that start and end at the home base (top left corner), stopping only in the corners of the arena (4 stops/round trip, including the stop at the home base). In the large arena, the spiny mouse takes one trip, stopping only at the corners of the arena (again, 4 stops). Thus, trip length and interstop distance are shorter in the small arena, and the number of trips and overall number of stops are smaller in the large arena. The shorter but more frequent trips in the small arena and longer but fewer trips in a large arena result in the same overall traveled distance and the same number of stops per trip.
When the number of trips increased with increase in number of stones, trip-length and inter-stop distance decreased, reflecting the tendency of spiny mice to stop at or near stones. Changes in the number of stops/trip were non-significant (Table 1). Overall, these changes imply that with increase in number of stones, spiny mice set out from the home base to more trips in the arena, but these trips were shorter in distance, had a shorter distance between successive stops, but preserved a relatively invariant number of stops per trip.
A similar trend was evident in the dark arenas, where with increase in number of stones, spiny mice took more trips that were shorter in length and in inter-stop distance. However, the non-significant decrease in the number of stops per trip that was noted in illuminated arenas with increased number of stones, reached statistical significance in the dark. Indeed, the number of stops/trip significantly decreased in 4-stone and in 16-stone arenas (Table 2). Overall, while the level of activity underwent conspicuous changes in illuminated arenas and remained steady in dark arenas, the temporal structure of locomotor behavior underwent similar changes in both illuminated and dark arenas.
Spatial distribution of locomotor activity and path shape
In empty illuminated arenas, spiny mice spent more than 80% of the time in the corners, the rest of the time mostly along the walls, and as little as 3% of the time in the center. Adding stones changed this pattern and the animals spent 13–26% of the time in the center, as well as stopping more frequently in the center (Table 1). In the dark, however, spiny mice spent 30–60% of the time and 30–70% of their stops in the center, with both percentage of time and stops increasing with increase in number of stones (Table 1).
In both small and large empty arenas, either dark or illuminated, spiny mice moved through the center in a convoluted path, changing frequently the direction of progression. When stones were added, trajectories comprised of more straight segments and fewer changes in direction of progression (Fig. 3). This change was reflected in the Meander index, which describes the angular change in direction of progression relative to distance moved. As shown, the meander was high without stones, and significantly decreased when stones were added (Tables 2 &3). Changes in the level of activity and its spatio-temporal structure are summarized in Table 3.
Figure 3 Trajectories of locomotion of exemplary spiny mice in lit arenas (top) and dark arenas (bottom). Each square shows one spiny mouse. As shown, in both small and large empty arenas, either dark or lit, spiny mice moved through the center in a convoluted path, changing frequently the direction of progression. Locomotion in the center increased in the dark or with the number of stones. With stones, trajectories of locomotion comprised more straight segments and fewer changes in direction of progression.
Table 3 Formal summary of the results shown in Tables 1 and 2.
Dark vs. light With lights on With lights off
Level of activity Distance traveled and speed Longer distances Increase with stone density Remains high
Temporal structure # of stops and trips More stops and trips More stones = more stops and trips
Trip length Shorter trips More stones = sorter trips
Stops/trip Fewer stops/trip Did not change Decreased
Spatial distribution Path shape Winding (zigzag) paths More stones = straighter path
Time and stops at the perimeter More time and stops in center More stones = more time and stops in center
Discussion
Spiny mice in the wild inhabit rocky mountains, dwelling in the crevices between and under rocks and boulders. It was therefore assumed that adding stones to an arena would create a complex environment, more resembling their natural habitat. Indeed, in an empty illuminated arena, spiny mice spent extended periods in the corners, traveling mainly along the walls, rarely entering the center of the empty arena where they traveled in a winding path. When stones were placed in the illuminated arena, the animals traveled significantly longer distances, as expected. While it was the density of stones rather than their number that accounted for the increased activity in the illuminated arena, introducing stones into a dark arena did not affect the level of activity, and the distance traveled was high regardless of number of stones or their density. In the following discussion it is proposed that increased activity is due to a sense of security and/or easier navigation provided by the stones whereas the convulted path in empty arena is a defensive strategy or a refelction of a dificulty of navigation in environment without landmarks or shelter (=stones).
Numerous field and laboratory studies found increased activity in nocturnal prey species tested in the dark, compared with their activity when tested under light [2,20]. In the same vein, foraging in rodents was shown to be closely associated with complex areas (shrubs) on bright nights but evenly distributed between sheltered and open areas on dark nights [8,21,22]. This avoidance of open areas probably reflects the finding that rodents are attacked and captured more frequently in the open [23]. It should be noted, however, that this anti-predatory pattern is effective against aerial raptors, but not necessarily against terrestrial predators, as indicated by the increased activity of snakes during dark nights [24], a time when rodents have higher activity. Therefore, the present findings that spiny mice avoid open illuminated spaces while demonstrating a higher level of locomotion in a more complex environment and/or in the dark areas, reinforces previous results on the effect of light level and habitat structure.
The present study demonstrates a change in path shape when locomoting in the center: spiny mice traveled along convoluted trajectories and rarely took straight paths (Fig. 3). These frequent changes in the direction of progression decreased with the increase in number of stones, and were especially conspicuous in empty dark arena, when activity in the center was prevalent. This behavioral pattern is reminiscent of the finding that gerbils' foraging path [25]. A mathematical model [26] suggested that a zigzag trajectory is advantageous when encountering a close or fast predator, whereas a straight trajectory is advantageous in facing a distant and relatively slow predator. Spiny mice may therefore move in a zigzag pattern as a defence against aerial raptors (fast predator) or snakes (close predator). Indeed, when spiny mice were attacked by a barn owl, they continued to locomote fast while frequently changing direction of progression, forming a convoluted path [27].
Another explanation for the changes in path shape is that stones are landmarks, and without them, especially in the dark, spiny mice may have difficulty in navigating [28] and therefore move in a convoluted path. Once landmarks (stones) are available, mice can more easily navigate and travel in straight paths, whereas when stones are absent they travel in the relatively homogenous environment along a winding path. A reminiscent mechanism was described in desert ants (Cataglyphis fortis) that return to nest directly but not necessarily in a straight path, presumably turning as frequently to the right as they do to the left, to reduce overall directional bias [29]. A survey of the mechanisms that may underlie intermittent progression suggests that pauses increase the capacity of sensory systems to detect relevant stimuli, and may involve perceptual processes such as velocity blur, relative motion detection, foveation, attention and interference between sensory systems [30]. When stones (=landmarks) are present, stops are frequent and spatial information can be collected during stops, alowing traveling along straight trajectories, whereas the lack of such spatial information processing may result in a winding path, as seen in empty and/or dark arena.
Light condition affected the spatial distribution of locomotor activity: while spiny mice remained most of the time along the walls of empty illuminated arenas, they increased the center time by 5–10 folds in complex environments. That the animals spent more time close to the walls in the empty illuminated arena compared with dark arena or complex arenas is unsurprising, probably linked to thigmotaxis, as shown in other rodent species (e.g., [20,31-33]). In the dark arena, however, center time and stops in the center were distinctly higher than in the illuminated arena, comprising 25%–70% of activity. This further supports the assumption that spiny mice move more in the center when afforded shelter by darkness and/or by the physical structure of the environment.
Stopping may also have an anti-predatory role [34] since owls usually attack moving prey, after being stimulated by its movement [35-37]. In consequence, a common defensive strategy in prey species is to freeze and remain immobile in the face of life threat, in order to eliminate the auditory and visual cues that predators use in pinpointing prey [38]. In following the above discussion on a possible defensive significance of convoluted paths, it is possible that complex environment in the dark does not provide the same sense of security than it does in an illuminated arena. This could be a result of snake activity, which is higher in dark and complex habitats but lower in the open [22,24,31]. It should be noted, however, that the above explanations are not mutually exclusive, and stopping may have a synergistic role in orientation, physiological recovery, and anti-predatory defense [34].
The present results in spiny mice are thus consistent with previous similar results in rats and voles [15,16]. in that they indicate that the animals preserve activity level and temporal structure under changing arena size. This observation may be a general property of rodents' open field behavior, which is gained by scaling interstop distance and number of trips to the home base. When environmental complexity was increased by adding stones, the number of trips increased while their length decreased. Therefore, the higher level of locomotor activity in complex environments was the result of more frequent but shorter trips and not of longer trips. This was obvious in the dark arena, where level of activity was high regardless of environment complexity, while the number of trips increased and their length decreased with increase in space complexity. These differences in the structure of trips may serve as a search-image parameter in other studies in spiny mice. For example, it is expected that foraging (e.g. traveling to food patches) will be longer in distance and less frequent in illuminated or exposed environments, but shorter and more frequent in a dark or sheltered area [39]. Long trips in the open are more risky, however, and spiny mice therefore need to undertake measures that reduce this risk. One possible way of reducing risk may be achieved by changing the distribution of activity and path shape, as described above.
Conclusions
The present results follow previous studies that demonstrated lower activity in illuminated and open areas compared with dark and complex areas. Observations on the behavior of spiny mice under these conditions revealed changes in the number and length of trips, in stopping frequency, and in path shape. Altogether, these changes reflect a flexible adaptation of locomotor activity to environmental conditions in a way that may be interpreted as aimed at efficient navigation, preserving the temporal structure of behavior, and reducion of predatory risk.
Methods
Study animals
The common spiny mouse (Acomys cahirinus) weighs 38–44 g and is 11 cm long, plus a 10-cm tail [40]. Spiny mice are an exceptional genus among murid rodents (Muridae) in being precocial and not having a nest. They differ from rats and mice in many aspects (see [41] for review); noteworthy are differences in depth perception [42], distance perception [43], exploration [44] and excitability [45]. We obtained 71 adult spiny mice bred in captive colonies at the research zoo of Tel-Aviv University. Fifty spiny mice were divided into five groups (n = 10; five males and five females per group); these groups were tested in a illuminated arena. The other 21 spiny mice were divided into three groups (n = 7; 3–4 males and 3–4 females per group); these groups were tested in a dark arena. The larger group size in light tests was due to the greater behavioral variability in light compared with dark tests. Several weeks before testing, the animals were housed in groups of 5–10, in metal cages measuring 40 cm × 70 cm and 25 cm, located outdoors in the zoo yard under natural (uncontrolled) temperature and light conditions. Overturned ceramic pots and wooden boxes were placed in each cage to provide shelter. Seeds, diced fresh vegetables, and live fly larvae were provided ad lib. Based on years of experience in maintaining colonies of spiny mice in our zoo, provision of water is unnecessary when sufficient fresh vegetables are provided.
Apparatus
A test arena was constructed by enclosing a tiled floor with plywood planks (50 cm high). Two arena sizes were used: 1 × 1 m and 2 × 2 m. Stones (tiles), 12 cm long, 12 cm wide and 6 cm high, were placed in the arena (details below). The arena was located inside an air-conditioned room (24°C), and could be illuminated by one of the following light-sources: (1) two 300 W light bulbs directed to the white ceiling in order to provide diffused illumination of the arena (Light tests); (2) two infrared lights (Tracksys, IR LED Illuminator; UK) that emit light in a range invisible to rodents (Dark tests; light level was 0.0425 Lux as measured with Profisix Sbc, Gossen). The video signal was recorded on a VCR (JVC HR-J737).
Procedure
Cages with spiny mice were brought to a room adjacent to the testing room 10 h before testing. For testing, a spiny mouse was removed in random order from the cage to a jar, and gently released from the jar into the center of the arena. Each spiny mouse was tested only once for 10 minutes, being randomly assigned to one of the above arenas. At the end of testing, animals were returned to the population at the research zoo. The first five groups of spiny mice (n = 10/group) were all tested in illuminated arenas under the following conditions: (1) small empty arena (no stones); (2) small arena with 4 stones; (3) large empty arena; (4) large arena with 4 stones; and (5) large arena with 16 stones (see Fig. 1). It should be noted that a setting of a small arena with 16 stones would have virtually prevent locomotion, and was therefore excluded. Three additional groups (n = 7/group) were tested in a large (2 × 2 m) dark arena with: i) no stones; ii) four stones; and iii) 16 stones, in order to compare the behavior in these three arenas with the behavior in the respective illuminated arenas.
Behavioral analysis
A tracking system (Ethovision by Noldus, NL) was used for data acquisition. The tracking system was set to score the spiny mouse as "not moving" (=stopping) when its center of gravity moved at a speed lower than 2 cm/sec, or as "moving" when the speed exceeded this limit during tracking at a rate of 25 frames/sec. Each arena was divided into four zone types. These were corners – a 20 × 20 cm square at each of the four corners of the arena; walls – a 20 cm strip along the walls between the corner zones; stones – a 20 × 20 cm square centered with each stone; center, the remaining area that comprised the spaces between the stones and away from the walls and the corners of the arena.
Based on our past studies, the parameters acquired from Ethovision were classified to represent three perspectives: i) level of activity, ii) temporal organization of locomotion, and iii) spatial distribution of locomotion. The parameters that were measured are described in Table 4. Briefly, the level of activity refers to the amount of activity regardless of temporal structure of spatial distribution. For example, the metric distance traveled was measured regardless of whether it comprised intermittent or continuous locomotion, or whether it was along the perimeter or in the center of the arena. The temporal structure refers to the order of bouts of locomotion and stopping periods. Parameters on temporal organization were derived in past studies, showing that locomotor behavior is organized in relation to a home base; a place where a rodent spends the longest cumulative non-locomoting periods [11,12,16]. From the home base the rodent takes round trips in the environment. The spatial distribution was aimed at distinguishing where activity had occurred. For example, the same amount of activity with the same temporal structure could be executed along the perimeter of the arena, or only in its center. Alternatively, a spiny mouse could travel in a straight or a winding trajectory. For these, the representation of the spatial distribution of activity was required.
Table 4 Parameters of locomotion that were measured for each spiny mouse.
Behavior Description
Level of activity
Distance traveled Overall distance (m.) that a spiny mouse traveled during the 10-min observation.
Traveling speed Overall traveled distance divided by the duration of locomoting periods (m/sec).
Temporal organization
Number of stops Incidence of "non-locomoting" intervals (stops), bounded by locomotion.
Number of trips Trips are intervals between consecutive stops at the home base, which is the place with the highest rank among zones according to the accumulated "non-locomoting" intervals. Thus, a trip comprised progression out from home base through consecutive stops in the arena, until returning to the home base.
Stops/trip Number of stops taken between two successive visits to the home base (= total number of stops divided by the total number of trips).
Trip length Metric distance traveled in a round-trip to the home base (total distance divided by the total number of trips).
Inter-stops distance The metric distance traveled between two consecutive stops (or, distance traveled in a "locomoting" interval = distance divided by number of stops).
Spatial distribution
Time spent along the perimeter (%) Calculated as percentage of the total time, in order to show how long spiny mice stayed at the vicinity of the walls of the arena, compared with the time spent in the center of the arena or near/on the stones.
Stops along the perimeter (%) Calculated as percentage of the total stops, in order to show how many stops took place along the vicinity of the walls of the arena, compared with stopping in the center of the arena or at/on the stones.
Meander The rate of change in direction of progression relative to the distance traveled, calculated automatically by Ethovision for each two successive time points by dividing the turn angle by the distance. Mean meander of each spiny mouse was used to calculate the mean of each group. + indicated a clockwise change in direction of progression, whereas - indicated a counterclockwise change. Thus, lower absolute (+ or -) meander values characterize locomotion along relatively straight trajectories, and higher meander absolute values describe circular or winding trajectories. It should be noted that meander is sensitive to tracking rate, animal size, and arena size. Therefore, meander may be compared only for the same animal size, same resolution, and same arena size.
Statistics
One way analysis of variance (ANOVA) was applied. Some of the data may not be strictly independent – i.e. trip length is the division of traveled distance by the number of trips, etc. – therefore, a Bonferroni correction was applied to set alpha level to 0.005 (0.05 divided by the 10 parameters that were used). Data calculated as proportions were transformed to the arcsine of the square-root-transformed raw data.
Acknowledgments
I am grateful to Sivan Reihan for her help in testing and data acquisition, to Naomi Paz for editing the manuscript, and to Barak Levy and the zookeepers of the I. Meier Segals Garden for Zoological Research for maintenance of the common spiny mice. This research was supported by The Israel Science Foundation, Grant 471/04.
==== Refs
Jones M Mandelik Y Dayan T Coexistence of temporally partitioned spiny mice: Roles of habitat structure and foraging behavior Ecology 2001 82 2164 2176
Nasello AG Machado C Bastos JF Felicio LF Sudden darkness induces a high activity low anxiety state in male and female rats Physiol Behav 1998 63 451 454 9469741 10.1016/S0031-9384(97)00462-9
Clarke JA Moonlight's influence on predator/prey interactions between short-eared owls (Asio flammeus) and deermice (Peromyscus maniculatus) Behav Ecol and Sociobiol 1983 13 205 209 10.1007/BF00299924
Price MV Waser NM Bass TA Effects of moonlight on microhabitat use by desert rodents J Mammal 1984 65 353 356
Bowers MA Seed removal experiments on desert rodents: the microhabitat by moonlight effect J Mammal 1988 69 201 204
Daly M Behrends PR Wilson MI Jacobs LF Behavioural modulation of predation risk: moonlight avoidance and crepuscular compensation in a nocturnal desert rodent Dipodomys merriami Anim Behav 1992 44 1 9
Vasquez RA Assessment of predation risk via illumination level – facultative central place foraging in the cricetid rodent Phyllotis darwini Behav Ecol and Sociobiol 1994 34 375 381 10.1007/s002650050053
Brillhart DB Kaufman DW Influence of illumination and surface-structure on space use by prairie deer mice (Peromyscus maniculatus bairdii) J Mammal 1991 72 764 768
Wilson RC Vacek T Lanier DL Dewsbury DA Open-field behavior in muroid rodents Behav Biol 1976 17 495 506 788698 10.1016/S0091-6773(76)90901-9
Genaro G Schmidek WR Exporation activity of rats in three different evironments Ethology 2000 106 849 859 10.1046/j.1439-0310.2000.00605.x
Eilam D Golani I Home base behavior of rats (Rattus norvegicus) exploring a novel environment Behav Brain Res 1989 34 199 211 2789700
Eilam D Open-field behavior withstands drastic changes in arena size Behav Brain Res 2003 142 53 62 12798265 10.1016/S0166-4328(02)00382-0
Drai D Kafkafi N Benjamini Y Elmer G Golani I Rats and mice share common ethologically relevant parameters of exploratory behavior Behav Brain Res 2001 125 133 140 11682104 10.1016/S0166-4328(01)00290-X
Whishaw IQ Tomie J Piloting and dead reckoning dissociated by fimbria-fornix lesions in a rat food carrying task Behav Brain Res 1997 89 87 97 9475617 10.1016/S0166-4328(97)00068-5
Golani I Benjamini Y Eilam D Stopping behavior: constraints on exploration in rats (Rattus norvegicus) Behav Brain Res 1993 53 21 33 8466665
Eilam D Dank M Maurer R Voles scale locomotion to the size of the open-field by adjusting the distance between stops: A possible link to path integration Behav Brain Res 2003 141 73 81 12672561 10.1016/S0166-4328(02)00322-4
Kronfeld-Schor N Dayan T Partioning of time as an ecological resource Ann Rev Ecol, Evol and Systematics 2003 34 153 181 10.1146/annurev.ecolsys.34.011802.132435
Shkolnik A Diurnal activity in a small desert rodents Int J Biometeorol 1971 15 115 120 5146797
Shargal E Kronfeld-Schor N Dayan T Population biology and spatial relationships of coexisting spiny mice (Acomys) in Israel J Mammal 2000 81 1046 1052
Topping MG Millar JS Goddard JA The effects of moonlight on nocturnal activity in bushy-tailed wood rats (Neotoma cinerea) Can J Zool 1999 77 480 485 10.1139/cjz-77-3-480
Kotler BP Brown JS Hasson O Factors affecting gerbil foraging behavior and rates of owl predation Ecol Dis 1991 72 2249 2260
Diaz M Rodent seed predation in cereal crop areas of central Spain – Effects of physiognomy, food availability, and predation risk Ecography 1992 15 77 85
Longland WS Price MV Direct observations of owls and heteromyid rodents – can predation risk explain microhabitat use Ecol Dis 1991 72 2261 2273
Clarke JA Chopko JT Mackessy SP The effect of moonlight on activity patterns of adult and juvenile prairie rattlesnakes (Crotalus viridis viridis) J Herepetol 1996 30 192 197
Dall SRX Kotler BP Bouskila A Attention, 'apprehenssion' and gerbil searching in patches Ann Zool Fen 2001 38 15 23
Furuichi N Dynamics between a predator and a prey switching two kinds of escape motions J Theor Biol 2002 217 159 166 12202110 10.1006/jtbi.2002.3027
Edut S Eilam D Rodent in open space adjust their behavioral response to the different risk levels during barn-owl attack BMC Ecology 2003 3 10 14614781 10.1186/1472-6785-3-10
Etienne AS Sitbon S Dahnhurni C Maurer R Golden hamsters on the 8-arm maze in light and darkness – the role of dead reckoning Quart J Exp Psychol Section B – Comp and Physiol Psych 1994 47 401 425
Muller M Wehner R Path integration in desert ants, Cataglyphis fortis Proc Natl Acad Sci USA 1988 85 5287 5290 16593958
Kramer DL McLaughlin RL The behavioral ecology of intermittent locomotion Am Zool 2001 41 137 153
Falkenberg JC Clarke JA Microhabitat use of deer mice: Effects of interspecific interaction risks J Mammal 1998 79 558 565
Sonnino S Spatial activity and habitat use of crested porcupine, Hystrix cristata L, 1758 (Rodentia, Hystricidae) in central Italy Mammalia 1998 62 175 189
Zadicario P Avni R Zadicario E Eilam D 'Looping' – an exploration mechanism in exploring a dark open field Behav Brain Res 2004
Vasquez RA Ebensperger LA Bozinovic F The influence of habitat on travel speed, intermittent locomotion, and vigilance in a diurnal rodent Behavioral Ecology 2002 13 182 187 10.1093/beheco/13.2.182
Metzgar LH An Experimental comparison of screech owl predation on resident and transient white-footed mice (Peromyscus leucopus) J Mammal 1967 48 387 391
Glickman SE Morrison BJ Some behavioral and neural correlates of predation susceptibility in mice Comm Behav Biol 1969 4 261 267
Kaufman DW Differential predation on active and inactive prey by barn owls Auk 1974 91 172 173
Mikkola H Willis I Owls of Europe 1983 England: T and A D Poyser Ltd
Mandelik Y Foraging microabitat use and foraging efficiencies in the common spiny mouse Acomys cahirinus Thesis submitted toward the MSc degree at Tel-Aviv university, Israel (in Hebrew, with abstract in English) 1999
Mendelssohn H Yom-Tov Y Fauna Palaestina: Mammalia of Israel 1999 Jerusalem: Keterpress Enterprises
Brunjes PC The precocial mouse, Acomys cahirinus Psychobiol 1990 18 339 350
Greenberg G Depth perception in Mongolian gerbils (Meriones unguiculatus) and spiny mice (Acomys russatus and A. cahirinus) J Comp Psychol 1986 100 81 84 3698584 10.1037//0735-7036.100.1.81
Goldman M Skolnick AJ Hernandez TP Tobach E Distance perception in the spiny mouse Acomys cahirinus – vertical jumping Percept Mot Skills 1992 75 883 895 1454491
Birke LI Sadler D Effects of odor familiarity on the development of systematic exploration in the spiny mouse, Acomys cahirinus Dev Psychobiol 1987 20 627 640 3691969
Marona-Lewicka D Michaluk J Antkiewicz-Michaluk L Vetulani J A comparison of locomotor responses to some psychotropic drugs and cerebral receptors in the Acomys cahirinus and the laboratory mouse Pol J Pharmacol Pharm 1987 39 293 302 2830608
| 15537426 | PMC539268 | CC BY | 2021-01-04 16:29:14 | no | BMC Ecol. 2004 Nov 10; 4:16 | utf-8 | BMC Ecol | 2,004 | 10.1186/1472-6785-4-16 | oa_comm |
==== Front
BMC Pulm MedBMC Pulmonary Medicine1471-2466BioMed Central London 1471-2466-4-131558142510.1186/1471-2466-4-13Research ArticleRandomised, controlled trial of N-acetylcysteine for treatment of acute exacerbations of chronic obstructive pulmonary disease [ISRCTN21676344] Black Peter N [email protected] Althea [email protected] Tracey E [email protected] Phillippa J [email protected] Robert P [email protected] Department of Medicine, University of Auckland, Private Bag 92019, Auckland, New Zealand2004 6 12 2004 4 13 13 10 5 2004 6 12 2004 Copyright © 2004 Black et al; licensee BioMed Central Ltd.2004Black 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
Prophylactic treatment with N-acetylcysteine (NAC) for 3 months or more is associated with a reduction in the frequency of exacerbations of chronic obstructive pulmonary disease (COPD). This raises the question of whether treatment with NAC during an acute exacerbation will hasten recovery from the exacerbation.
Methods
We have examined this in a randomised, double-blind, placebo controlled trial. Subjects, admitted to hospital with an acute exacerbation of COPD, were randomised within 24 h of admission to treatment with NAC 600 mg b.d. (n = 25) or matching placebo (n = 25). Treatment continued for 7 days or until discharge (whichever occurred first). To be eligible subjects had to be ≥ 50 years, have an FEV1 ≤ 60% predicted, FEV1/VC ≤ 70% and ≥ 10 pack year smoking history. Subjects with asthma, heart failure, pneumonia and other respiratory diseases were excluded. All subjects received concurrent treatment with prednisone 40 mg/day, nebulised salbutamol 5 mg q.i.d and where appropriate antibiotics. FEV1, VC, SaO2 and breathlessness were measured 2 hours after a dose of nebulised salbutamol, at the same time each day. Breathlessness was measured on a seven point Likert scale.
Results
At baseline FEV1 (% predicted) was 22% in the NAC group and 24% in the control group. There was no difference between the groups in the rate of change of FEV1, VC, SaO2 or breathlessness. Nor did the groups differ in the median length of stay in hospital (6 days for both groups).
Conclusions
Addition of NAC to treatment with corticosteroids and bronchodilators does not modify the outcome in acute exacerbations of COPD.
==== Body
Background
Exacerbations are an important cause of morbidity in Chronic Obstructive Pulmonary Disease. Seemungal et al found that exacerbations were an important determinant of quality of life in COPD [1]. In addition hospital admissions with exacerbations account for a large proportion of the expenditure on the treatment of COPD [2]. This has led to a search for strategies to prevention exacerbations and to hasten their resolution when they do occur. A systematic review found that treatment with mucolytics for 2 months or more reduced the frequency of exacerbations by 29% [3]. The majority of the studies included in the review were with N-acetylcysteine. These findings are supported by a recent pharmacoepidemiologic study [4]. Aside from its action as a mucolytic, N-acetylcysteine is an antioxidant [5,6] and has anti-inflammatory actions [7] and this could contribute to its actions in preventing exacerbations of COPD. Exacerbations of COPD are characterized by increased infiltration of the airways with neutrophils and eosinophils [8] and by increased production of reactive oxygen species [9]. Reactive oxygen species can activate the epidermal growth factor receptor to promote mucus secretion [10] and this is one of the potential mechanisms by which an increase in reactive oxygen species could lead to a worsening of an acute exacerbation. In view of this we wondered whether N-acetylcysteine might also be useful in the treatment of patients presenting with an acute exacerbation of COPD. We undertook a randomised, double-blind, placebo-controlled, parallel group study of oral N-acetylcysteine 600 mg b.d. in addition to standard treatment in patients admitted to hospital with an acute exacerbation of COPD.
Methods
Patients were eligible for inclusion in the study if they had a physician diagnosis of COPD, were ≥ 50 years of age, had a smoking history ≥ 10 pack years and had been admitted to hospital with an acute exacerbation of their COPD in the previous 24 hours. In addition they were required to have FEV1 ≤ 60% predicted and FEV1/VC ≤ 0.7 at time of inclusion. Patients were excluded if they had any of the following conditions: asthma (as the primary diagnosis), heart failure, bronchiectasis, bronchial carcinoma, interstitial lung disease, pneumonia. They were also excluded if they were unable to comply with the study procedures because they did not speak English or were demented or if they had any other medical problems that in the opinion of the investigator would interfere with the conduct of the study.
Subjects were treated with N-acetylcysteine 600 mg twice daily (b.d.) or matching placebo. Treatment was continued for 7 days or until discharge whichever occurred first. Effervescent N-acetylcysteine tablets were purchased from Zambon pharmaceuticals (Milan, Italy) and were repackaged in size 0 gelatine capsules. Each capsule contained 300 mg of N-acetylcysteine. Placebo capsules were prepared containing lactose as a filler. A randomisation schedule was drawn up by the hospital pharmacy and patients were allocated sequential randomisation numbers as they entered the study.
In addition to N-acetylcysteine the patients received standard treatment for their exacerbation as specified by the hospital guidelines. This was oxygen therapy, prednisone 40 mg o.d. for one week and nebulised bronchodilators i.e. salbutamol 5 mg four times daily (q.i.d.) and ipratropium 0.5 mg q.i.d. Antibiotics were prescribed if the patients had increased volume and/or purulence of sputum. Mucolytics were not permitted during the trial except as study medicine.
The primary endpoint was breathlessness measured on a seven point Likert scale (Table 1). Secondary endpoints were FEV1, oxygen saturation and length of hospital stay. The study assessments were performed at the same time each day and two hours after the last dose of nebulised bronchodilator. Breathlessness was assessed prior to spirometry. Spirometry was performed according to American Thoracic Society criteria using a Vitalograph spirometer. Oxygen saturation was measured using a Nelcor N-20 pulse oximeter. Supplemental oxygen was stopped for 10 minutes before any of the measurements were performed. In addition the subjects were interrogated on each occasion about any possible adverse effects.
Table 1 Likert Scale for Breathlessness
1 Extremely short of breath
2 Very short of breath
3 Quite a bit short of breath
4 Moderate shortness of breath
5 Some shortness of breath
6 A little shortness of breath
7 Not at all short of breath
The study was conducted at Auckland Hospital between June 2001 and October 2001. The study was approved by the Auckland Ethics Committee and all participants provided written informed consent.
Power calculation
The power calculation was based on a previous study where we treated subjects for acute exacerbations of COPD with theophylline for up to seven days [11]. In that study the average improvement in Likert score from baseline to the end of the study was 1.95 with theophylline and 1.05 with placebo. We assumed the subjects in this study would have a similar distribution of Likert scores for breathlessness. On this basis, 50 participants gave us 88% power to detect a similar difference between the groups as in the previous study, at a 5% level of significance.
Statistical analysis
The two treatment groups were compared at baseline using Student's t-test for normally distributed variables, Wilcoxon test for non-parametric data and Fisher's exact test for categorical data.
The effect of N-acetylcysteine on breathlessness, lung function and oxygen saturation was analysed by fitting the patient data to a random coefficient model using a mixed linear model approach (least squares regression). The baseline (Day 1) measurements were used as co-variates in the analysis. This allowed individual slopes and intercepts to be formed for each patient and their random variation incorporated in the model. This model adjusted for the varying number of observations available on the different patients. Non-normal dependent variables were rendered normal by transformation. Significant and main interaction effects were investigated by the method of Tukey.
All tests were two-tailed and a 5% significance level was maintained throughout these analyses. The analyses were carried out using SAS Version 8.0 (SAS Institute Inc, Cary, NC, USA.)
The life test procedure of SAS was used to compute nonparametric estimates of the length of stay function by the Kaplan-Meier method. Comparison between these functions was made using the Wilcoxon and log rank tests.
Results
Two hundred and ten patients who had been admitted to hospital with an exacerbation of COPD were screened for the study. Fifty subjects were randomised to treatment with 25 subjects receiving N-acetylcysteine and 25 subjects receiving placebo. The commonest reasons for exclusion were concomitant heart failure (n = 41) or pneumonia (n = 35).
The two groups were similar at baseline in terms of age, smoking history, lung function, oxygen saturation and breathlessness (Table 2). There were more men in the placebo group but the difference was not statistically significant (p = 0.23). There was no difference between the groups in the use of inhaled bronchodilators prior to admission (Table 3). More subjects in the N-acetylcysteine group had been on treatment with inhaled corticosteroids prior to admission but this difference was not statistically significant (p = 0.16). Although we did not document the amount of sputum produced by the subjects most subjects presented both with an increased volume of sputum as well as breathlessness.
Table 2 Baseline characteristics of the subjects
N-Acetylcysteine (n = 25) Placebo (n = 25)
Gender Male/Female 11/14 19/6
Age Years (SD) 73.6 (7.8) 73.0 (8.2)
Smoking History Pack Years (SD) 44.4 (36.2) 53.7 (36.8)
FEV1 % predicted (SD) 22 (10) 24 (12)
VC % predicted (SD) 56 (18) 64 (22)
SaO2 % (SD) 90.2 (4.0) 90.4 (2.7)
Breathlessness Likert Score (IQ range) 4 (3–6) 4 (3–5)
Values are shown as mean and standard deviation except for Likert scores that are shown as median and interquartile range. There were no statistically significant differences between the two groups for any of the measures.
Table 3 Concurrent medications
N-acetylcysteine (n = 25) Placebo (n = 25)
Inhaled steroids 15 9
Oral prednisone 9 6
Short acting inhaled beta-agonists 20 19
Ipratropium bromide 14 14
Long acting inhaled beta-agonists 7 7
Theophylline 4 1
The patients on treatment with oral prednisone included patients on long term treatment with oral steroids and those who were prescribed prednisone for this exacerbation prior to admission. There was no significant difference between the N-acetylcysteine and placebo groups for any of the concomitant medicines.
All of the subjects completed the study with none being withdrawn early. The rate of change in the Likert scores, lung function and oxygen saturation is shown in Table 4. For the Likert score, FEV1, and SaO2 the rate of change was greater with placebo than with N-acetylcysteine but none of these differences were statistically significant. Table 5 shows the absolute changes in Likert score, FEV1 and SaO2 from the beginning to end of the study.
Table 4 Slope of least squares regression line
N-acetylcysteine (n = 25) Placebo (n = 25)
Likert Score 0.16 (0.42) 0.35 (0.45)
FEV1 % predicted 0.001 (0.015) 0.019 (0.019)
Sa02 0.40 (0.89) 0.88 (1.43)
Mean and standard deviations for slopes of the least square regression lines for the effects of NAC and placebo on Likert scores for breathlessness, FEV1 % predicted and SaO2. There were no significant differences between NAC and placebo.
Table 5 Change in outcome measures from beginning to end of study
N-acetylcysteine (n = 25) Placebo (n = 25)
Likert Score 0.7 0.8
FEV1 (litres) 0.03 0.15
Sa02 (%) 1.2 1.8
The average change in Likert score, FEV1, VC and SaO2 from entry into the study to end of study (discharge or Day 7) are shown.
The Kaplan-Meier analysis showed a similar time course until discharge for the treatment and placebo arms (Figure 1). Neither the log-rank statistic (p = 0.33), which places more weight on longer lengths of stay in hospital, nor the Wilcoxon test (p = 0.30) which places more weight on shorter stays in hospital were significant. The median length of stay was 6.0 in the NAC group and 5.5 in the placebo arm.
Figure 1 The percentage of patients remaining in the study (i.e. who had not been discharged from hospital) on each day. N-acetylcysteine is shown by a dotted blue line and placebo by a solid red line. The life test procedure of SAS was used to compute nonparametric estimates of the length of stay function by the Kaplan-Meier method. Comparison between these functions was made using the Wilcoxon and log rank tests. Neither the log-rank statistic (p = 0.33) nor the Wilcoxon test (p = 0.30) were significant. The median length of stay was 6.0 days in the NAC group and 5.5 days in the placebo arm.
Three subjects reported adverse events in each group. One of the subjects treated with N-acetylcysteine reported nausea compared with two of the subjects treated with placebo. There were no serious adverse events.
Discussion
N-acetylcysteine has been consistently shown to reduce the number of exacerbations of COPD when it is taken for 3 months or more. In contrast we failed to show any benefit when N-acetylcysteine was administered as a treatment for acute exacerbations of COPD. There are a number of possible explanations for the failure to see any benefit.
We cannot exclude the possibility that there was a Type II error and that there is indeed a beneficial effect of N-acetylcysteine in the treatment of acute exacerbations. A larger study would be needed to rule out this possibility but there was less improvement in breathlessness, lung function and oxygen saturation with N-acetylcysteine than with placebo that argues against this explanation.
Another possibility that needs to be considered is that we used too low a dose of N-acetylcysteine and that the concentrations of N-acetylcysteine in the lung were not high enough to exert adequate antioxidant or anti-inflammatory effects. N-acetylcysteine is metabolized to cysteine and this in turn acts as a precursor of reduced glutathione which is an antioxidant [12]. Bridgeman and colleagues studied the effects of administering either 600 mg once daily or 600 mg three times daily [13]. After a single dose of 600 mg, N-acetylcysteine was detected in plasma for 1.5 hours. Plasma cysteine concentrations were also elevated but had returned to baseline by four hours. Glutathione concentrations were variably increased following a single dose of N-acetylcysteine but when N-acetylcysteine was given as 600 mg three times daily (t.i.d) for 5 days the glutathione concentrations were consistently and significantly elevated 12 hours post dose. In this study there was no increase in cysteine or reduced glutathione in either bronchoalveolar lavage fluid or lung tissue (from subjects undergoing pneumonectomy) when the samples were obtained 16–20 hours after the last dose of N-acetylcysteine. In an earlier study, however, reduced glutathione had been shown to be increased in bronchoalveolar lavage fluid 1 to 3 hours after a single dose of 600 mg of N-acetylcysteine [14]. It is likely that the dosing regimen that we used would lead to increases in cysteine and glutathione in both plasma and in the lungs but this may well not have been sustained over the whole of 24 hours. This leaves unanswered the question of whether the changes that did occur in N-acetylcysteine, cysteine and glutathione would have been sufficient to alter the course of the exacerbation. The results of studies where N-acetylcysteine was used as a prophylactic agent to prevent exacerbations of chronic bronchitis and/or COPD would argue that we did use an adequate dose. In these studies doses of N-acetylcysteine between 300 mg b.d. to 600 mg b.d. were effective and the dose that we used in this study is at the upper end of this range. N-acetylcysteine has also been shown to be effective for other indications when it has been used in this dose. Several studies have shown that N-acetylcysteine 600 mg b.d. protects against contrast nephropathy [15,16]. Whether or not a higher dose of N-acetylcysteine would have been any more effective in the treatment of acute exacerbations of COPD can only be answered by conducting additional studies.
In contrast to N-acetylcysteine, prednisone and prednisolone are effective treatments for acute exacerbations of COPD. Davies et al studied 56 patients admitted to hospital with an exacerbation of COPD and found that prednisolone led to a greater improvement in lung function and shortened the hospital stay [17]. Other studies have confirmed the efficacy of corticosteroids in severe exacerbations of COPD [18]. There is evidence of increased numbers of eosinophils in the airways during exacerbations of COPD. Corticosteroids are very effective at suppressing eosinophilic inflammation in the airways and this may account for the benefit seen in exacerbations of COPD. When children with an exacerbation of asthma are treated with prednisone, there is a marked reduction in the concentration of 8-isoprostane in exhaled breath condensate [19]. 8-isoprostane is a marker of oxidative stress. If N-acetylcysteine prevents exacerbations of COPD because it is an anti-inflammatory agent and/or antioxidant, it may be difficult to see additional benefit in established exacerbations of COPD when the patients are also treated with prednisone, which has anti-inflammatory actions and the potential to reduce formation of reactive oxygen species from inflammatory cells.
There have been a number of other studies looking at the effect of mucolytics in acute exacerbations of chronic bronchitis although none of these studies used N-acetylcysteine. Each of these studies has limitations. Langlands treated 27 patients, who had been admitted to hospital with an exacerbation of chronic bronchitis, with bromhexine 8 mg t.i.d. for two weeks (13 patients received bromhexine and 14 received placebo) [20]. In this study lung function was only measured twice a week during the study but the difference between treatments was not statistically significant. Maesen and his colleagues studied 22 patients admitted to hospital with an exacerbation of chronic bronchitis and purulent sputum [21]. All subjects received erythromycin and half were treated with bromhexine. Lung function was not measured but treatment with bromhexine did not influence the bacteriological response to erythromycin. Fimiguerra et al randomized 40 patients who had been admitted to hospital with an exacerbation of chronic bronchitis to treatment with amoxicillin alone (20 patients) or a combination of amoxicillin and domiodol (20 patients) for 10 days [22]. There was a three day washout period before treatment was initiated but it is not clear if this means that patients were in hospital for three days before treatment was started. Lung function was only measured at the beginning and end of treatment but there was no difference between the groups in changes in FEV1 and VC. Sputum volumes, however, were greater with the combination of the mucolytic and antibiotic. Ricevuti et al treated 24 patients with an exacerbation of chronic bronchitis [23]. Half of the patients were randomised to a combination of erdosteine and amoxicillin for seven days and the other received amoxicillin alone for the same period of time. Sputum viscosity and temperature resolved significantly more quickly with the combination but lung function was not measured in this study. None of these studies measured lung function on a daily basis and none assessed changes in breathlessness. This makes it difficult to know if treatment with these mucolytics influenced the rate of resolution of the exacerbations. On balance however these studies do not strongly suggest that mucolytics influence the resolution of acute exacerbations of COPD.
Conclusions
Our study does not suggest that 600 mg b.d. of N-acetylcysteine is effective in the treatment of patients who are admitted to hospital with an acute exacerbation of COPD and who receive concurrent treatment with corticosteroids. In future studies it may be appropriate to use a higher dose of N-acetylcysteine and to compare N-acetylcysteine with placebo in patients with mild exacerbations who do not require treatment with corticosteroids. Consideration could also be given to comparing the effects of N-acetylcysteine with prednisone or prednisolone in patients who would usually be treated with oral corticosteroids.
Abbreviations
NAC N-acetylcysteine
COPD chronic obstructive pulmonary disease
FEV1 forced expiratory volume in one second
VC vital capacity
SaO2 oxygen saturation
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
PNB conceived the idea for the study and was responsible for the study design and writing the manuscript. He was also was involved in the conduct of the study and the analysis of the data. AM-D was involved in the conduct of the study, in the analysis of the data and with writing the manuscript. PJP was involved with the design and conduct of the study. TEM and RPY were involved with the conduct of the study. All of the authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The study was funded by a grant from the Health Research Council of New Zealand. We would like to thank Greg Gamble for help with statistical analysis, and the pharmacists, nursing and medical staff at Auckland Hospital for their assistance with this study.
==== Refs
Seemungal TA Donaldson GC Paul EA Bestall JC Jeffries DJ Wedzicha JA Effect of exacerbations on quality of life in patients with chronic obstructive pulmonary disease Am J Respir Crit Care Med 1998 157 1418 1422 9603117
Wouters EF Economic analysis of the Confronting COPD survey: an overview of results Respir Med 2003 97 S3 14 12647938 10.1016/S0954-6111(03)80020-3
Poole PJ Black PN Systematic review of mucolytics for prevention of exacerbations in chronic bronchitis BMJ 2001 322 1271 4 11375228 10.1136/bmj.322.7297.1271
Gerrits CM Herings RM Leufkens HG Lammers JW N-acetylcysteine reduces the risk of re-hospitalisation among patients with chronic obstructive pulmonary disease Eur Respir J 2003 21 795 8 12765423 10.1183/09031936.03.00063402
Nagy AM Vanderbist F Parij N Maes P Fondu P Neve J Effect of the mucoactive drug Nacystelyn on the respiratory burst of human blood polymorphonuclear neutrophils Pulm Pharmacol Ther 1997 10 287 92 9778492 10.1006/pupt.1998.0106
Kasielski M Nowak D Long-term administration of N-acetylcysteine decreases hydrogen peroxide exhalation in subjects with chronic obstructive pulmonary disease Respir Med 2001 95 448 56 11421501 10.1053/rmed.2001.1066
Blesa S Cortijo J Mata M Serrano A Closa D Santangelo F Estrela JM Suchankova J Morcillo EJ Oral N-acetylcysteine attenuates the rat pulmonary inflammatory response to antigen Eur Respir J 2003 21 394 400 12661991
Saetta M Di Stefano A Maestrelli P Turato G Ruggieri MP Roggeri A Calcagni P Mapp CE Ciaccia A Fabbri LM Airway eosinophilia in chronic bronchitis during exacerbations Am J Respir Crit Care Med 1994 150 1646 52 7952628
Dekhuijzen PN Aben KK Dekker I Aarts LP Wielders PL van Herwaarden CL Bast A Increased exhalation of hydrogen peroxide in patients with stable and unstable chronic obstructive pulmonary disease Am J Respir Crit Care Med 1996 154 813 6 8810624
Kim S Nadel JA Role of neutrophils in mucus hypersecretion in COPD and implications for therapy Treat Respir Med 2004 3 147 59 15219174
Ram FSF Poole PJ Bagg W Stewart J Black PN Randomised, controlled trial of theophylline for the treatment of exacerbations of chronic obstructive pulmonary disease (abstract) Am J Respir Crit Care Med 2000 161 A489
Bonanomi L Gazzaniga A Toxicological, pharmacokinetic and metabolic studies on acetylcysteine Eur J Respir Dis Suppl 1980 111 45 51 6938410
Bridgeman MME Marsden M Selby C Morrison D MacNee W Effect of N-acetylcysteine on the concentrations of thiols in plasma, bronchoalveolar lavage fluid, and lung tissue Thorax 1994 49 670 675 8066561
Bridgeman MME Marsden M MacNee W Flenley DC Ryle AP Cysteine and glutathione concentrations in plasma and bronchoalveolar lavage fluid after the treatment with N-acetylcysteine Thorax 1991 46 39 42 1871695
Tepel M van der Giet M Schwarzfeld C Laufer U Liermann D Zidek W Prevention of radiographic-contrast agent- induced reduction in renal function by acetylcysteine N Engl J Med 2000 343 180 10900277 10.1056/NEJM200007203430304
Kay J Chow WH Chan TM Lo SK Kwok OH Yip A Fan K Lee CH Lam WF Acetylcysteine for prevention of acute deterioration of renal function following elective coronary angiography and intervention: A randomized controlled trial JAMA 2003 289 553 8 12578487 10.1001/jama.289.5.553
Davies L Angus RM Calverley PM Oral corticosteroids in patients admitted to hospital with exacerbations of chronic obstructive pulmonary disease: a prospective randomised controlled trial Lancet 1999 354 456 60 10465169 10.1016/S0140-6736(98)11326-0
Wood-Baker R Walters EH Gibson P Oral corticosteroids for acute exacerbations of chronic obstructive pulmonary disease Cochrane Database Syst Rev 2001 2 CD001288 11405984
Baraldi E Carraro S Alinovi R Pesci A Ghiro L Bodini A Piacentini G Zacchello F Zanconato S Cysteinyl leukotrienes and 8-isoprostane in exhaled breath condensates of children with asthma exacerbations Thorax 2003 58 505 509 12775861 10.1136/thorax.58.6.505
Langlands JHM Double-blind clinical trial of bromhexine as a mucolytic drug in chronic bronchitis Lancet 1970 i 448 450 4189753 10.1016/S0140-6736(70)90835-4
Maesen FPV Davies BI Brouwers J Rubingh G Erythromycin and bromhexine in acute exacerbations of chronic bronchitis Eur J Respir Dis 1982 63 325 329 6981522
Finiguerra M Conti P Figura I Legnani W Morandini GC Clinical study on the effects of an antibiotic and mucolytic association (amoxicillin and domiodol) in hypersecretory chronic bronchopulmonary diseases Curr Ther Res 1982 31 895 905
Ricevuti G Mazzone A Uccelli E Gazzani G Fregnan GB Influence of erdosteine, a mucolytic agent, on amoxicillin penetration into sputum in patients with an infective exacerbation of chronic bronchitis Thorax 1988 43 585 590 3051508
| 15581425 | PMC539269 | CC BY | 2021-01-04 16:30:10 | no | BMC Pulm Med. 2004 Dec 6; 4:13 | utf-8 | BMC Pulm Med | 2,004 | 10.1186/1471-2466-4-13 | oa_comm |
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-5-381553325510.1186/1471-2474-5-38Research ArticleSick leave among home-care personnel: a longitudinal study of risk factors Horneij Eva L [email protected] Irene B [email protected]öm Eva B [email protected] Charlotte [email protected] Ramlösa Clinic, Ramlösa Brunnshotell, Helsingborg, Sweden2 Department of Physical Therapy, Lund University, Lund, Sweden3 Section for Personal Injury Prevention, Karolinska Institute, Stockholm, Sweden2004 8 11 2004 5 38 38 24 6 2004 8 11 2004 Copyright © 2004 Horneij 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
Sick leave due to neck, shoulder and back disorders (NSBD) is higher among health-care workers, especially nursing aides/assistant nurses, compared with employees in other occupations. More information is needed about predictors of sick leave among health care workers. The aim of the study was to assess whether self-reported factors related to health, work and leisure time could predict: 1) future certified sick leave due to any cause, in nursing aides/assistant nurses (Study group I) and 2) future self-reported sick leave due to NSBD in nursing aides/assistant nurses (Study group II).
Methods
Study group I, comprised 443 female nursing aides/assistant nurses, not on sick leave at baseline when a questionnaire was completed. Data on certified sick leave were collected after 18 months. Study group II comprised 274 of the women, who at baseline reported no sick leave during the preceding year due to NSBD and who participated at the 18 month follow-up. Data on sick leave due to NSBD were collected from the questionnaire at 18 months. The associations between future sick leave and factors related to health, work and leisure time were tested by logistic regression analyses.
Results
Health-related factors such as previous low back disorders (OR: 1.89; 95% CI 1.20–2.97) and previous sick leave (OR 6.40; 95%CI 3.97–10.31), were associated with a higher risk of future sick leave due to any cause. Factors related to health, work and leisure time, i.e. previous low back disorders (OR: 4.45; 95% CI 1.27–15.77) previous sick leave, not due to NSBD (OR 3.30; 95%CI 1.33–8.17), high strain work (OR 2.34; 95%CI 1.05–5.23) and high perceived physical exertion in domestic work (OR 2.56; 95%CI 1.12–5.86) were associated with a higher risk of future sick leave due to NSBD. In the final analyses, previous low back disorders and previous sick leave remained significant in both study groups.
Conclusion
The results suggest a focus on previous low back disorders and previous sick leave for the design of early prevention programmes aiming at reducing future sick leave due to any cause, as well as due to NSBD, among nursing aides/assistant nurses. A multifactorial approach may be of importance in the early prevention of sick leave due to NSBD.
==== Body
Background
Over the past century, sick leave, mainly attributed to musculoskeletal disorders (MSD), has increased in Sweden, especially among women [1]. In 2003, the proportion of women on sick leave was higher than in any previous year [1], indicating more individual suffering and also increased cost for the community. The prevalence of neck, shoulder and back pain is higher among health care workers, especially among nursing aides, compared with employees in other occupations [2-4]. In 2002, women working within the health-care sector in Sweden had the highest proportion of sick leave, mainly attributed to MSD [1]. The need for effective early prevention strategies is evident. However, if customised prevention programmes are to be possible, more information is needed about predictors of future sick leave among women working in the nursing sector.
Many studies have focused on the association between risk factors and the reporting of neck, shoulder and back pain [5-8]. However, back pain is a recurrent problem, which may or may not influence participation in daily activities and the ability to work [9,10]. It has been shown that predictors of perceived pain may not be the same as predictors of future sick leave [11]. Thus, when constructing intervention programmes, focusing on individuals at risk of future sick leave, predictors associated with perceived pain should be distinguished from those of future sick leave.
Studies have pointed to different factors in different populations to relate with future sick leave. Demographic factors such as age may be associated with sick leave mainly in individuals with chronic neck-or-back disorders [12]. In some studies this association was, however, not verified, which may be due to a "healthy worker effect" [11,13]. Medical factors, such as, for example, previous experience of back disorders, perceived health or prior sick leave predicted future sick leave [11,12,14-16]. Self-reported high physical load increased the risk of prolonged sick leave in populations already on sick leave due to back pain [17,18]. Perceived high work load in combination with low decision latitude, hereafter called high-strain work, is assumed to have negative consequences on health [19,20]. Low social support may be a risk factor for the development of back pain and also for future sick leave [19,20]. The association between high work load, low decision latitude and low social support and future sick leave is, however, unclear [14,19].
Mental problems, such as, for example, anxiety, also increased the risk of future sick leave among nurses [19]. Moreover, economic problems exacerbated the risk of prolonged sick leave among patients with low back disorders [21]. A strained economic situation also tends to lead to a decreased probability that woman suffering low back pain will seek medical attention [22].
The effects of physical exercising on future sick leave are not clear. Sedentary activities in leisure time were, however, associated with a higher prevalence of back symptoms and sick leave [23]. Eriksen et al. [16,24] found that regular physical activity, such as brisk walks, aerobics or other forms of exercise for 20 minutes or more at least once a week predicted fewer sick leaves after 3 and 15 months among nursing aides.
To summarise, a large number of factors, demographic, medical, physical, psychosocial, psychological as well as socioeconomic, have been documented as risk factors for future sick leave in different working populations [14].
It has been concluded that there is a lack of studies on neck-and-back disorders among women working within special groups as for example the health-care sector [14]. Because studies on home-care personnel, analysing risk factors associated with neck, shoulder and back disorders, have mostly been cross-sectional, it is not possible to determine causality.
It is of interest to analyse whether the above-mentioned factors associated with future sick leave in different populations, can also predict future sick leave among nursing aides/assistant nurses working within the home-care service.
Working as a nursing aide/assistant nurse within the home-care service is generally physically heavy work, requiring repeated transfers and lifts of patients. The physical load on the spine depends on several factors, for example, the weight of the patient, work place design, work organisation, work technique, work equipment, the cooperation of the patient etc [25]. Even transfers of a light and cooperative patient imply high spinal loads and a risk of causing low back disorders [26]. An appropriate work technique may decrease the biomechanical load on the spine. However, work technique differs among individuals and should thus be studied on an individual level [25]. In the present study, self-reported measures were analysed and physical load on the spine were not included. Self-reported working positions did not associate with prolonged sick leave in individuals with chronic back pain [14]. In longitudinal studies on nurses and nursing aides, frequent lifting or repositioning of patients did not predict future sick leave [11,16]. However, the perception of physical exertion at work was a risk indicator for low back symptoms among nursing aides working in geriatric care [6]. The perception of physical exertion correlated also with the onset of low-back pain among nursing aides [7]. The risk of seeking care was higher among the nursing personnel who perceived high physical exertion in domestic work [27]. The impact of perception of physical exertion on future sick leave among nursing aides/assistant nurses is, to our knowledge, not known.
The aim of the present study was to assess whether a selection of self-reported measures of factors related to health, work and leisure time could predict
1. future, certified sick leave after 18 months, due to any cause, in a group of working nursing aides/assistant nurses (Study group I).
2. future, self-reported sick leave after 18 months, due to neck, shoulder and/or back disorders, in a group of working nursing aides/assistant nurses (Study group II).
The aim was based on the hypothesis that factors related to the workplace as well as factors related to perceived health and to leisure time are associated with future sick leave in female nursing aides/assistant nurses.
It was also hypothesised that predictors of future sick leave due to neck, shoulder and/or back disorders are different from predictors of sick leave due to any cause.
Methods
This study was one part of a larger project, aiming at preventing or reducing disorders of the neck, shoulder or back among female nursing aides/assistant nurses, working within the home-care services. The definition of the titles of home-care personnel is heterogeneous. In the present paper, nursing aides had little or no formal training, while the assistant nurses had undergone two – three years secondary education in nursing or had long experience of the work as a nursing aide and about one-year further education.
The study was approved by the Ethical Committee of the Faculty of Medicine, University of Lund, Sweden (LU 286-95). All participants gave their written consent before participation.
Subjects
The municipal home-care services were organised in six units situated in different geographically defined areas of a medium-sized town in the south of Sweden. Initially, all of the 659 women working in five of the six units were invited to participate in the prevention project. (One unit was excluded due to its participation in another study). The inclusion criteria were: Swedish speaking, permanently employed, not pregnant, in work and working at least 50% of full time. Participation was accepted by 534 (81%) of the women. The main reasons not to participate were: the opinion that the project was important only for younger staff, dissatisfaction with the work situation, participation in compulsory, further education for nursing aides to qualify as assistant nurse, lack of time or family reasons. As we wanted to elucidate predictors of future sick leave at an 18 month follow-up, women who at that time had retired from work, had left work or were off duty, were excluded from the analyses. Two women were deceased (n = 91). Thus, the final sample consisted of 443 nursing aides/assistant nurses (Figure 1).
At baseline, totally 282 of the originally 534 participants were randomized to one of two intervention groups, aiming at preventing neck, shoulder and back disorders (SM = Stress Management or IT = Individual Physical Training programmes) or to a Control group. Intervention groups and the Control group were evaluated and compared at the 18-month follow-up. No significant differences between the groups could be shown [39]. Thus, in the present study the groups on intervention programmes and the Control group were treated as a whole and called intervention.
Study group I
Study group I comprised 443 women who were not on sick leave when they completed the baseline questionnaire. Information on certified sick leave due to any cause during the six months preceding the 18-month follow-up was obtained from the National Social Insurance Board. The number of women who participated in intervention was 241 (54%). Demographic data, disorders, sick leave and participation in intervention for Study group I are presented in Table 1.
Study group II
Study group II comprised 274 women. At baseline, 383 women had reported no sick-leave due to symptoms from the neck, shoulder and/or back during the preceding 12 months. At the 18-month follow-up, 109 of these women did not fill in the questionnaire (Figure 1). The main known causes for dropping out at the 18-month follow-up were, not being able to fill in the questionnaires in time, mainly due to vacations, illnesses and refusal to take part, mainly due to lack of time. The reason for not participating was unknown for 69 of the women (18%). However, when further contacts were established with these women at a later stage, we were told that the reason for not responding was frequent reorganisations at work. Compared with participants, non-participants were, at baseline, more dissatisfied with social support at work (p = 0.03). No other differences between the participants and the drop-outs were shown. The number of women who participated in some intervention was 173 (64%). Demographic data, disorders, sick leave and participation in intervention for Study group II are presented in Table 1.
Assessments
All participants were asked to fill in a questionnaire at the start of the study and after 6, 12 and 18 months. At baseline, the questionnaires were administered by the project nurse and filled in at the work place (Study group I and Study group II). At the 18-month follow-up, questionnaires were sent out to a contact person at each work place, who was made responsible for the distribution to the participants in the study. Each respondent received a stamped envelope in which she returned her own questionnaire to the project nurse (Study group II).
Dependent variables
Sick leave due to any cause at the 18-month follow-up (Study group I)
Data about sick leave during the six months preceding the 18-month follow-up were obtained from the register of the National Social Insurance Board. In Sweden, the municipal authorities systematically record and report each day an employee is on sick leave to the National Social Insurance Board. Diagnoses are not reported to the Board. All days were counted as whole days irrespective of whether they were whole days or part of days. Due to skewed data, sick leave was dichotomized in 0 days/≥1 day (third quartile).
Sick leave due to disorders of the neck, shoulders and/or back at the 18-month follow-up (Study group II)
At all follow-ups the nursing aides/assistant nurses were asked: "Have you been on sick leave any time during the previous six months due to neck, shoulder and/or back disorders?". Response options were yes/no and constituted the dependent variable for Study group II at the 18-month follow-up. The validity of the responses was checked for all follow-up questionnaires against the general Nordic Musculoskeletal Questionnaire (NMQ) [28] i.e. the question concerning whether the disorder had been incapacitating during the preceding six months. A pattern was seen for women who reported sick leave due to neck, shoulder and/or back disorders at follow-ups by the fact that they reported in the NMQ that the disorders also had been incapacitating some time during the 6 months preceding the 18 month follow-up.
Independent variables
Self-reported measures associated with health
Mental health
Anxiety and depression were assessed by the Hospital Anxiety and Depression Scale (HAD) [29] which consists of two subscales – one for anxiety and one for depression. Anxiety and depression were closely related (p = 0.000) and as depression levels are generally lower compared with anxiety levels in working populations [30], which was also the case in the present study, we selected anxiety as a measure of mental well-being. The subscale contains seven items ranging from 0 – 3, with higher scores reflecting greater anxiety. A sum of eight or more has been shown in comparisons with psychiatric interviews to reflect anxiety [29]. The Swedish version was tested and evaluated by Lundqvist et al. [31].
Musculoskeletal disorders
The prevalence of musculoskeletal disorders from the neck, shoulders and back was assessed by the general Nordic Musculoskeletal Questionnaire (NMQ) [28]. Participants were asked about pain, aches or discomfort some time during the preceding 12 months. The response options were yes/no.
Sick leave during the 12 months preceding baseline
Information on sick leave was obtained from the National Social Insurance Board.
Individuals in Study group II, who had been on sick leave, but who in the questionnaires had reported no sick leave due to symptoms from the neck, shoulder and/or back were assumed to be on sick leave due to other reasons. Work loss was recorded as whole working days. As all data on sick leave were extremely skewed, we chose to dichotomise the material. The third quartile was used as cut-off point, which for Study group I was 0 days/≥1 days and for Study group II <9 days/≥9 days.
Self-reported measures associated with work
Perceived physical exertion at work
The participants were asked: "What degree of physical exertion do you usually perceive in your present job?" [32]. The question was assessed according to Borg [33] and ranged from 6 (no exertion at all) to 20 (maximal exertion). High physical exertion at work was defined as 15 on the scale (corresponding "hard") or more (third quartile) [6].
Perceived work-related psychosocial factors
Psychosocial factors at work, such as social support, decision latitude and psychological load were assessed by a questionnaire developed by Rubenowitz [34,35]. The questionnaire considers five psychosocial factors: "Influence and control over work", "Supervisor climate", "Stimulus from the work itself", "Relation to fellow workers" and "Psychological load". Each factor comprises five items and each item has five fixed response alternatives from 1 to 5, where 1 means very unsatisfactory and 5 very satisfactory. A separate score, ranging from 1 to 5, is calculated on the mean of each factor. In the present analysis, "Supervisor climate" and "Relation to fellow workers" were defined as "Social support" and "Influence and control over work" and "Stimulus from the work itself" were defined as "Decision latitude" [35]. "Psychological load" together with "decision latitude" were defined as strain. High strain was equal to high psychological load in combination with low decision latitude [36]. The first quartile of the psychosocial factors were categorised as poor [34].
Self-reported measures associated with leisure time
Exercise and physical activity
The participants were asked: "To what extent have you performed physical activities or fitness training during the previous six months?" [32]. The question comprised eight options. A sedentary life style was assumed by the response "No exercise and very little physical activity" (score = 1).
Perceived physical exertion in domestic work
The participants were asked about physical exertion in domestic work in the same way as for physical exertion at work: "What degree of physical exertion do you usually perceive in your daily domestic work?" [32]. Response options ranged from 6 (no exertion at all) to 20 (maximal exertion) [33]. The cut-off point was 13 (third quartile) corresponding "somewhat hard".
Perceived psychological stress outside work
This question, originally developed for the Malmo Diet and Cancer study on 53000 Swedish men and women, was defined by "Have you lately felt mentally stressed or been under psychological pressure due to problems outside work?" Response options were yes/no [37].
Economic situation
Perceived satisfaction with her own economic situation was assessed by a seven-point scale where the first point represented "very bad" and the seventh point "Excellent, could not be better" [38].
Statistical analysis
All logistic regression analyses were adjusted for age and intervention.
Three models were tested, namely, for factors related to health, to work and to leisure time. The associations between the independent variables and the outcome variables were first analysed for each of the three models by univariate logistic regression. Secondly, each model was analysed with all variables included in a backward stepwise multivariate regression analysis with a likelihood ratio test. Finally, a multivariate model including all the variables from the three models together was tested. The criteria for inclusion and exclusion were p = 0.05. The logistic regression analyses were tested for goodness of fit by means of the Hosmer and Lemeshow method. Factors were dichotomized by using cut-off points described in the literature. When no such literature was found, cut-off points were taken as the first or the third quartile for frequencies for Study group II. There were minimal differences between the quartiles of the two study groups. Thus, for the purpose of comparing the two groups, the same cut-off points were used. However, for sick-leave before baseline we chose two different cut-off points, as the third quartile differed between the two study groups, being 0 days for Study group I and 9 days for Study group II.
The sample size may vary in the different analyses due to missing values. Comparisons between drop-outs and participants were analysed by t-test or the chi-square test. Correlations between the exposure variables were calculated with Pearson correlation coefficients (r). All statistical calculations were performed using the SPSS 11.5.1 Software for Windows (SPSS Inc., Chicago, IL, USA).
Results
In the univariate logistic regression analysis, there were minimal differences between odds ratios adjusted and not adjusted for age and intervention. Thus, in Tables 2,3,4,5, only adjusted odds ratios are presented.
Predictors of future sick leave due to any cause (Study group I)
Age or participation in intervention (presented under section Methods-Subjects) did not have any effect on future sick leave due to any cause.
Self-reported factors related to health
Perceived disorders of the low back and sick leave during the 12 months preceding baseline predicted future sick leave due to any cause in the unadjusted and the adjusted univariate analyses as well as in the multivariate analysis (Table 2 and 5). Perceived disorders of the neck or shoulders did not predict future sick leave. Neck, shoulder and back disorders at baseline correlated with each other. Of the women who reported neck disorders, 88% also reported shoulder disorders (r = 0.57) and 75% reported low back disorders (r = 0.33). Of the women who reported shoulder disorders 71% reported low back disorders (r = 0.28).
Self-reported factors related to work
Factors related to work did not predict future sick leave due to any cause (Table 3 and 5).
Self-reported factors related to leisure time
Factors related to leisure time did not predict future sick leave in the univariate or in the multivariate analysis (Tables 4 and 5).
Final model Study group I
When all variables from the three models were entered into the final model, disorders of the low back during the preceding 12 months (OR: 1.95; CI: 1.23–3.10) and sick leave during the 12 months preceding baseline (OR: 6.61; CI: 4.08–10.73) were the only factors predicting sick leave due to any cause after 18 months.
Predictors of future sick leave due to disorders of the neck, shoulders and/or back (Study group II)
A tendency towards a reduced risk of future sick leave due to neck, shoulder and/or back disorders was found for the women who participated in some intervention (OR: 0.46; CI: 0.21–1.01). Age did not have an impact on sick leave (OR: 1.02; CI: 0.98–1.06).
Self-reported factors related to health
Anxiety, low back disorders during the 12 months preceding baseline as well as sick leave due to reasons other than disorders of the neck, shoulders and/or back were significant predictors of sick leave after 18 months due to disorders of the neck, shoulders and/or back (p < 0.05) before and after adjustment for age and intervention (Table 2). When all characteristics related to health were entered into the multiple logistic regression model, disorders of the low back and sick leave due to reasons other than disorders of the neck, shoulders and/or back remained significant (Table 5). Neck disorders at baseline, correlated with shoulder disorders (r = 0.57) and with back disorders (r = 0.36). Of the nursing aides/assistant nurses who reported neck disorders, 87% also reported disorders of the shoulders and 75% reported disorders of the low back. Of those who reported disorders of the shoulders at baseline, 70% also reported disorders of the low back (r = 0.29).
Self-reported factors related to work
Before adjustment for age and intervention, high psychological load and high-strain work predicted future sick leave due to neck, shoulder and/or back disorders (OR: 2.28; CI:1.04–5.03 and OR: 2.42; CI: 1.09–5.37). However, after adjustments, only high-strain work demonstrated significant association with future sick leave (Table 3).
Two multivariate regression analyses were performed: one by entering psychological load and decision latitude separately into the model and one by analysing the same factors together (called strain). Only high-strain work predicted future sick leave due to disorders in the neck, shoulder and/or back (Table 5).
Self-reported factors related to leisure time
High perceived physical exertion in domestic work and perceived psychological stress outside work were associated with future sick leave due to neck, shoulder and/or back disorders both before and after adjustments for age and intervention (Table 4). In the multivariate model, high perceived physical exertion in domestic work remained significant predictor of future sick leave (Table 5).
Final model, Study group II
In the final model, where all variables from the three models were entered into the logistic regression analysis, perceived low back disorders some time during the 12 months preceding baseline and sick leave of > 9 days in the year before baseline remained significant predictors of sick leave due to neck, shoulder and/or back disorders at 18 months after baseline (OR: 7.36; CI: 1.67–32.43 and OR: 2.84; CI: 1.13–7.11 respectively).
Discussion
The results of the present study indicated that only factors related to health were associated with future sick leave due to any cause in a group of nursing aides/assistant nurses. On the other hand, factors related to health, the workplace as well as leisure time were associated with future sick leave due to neck, shoulder and/or back disorders, indicating a multifactorial background for these disorders. Low back disorders and sick leave during the 12 months preceding baseline, were significantly associated with future sick leave in both Study group I and Study group II. Thus, the hypothesis that risk factors for future sick leave due to disorders in the neck, shoulders and/or back are different from sick leave due to any cause was only partly demonstrated. This is in concordance with other studies, which have also documented earlier experience of back pain and sick leave preceding baseline as predictors of future sick leave due to low back pain among nursing personnel [11] as well as in general populations [12,40]. Natvig et al. [15], showed that low back pain as a part of widespread pain predicted long-term disability due to any cause, while local low back pain did not. In the present study perceived disorders of the neck, shoulder and back correlated with each other, indicating wide spread disorders. Further, only disorders of the back and not of the neck or of the shoulders predicted sick leave in Study group I as well as in Study group II. One explanation for this finding may be the content of the work performed by a nursing aide/assistant nurse, with repeated repositioning, transfers and lifts of patients, which result in high spinal loads mainly on the low back [25,26]. Thus, it may be more difficult to perform this work with a vulnerable low back in comparison with vulnerable neck/shoulders. In the final analyses of Study group I and Study group II, including all variables from the three models, only low back disorders and previous sick leave remained significant. However, in Study group II, due to the small number of women on sick leave due to neck, shoulder and/or back disorders at the 18-month follow-up, results of this final analysis should be interpreted with caution.
Generally, most studies on predictors of future sick leave are made on men and few studies focusing on nursing personnel have been published. Thus, comparisons of the results from the current study with similar studies are limited.
The present study was performed during a period with frequent reorganisations and reductions of staff within the home-care services in the city studied, as in general in Sweden. A greater amount of work was performed faster than before and the category of patients, for whom the community home-care organisation took responsibility for, were more handicapped than previously. Thus, the physical and psychological stresses were generally exacerbated during the period studied. In the present study, among the factors related to work, only high-strain work (high psychological load in combination with low decision latitude) predicted future sick leave due to neck, shoulder and/or back. Marras et al. [41] showed that lifting combined with psychosocial stress, increases the load on the spine indicating an increased risk of back pain when working in physically demanding jobs in combination with high strain. Bourbonnais and Mondor [19] found that high-strain work reinforced the risk of future, short-term sick leave due to any cause among nurses. The inclusion of strain-reducing strategies in stress-management programmes for nursing aides/assistant nurses may, thus, decrease the risk of future sick leave due to back disorders, but should be further studied.
Women still do more unpaid work than men. In a 23-year perspective study, the physical work load decreased among men but not among women [42]. In the present study, we found that high physical exertion in domestic work predicted sick leave due to neck, shoulder and/or back disorders. Josephson et al. [27] found that the tendency to seek care was greater among nursing personnel who perceived that they did far too much domestic work. These findings point to the importance of also including factors outside work, even when customising early prevention programmes at the work place with the aim to reduce future sick leave due to neck, shoulder and back disorders among female nursing aides/assistant nurses.
We could not verify inactivity to be a predictor of future sick leave. The cut-off point was based on the assumption that a sedentary life style would predict future sick leave as shown in the study by Hildebrandt et al. [23]. In the study by Eriksen et al. [24], brisk walks, aerobics or gymnastics and other physical leisure-time activities for 20 minutes or more at least once a week protected against sick leave of more than 14 days, independent of diagnosis, after a 15-month follow-up period among nursing aides. The work as a nursing aide/assistant nurse is physically active, involving a great deal of walking. Few women reported a sedentary life at baseline. Thus, inactivity may not have been an appropriate cut-off point.
Moreover, in the present study, social support did not predict future sick leave. Since there was a significant initial difference in perceived social support between participants and non-participants in Study group II, the conclusions of this analysis are limited. However, neither in the Study group I, was social support related to sick leave. This finding is contradictory to the results of the study by Eriksen et al. on 5563 Norwegian nursing aides [16]. These authors found that social support was the most important work-related predictor of future sick leave due to any cause. Bourbonnais and Mondor [19] could also state a relationship between low social support and future sick leave among Canadian nurses. However, after adjustment for job strain, this relationship was no longer significant.
Methodological considerations
Data on the dependent variable sick leave, for Study group I were obtained from the National Social Insurance Board and for Study group II were taken from the questionnaires. Systematically recorded, certified sick leave is assumed to be more valid than data on sick leave based on questionnaires [43]. For Study group II, questionnaires were filled in at the 6, 12 and 18 month follow-ups. In order to strengthen the validity of the question "Have you been on sick leave any time during the preceding six months due to neck, shoulder and/or back disorders", answers at the follow-ups were checked against the NMQ [28]. Participants who at baseline reported no sick leave due to neck, shoulder and/or back disorders could nevertheless, in the NMQ, indicate perceived disorders from the same regions but that these disorders had not been incapacitating. However, a pattern was seen for women who reported sick leave due to neck, shoulder and/or back disorders at follow-ups by the fact that they also, in the NMQ, reported that these disorders had been incapacitating. It may also be assumed that the recall bias was reduced and the sensitivity of the question about sick leave enhanced, as we did not ask the respondent to specify the number either of sick-days or of sick-leave episodes [43].
Age was not related to any reason for sick leave, which is in contrast with the results of a review study by Turner et al. [12] on people working in various occupations. They found an increased risk of sick leave with older age. However, the results of the present study are in concordance with other studies on home-care personnel and future sick leave [11,13], indicating a healthy worker effect among nursing aides/assistant nurses, which in turn may result in an underestimation of risk factors for future sick leave.
In Study group II, a large number of participants dropped out at the 18-month follow-up, which constitutes a risk of selection bias. The number of drop-outs is comparable with that found in other studies on nursing personnel [19,30,44]. In addition, except for social support at work (discussed earlier), participants did not differ from non-participants on any of the dependent variables. Thus, we regard the 274 participants in Study group II to be representative as a working group of nursing aides/assistant nurses.
The feasibility to generalise the results from Study group I to other women working as nursing aides/nursing assistants should be satisfactory, as all women participating at baseline were analysed at18 months and the data of the dependent variable, certified sick leave, were provided by the National Social Insurance Board.
In Study group I, 54% and in Study group II, 64% of the nursing aides/assistant nurses had participated in some intervention. The main purpose of this intervention project was not to reduce sick leave. The participants were informed that the aim of the project was to prevent or reduce pain and discomfort of the neck, shoulder and back. Sick leave was not mentioned. Thus, in our opinion, participation in intervention groups should not have biased the reporting of sick leave due to neck, shoulder and back disorders.
Conclusions
The present study indicated that previous disorders of the low back and previous sick leave were the strongest predictors of future sick leave due to any cause as well as of future sick leave due to neck, shoulder and/or back disorders in a group of nursing aides/assistant nurses, who at baseline were at work. Previous neck or shoulder disorders did not predict future sick leave.
Moreover, factors related to health, work and leisure time were all related to sick leave due to neck, shoulder and/or back disorders, while factors related only to health predicted sick leave due to any cause.
The results point to the importance of a primary focus on previous low back disorders and previous sick leave when designing early prevention programmes for future sick leave among this working population. The results might also point to the importance of a multifactorial approach when customising early prevention programmes with the purpose to decrease future sick leave due to neck, shoulders and/or back disorders in a group of women working as nursing aides/assistant nurses.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EHj participated in the design of the study, participated in collecting the data, performed the statistical analyses, and drafted the manuscript. IJ and EHm participated in the design of the study and in the progress and revision of the manuscript. CE participated in the progress 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:
Acknowledgements
We wish to thank The Swedish Labour Market Insurance, AFA for financial support. We also thank Per-Erik Isberg for his advice on statistics.
Figures and Tables
Figure 1 Study profile
Table 1 Demographic data, sick leave, proportions of symptoms from the neck, shoulder and low back during the 12 months preceding baseline, and proportions of women participating in some intervention.
Study group Ia
n = 443 Study group IIb
n = 274
Age, mean (SD)
45 years
(± 10) 45 years
(± 10)
Symptoms some time during the 12 months before baseline
Neck 59% 57%
Shoulders 66% 65%
Low back 62% 60%
Number of days on sick leave due to any cause 12 months before baseline/6 months before 18-month follow-up, mean (SD)
14 (38)/
12 (31)
Numbers of days on sick leave 12 months before baseline (other reasons than disorders of the neck, shoulder and/or back)/ due to any cause 6 months before 18-month follow-up, mean (SD)
10 (30)/
9 (30)
Duration of employment
0 – 4 years 3% 3%
5 – 10 years 22% 20%
11 – 20 year 47% 43%
>20 years 28% 34%
Proportion of full-time work, mean (SD)
81% (± 17) 81% (± 17)
N:o of persons in household (including the person involved), mean (SD)
2,7 (± 1.3) 2,8 (± 1.3)
Living alone
17% 16%
Participation in intervention
Stress Management group 18% 21%
Individual Physical Training group 16% 18%
Control group 20% 25%
a All participants at baseline. b Participants who at baseline reported no sick leave due to neck, shoulder and/or back disorders during the preceding 12 months and who answered the questionnaire at the 18-month follow-up.
Table 2 Self-reported factors related to health and associations with sick leave after 18 months due to any cause (Study group I)a and due to neck, shoulder and/or back disorders (Study group II)b.
Number observed
Number observed
Odds Ratio
Odds Ratio
P value
P value
Variable
Study gp Ia (n = 443)
N (cases)
Study gp IIb (n = 274)
N (cases)
Study gp I OR
(95%CI)
Study gp II OR (95%CI)
Study gp I
Study gp II
Anxiety [0c -21]
- <8 332 (164) 217 (16)
- ≥8 102 (55) 55 (11) 1.18
(0.76–1.85) 3.04 (1.31–7.06) 0.462 0.010
Neck disorders during the year before baseline
- No 175 (79) 113 (10)
- Yes 254 (137) 152 (16) 1.39
(0.94–2.06) 1.10 (0.47–2.55) 0.096 0.830
Shoulder disorders during the year before baseline
- No 146 (67) 96 (10)
- Yes 281 (147) 172 (16) 1.29
(0.86–1.93) 0.79 (0.34–1.85) 0.214 0.592
Low back disorders during the year before baseline
- No 164 (64) 107 (4)
- Yes 265 (152) 160 (22) 2.09 (1.40-3.12) 3.79 (1.26–11.44) <10-3 0.018
Study group I
Any sick leave during the year before baseline
- 0 days 162 (35)
- ≥1 days 281 (189) 7.67
(4.86–12.12) <10-3
Study group II
Sick leave during the year before baseline (other reasons than disorders of the neck, shoulder and/or back)
- <9 days 204 (16)
- ≥9 days 70 (12) 2.64 (1.12–6.0) 0.021
a All participants at baseline. b Participants who at baseline reported no sick leave due to neck, shoulder and/or back disorders during the preceding 12 months and who also answered the questionnaire at the 18-month follow-up. Minimum and maximum values are presented within square brackets. c Best possible value. CI = 95% confidence interval. Odds Ratios are adjusted for age and intervention
Table 3 Self-reported factors related to work and associations with sick leave after 18 months due to any cause (Study group I)a and due to neck, shoulder and/or back disorders (Study group II)b.
Number observed
Number observed
Odds Ratio
Odds Ratio
P value
P value
Variable
Study gp Ia
(n = 443) N (cases)
Study gp IIb
(n = 274) N (cases)
Study gp I OR
(95%CI)
Study gp II OR
(95%CI)
Study gp I
Study gp II
Perceived physical exertion at work
[6c -20]
- <15 212 (102) 134 (12)
- ≥15 229 (121) 140 (16) 1.21
(0.83–1.76) 1.33
(0.60–2.94) 0.318 0.487
Social support [1d-10]
- >7.2 329 (168) 69 (9)
- ≤7.2] 114 (56) 201(19) 0.94
(0.61–1.44) 0.77
(0.32–1.81) 0.770 0.54
Decision latitude [1d-10]
- >6.2 311 (159) 86 (12)
- ≤6.2 132 (65) 185 (16) 0.97
(0.64–1.46) 0.61
(0.27–1.37) 0.873 0.234
Psychological load [1d-5]
- >2.6 303 (145) 183 (14)
- ≤2.6 140 (79) 88 (14) 1.44
(0.96–2.17) 2.15
(0.97–4.77) 0.077 0.060
Strain (Decn-Latde + Psych load) [1d-15]
- >8.8 321 (160) 194 (15)
- ≤8.8 122 (64) 77 (13) 1.16
(0.76–1.76) 2.35
(1.05–5.26) 0.500 0.037
a All participants at baseline. b Participants who at baseline reported no sick leave due to neck, shoulder and/or back disorders during the preceding 12 months and who also answered the questionnaire at the 18-month follow-up. Minimum and maximum values are presented within square brackets. c Best possible value. d Worst possible value. CI = 95% confidence interval. Odds Ratios are adjusted for age and intervention
Table 4 Self-reported factors related to leisure time and associations with sick leave after 18 months due to any cause (Study group I)a and due to neck, shoulder and/or back disorders (Study group II)b.
Number observed
Number observed
Odds Ratio
Odds Ratio
P value
P value
Variable
Study gp Ia
(n = 443)
N (cases)
Study gp IIb
(n = 274)
N (cases)
Study gp I OR
(95%CI)
Study gp II OR (95%CI)
Study gp I
Study gp II
Physical activity and exercise
- Physical activity and/or at least some exercise 422 (215) 261 (27)
- No exercise and very little physical activity 17 (7) 12 (1) 0.67
(0.25–1.79) 0.77
(0.09–6.28) 0.418 0.805
Perceived physical exertion in domestic work [6c -20]
- <13 285 (147) 180 (13)
- ≥13 156 (77) 88 (15) 0.96
(0.65–1.44) 2.36
(1.05–5.31) 0.855 0.038
Perceived psychological stress outside work
- No 305 (160) 185 (13)
- Yes 137 (63) 88 (14) 0.79
(0.53–1.19) 2.40
(1.06–5.43) 0.253 0.035
Satisfaction with economy [1d-7]
- >4 282 (138) 179 (19)
- ≤4 157 (84) 93 (8) 1.29
(0.86–1.93) 0.77
(0.32–1.90) 0.218 0.577
a All participants at baseline. b Participants who at baseline reported no sick leave due to neck, shoulder and/or back disorders during the preceding 12 months and who also answered the questionnaire at the 18-month follow-up. Minimum and maximum values are presented within square brackets. c Best possible value. d Worst possible value. CI = 95% confidence interval. Odds Ratios are adjusted for age and intervention
Table 5 Significant predictors of future sick leave due to any cause (Study group I)a and due to neck, shoulder and/or back disorders (Study group II)b for factors related to health, work and leisure time respectively. Multiple regression analyses for each model.
Study group Ia
Study group IIb
Study group I
Study group II
Models
(n = 443) OR
(95%CI)
(n = 274) OR
(95%CI)
P value
P value
Predictors related to health
- Low back disorders during the year before baseline 1.89
(1.20–2.97) 4.45
(1.27–15.77) 0.006 0.007
- Any sick leave during the year before baseline 0 days/≥1 days 6.40 (3.97–10.31) <10-3
- Sick leave during the year before baseline (other reasons than disorders of the neck, shoulder and/or back) < 9 days/≥9 days 3.30
(1.33–8.17) 0.011
Predictors related to work
- High strain work - 2.34
(1.05–5.23) - 0.041
Predictors related to leisure time
- High perceived physical exertion in domestic work - 2.56
(1.12–5.86) - 0.026
a All participants at baseline. b Participants who at baseline reported no sick leave due to neck, shoulder and/or back disorders during the preceding 12 months and who also answered the questionnaire at the 18-month follow-up. OR = Odds Ratio. CI = 95% confidence interval. Odds Ratios are adjusted for age and intervention
==== Refs
The National Social Insurance Board (in Swedish)
Cohen-Mansfield J Culpepper WJ IICarter P Nursing staff back injuries. Prevalence and costs in long tem care facilities AAOOHN J 1996 44 9 17
Guo HR Working hours spent on repeated activities and prevalence back pain Occup Environ Med 2002 59 680 688 12356929 10.1136/oem.59.10.680
Silverstein B Viikari-Juntura E Kalat J Use of a prevention index to identify industries at high risk for work-related musculoskeletal disorders of the neck back, and upper-extremity in Washington state, 1990–1998 Am J Ind Med 2002 41 149 169 11920960 10.1002/ajim.10054
Alexopoulos EC Burdorf A Kalokerinou A Risk factors for musculoskeletal disorders among nursing personnel in Greek hospitals Int Arch Occup Environ Health 2003 76 289 294 12739172
Josephson M Hagberg M Wigaeus Hjelm E Self-reported physical exertion in geriatric care – a risk indicator for low back symptoms? Spine 1996 21 2781 2785 8979326 10.1097/00007632-199612010-00014
Gonge H Donbæk Jensen L Bonde JP Do psychosocial strain and physical exertion predict onset of low-back pain among nursing aides? Scand J Wok Environ Health 2001 27 388 394
Feyer AM Herbison P Williamson AM de Silva I Mandryk J Hendrie L Hely MCG The role of physical and psychological factors in occupational low back pain: a prospective cohort study Occup Environ Med 2000 57 116 120 10711279 10.1136/oem.57.2.116
Maul I Läubli T Klipstein A Krueger H Course of low back pain among nurses: a longitudinal study across eight years Occup Environ Med 2003 60 497 503 12819283 10.1136/oem.60.7.497
Waddell G Law M The clinical course of low back pain In The back pain revolution 1998 Edingburgh: Churchill Livingstone 103 117
Smedley J Egger P Cooper C Coggon D Prospective cohort study of predictors of incident low back pain in nurses BMJ 1997 314 1225 1228 9154024
Turner JA Franklin G Turk DC Predictors of chronic disability in injured workers: a systematic literature synthesis Am J Ind Med 2000 38 707 722 11071692 10.1002/1097-0274(200012)38:6<707::AID-AJIM10>3.0.CO;2-9
Hofman F Stössel U Michaelis M Nübling M Siegel A Low back pain and lumbago-sciatica in nurses and a reference group of clerks: results of a comparative prevalence study in Germany Int Arch Occup Environ Health 2002 75 484 490 12172895 10.1007/s00420-002-0332-6
The Swedish Council on Technology Assessment in Health Care Prescribed sick leave – causes, consequences and practices. A systematic review SBU 2003 2003 Gothenburg
Natvig B Eriksen W Bruusgaard D Low back pain as predictor of long-term work disability Scand J Public Health 2002 30 288 292 12680505 10.1080/14034940210133951
Eriksen W Bruusgaard D Knardahl S Work factors as predictors of sickness absence: a three month prospective study of nurses' aides Occup Environ Med 2003 60 271 278 12660375 10.1136/oem.60.4.271
Krause N Dasinger LK Deegan LJ Rudolph L Brand RJ Psychosocial job factors and return-to-work after compensated low back injury: a disability phase-specific analysis Am J Ind Med 2001 40 374 92 11598987 10.1002/ajim.1112
Fransen M Woodward M Norton R Coggan C Dawe M Sheridan N Risk factors associated with the transition from acute to chronic occupational back pain Spine 2002 27 92 98 11805644 10.1097/00007632-200201010-00022
Bourbonnais R Mondor M Job strain and sickness absence among nurses in the province of Quebec Am J Ind Med 2001 39 194 202 11170161 10.1002/1097-0274(200102)39:2<194::AID-AJIM1006>3.0.CO;2-K
Hoogendoorn WE van Poppel MN Bongers PM Koes BW Bouter LM Systematic review of psychosocial factors at work and private life as risk factors for back pain Spine 2000 25 2114 2125 10954644 10.1097/00007632-200008150-00017
Lancourt J Kettelhut M Predicting return to work for lower back pain patients receiving workers' compensation Spine 1992 17 629 640 1385657
Mortimer M Ahlberg G and the MUSIC-Norrtälje study group To seek or not to seek? Care-seeking behaviour among people with low-back pain Scand J Publ Health 2003 31 194 203 10.1080/14034940210134086
Hildebrandt VH Bongers PM Dul J van Dijk FJH Kemper HCG The relationship between leisure time, physical activities and musculoskeletal symptoms and disability in worker population Int Arch Environ Health 2000 73 507 518 10.1007/s004200000167
Eriksen W Bruusgaard D Physical leisure-time activities and long-term sick leave: a 15-month prospective study of nurses' aides J Occup Environ Med 2002 44 530 538 12085479
Kjellberg K Work technique in lifting and patient transfer tasks PhD thesis 2003 The Sahlgrenska Academy at Göteborg University, Department of Occupational Medicine, National Institute for Working Life, Department for Work and Health
Marras WS Davis KG Kirking BC Bertsche PK A comprehensive analysis of low-back disorder risk and spinal loading during the transferring and repositioning of patients using different techniques Ergonomics 1999 42 904 926 10424181 10.1080/001401399185207
Josephson M Gustafsson H Ahlberg-Hultén G Härenstam A Theorell T Wiktorin C Vingård E and MUSIC_Norrtälje study group Paid and unpaid work, and its relation to care seeking for low back and neck/shoulder disorders among women Women Health 2003 37 17 30 12733551 10.1300/J013v37n02_02
Kuorinka I Jonsson B Kilbom Å Vinterberg H Biering-Sørensen F Andersson G Jørgensen K Standardised Nordic Questionnaire for the analyses of musculoskeletal symptoms Appl Ergonomics 1987 18 233 237 10.1016/0003-6870(87)90010-X
Zigmond AS Snaith RP The Hospital Anxiety and depression scale Acta Psychiatr Scand 1983 67 361 370 6880820
Evans O Steptoe A The contribution of gender-role orientation, work factors and home stressors of psychological well-being and sickness absence in male-and female-dominated occupational groups Soc Sci Med 2002 54 481 492 11848269 10.1016/S0277-9536(01)00044-2
Lundqvist C Sjösteen A Blomstrand C Lind B Sullivan M Spinal cord injuries. Clinical, functional and emotional status Spine 1991 16 78 83 2003241
Wiktorin C Wigaeus Hjelm E Winkel J Köster M and Stockholm MUSIC I study group Reproducibility of a questionnaire for assessment of physical load during work and leisure time J Occup Med 1996 38 190 201 10.1097/00043764-199602000-00017
Borg G Perceived exertion as an indicator of somatic stress Scand J Rehab Med 1970 2 92 98
Johansson JÅ Psychosocial work factors, physical work load and associated musculoskeletal symptoms among home care workers Scand J Psychol 1995 36 113 129 7644896
Johansson JÅ The impact of decision latitude, psychological load and social support at work on musculoskeletal symptoms Eur J Public Heath 1995 5 169 174
Karasek R Theorell T Healthy work. Stress, productivity, and the reconstruction of working life 1990 New York: Basic Books, Inc
Berglund G Elmstadt S Janzon L Larsson SA The Malmo Diet and Cancer Study. Design and feasibility J Intern Med 1993 233 45 51 8429286
Sullivan M Karlsson J Bengtsson C Furunes B Lapidus L Lissner L The Göteborg Quality of Life Instrument" – A psychometric evaluation of assessments of symptoms and well-being among women in general population Scand J Prim Care 1993 11 267 275
Horneij E Hemborg B Jensen I Ekdahl C No significant differences between intervention programmes on neck, shoulder and low back pain: a prospective randomized study among home-care personnel J Rehabil Med 2001 33 170 176 11506215 10.1080/165019701750300636
Florvall Müller C Monrad T Biering Sørensen F Darre E Deis A Kryger P The influence of previous low back trouble, general health, and working conditions on future sick-listing because of low back trouble. A 15-year follow-up study of risk indicators for self-reported sick-listing caused by low back trouble Spine 1999 24 1562 1570 10457576 10.1097/00007632-199908010-00010
Marras WS Davis KG Heaney CA Maronitis AB Allread WG The influence of psychological stress, gender, and personality on mechanical loading of the lumbar spine Spine 2000 25 3045 3054 11145816 10.1097/00007632-200012010-00012
Torgen M Kilbom Å Physical work load between 1970 and 1993 – did it change? Scand J Work Environ Health 2000 26 161 168 10817382
van Poppel MN de Wet HC Koes BW Smid T Bouter LM Measuring sick leave: a comparison of self-reported data on sick leave and data from company records Occup Med 2002 52 485 490 10.1093/occmed/52.8.485
Lagerström M Hagberg M Evaluation of a 3 year education and training program for nursing personnel at a Swedish hospital AAOHN J 1997 45 83 92 9146108
| 15533255 | PMC539270 | CC BY | 2021-01-04 16:03:42 | no | BMC Musculoskelet Disord. 2004 Nov 8; 5:38 | utf-8 | BMC Musculoskelet Disord | 2,004 | 10.1186/1471-2474-5-38 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-4-871556657210.1186/1471-2407-4-87Research ArticleAntitumor effectiveness of different amounts of electrical charge in Ehrlich and fibrosarcoma Sa-37 tumors Ciria HC [email protected] MS [email protected] LB [email protected]ón RP [email protected] MF [email protected] OG [email protected]ález TR [email protected]ópez DS [email protected] JM [email protected] Sección de Bioelectricidad. Departamento de Bioingeniería y Equipos, Centro Nacional de Electromagnetismo Aplicado, Universidad de Oriente, Santiago de Cuba 90400, Cuba2 Hospital Oncológico Docente Provincial Conrado Benítez, Santiago de Cuba 90100, Cuba3 Departamento de Inmunología, Hospital Provincial Clínico Quirúrgico Docente Saturnino Lora, Santiago de Cuba 90500, Cuba4 Hospital Infantil Norte Docente "Juan Martínez de la Cruz Maceira". Santiago de Cuba, Cuba5 Dirección Municipal de Salud Pública. Santiago de Cuba, Cuba6 Departamento de Investigación en Física, Universidad de Sonora, Apdo. Postal A-088, 83190 Hermosillo, Sonora, México2004 26 11 2004 4 87 87 15 7 2004 26 11 2004 Copyright © 2004 Ciria et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In vivo studies were conducted to quantify the effectiveness of low-level direct electric current for different amounts of electrical charge and the survival rate in fibrosarcoma Sa-37 and Ehrlich tumors, also the effect of direct electric in Ehrlich tumor was evaluate through the measurements of tumor volume and the peritumoral and tumoral findings.
Methods
BALB/c male mice, 7–8 week old and 20–22 g weight were used. Ehrlich and fibrosarcoma Sa-37 cell lines, growing in BALB/c mice. Solid and subcutaneous Ehrlich and fibrosarcoma Sa-37 tumors, located dorsolaterally in animals, were initiated by the inoculation of 5 × 106 and 1 × 105 viable tumor cells, respectively. For each type of tumor four groups (one control group and three treated groups) consisting of 10 mice randomly divided were formed. When the tumors reached approximately 0.5 cm3, four platinum electrodes were inserted into their bases. The electric charge delivered to the tumors was varied in the range of 5.5 to 110 C/cm3 for a constant time of 45 minutes. An additional experiment was performed in BALB/c male mice bearing Ehrlich tumor to examine from a histolological point of view the effects of direct electric current. A control group and a treated group with 77 C/cm3 (27.0 C in 0.35 cm3) and 10 mA for 45 min were formed. In this experiment when the tumor volumes reached 0.35 cm3, two anodes and two cathodes were inserted into the base perpendicular to the tumor long axis.
Results
Significant tumor growth delay and survival rate were achieved after electrotherapy and both were dependent on direct electric current intensity, being more marked in fibrosarcoma Sa-37 tumor. Complete regressions for fibrosarcoma Sa-37 and Ehrlich tumors were observed for electrical charges of 80 and 92 C/cm3, respectively.
Histopathological and peritumoral findings in Ehrlich tumor revealed in the treated group marked tumor necrosis, vascular congestion, peritumoral neutrophil infiltration, an acute inflammatory response, and a moderate peritumoral monocyte infiltration. The morphologic pattern of necrotic cell mass after direct electric current treatment is the coagulative necrosis. These findings were not observed in any of the untreated tumors.
Conclusion
The data presented indicate that electrotherapy with low-level DEC is feasible and effective in the treatment of the Ehrlich and fibrosarcoma Sa-37 tumors. Our results demonstrate that the sensitivity of these tumors to direct electric current and survival rates of the mice depended on both the amount of electrical charge and the type of tumor. Also the complete regression of each type of tumor is obtained for a threshold amount of electrical charge.
==== Body
Background
The use of electric current in the treatment of malignant tumors has been known since the beginning of the 19th century. Several investigators have reported encouraging results from experimental low-level direct current therapy (DEC) in different types of tumor [1-3]. These studies have shown that DEC has an antitumor effect in different animal tumor models and in clinic; however, it has not yet been universally accepted.
The dose-response relationships obtained in these studies indicate that the DEC effectiveness depends on both the type of tumor and therapeutic scheme (amount of electrical charge and electrode array). Lack of guidance has become an obstacle to introduce the electrochemical treatment (EChT) in the clinic oncology. This is due to the lack of standardization of the EChT method regarding DEC doses and electrode array. Ren et al. [4] studied the influence of the dose and electrode spacing in the breast cancer and concluded that an increase of the dose lead to an increase in both the necrosis percentage and increased survival rate. However, they did not find significant spacing effect on the tumor necrosis percentage. On the other hand, Chou et al. [5] revealed that the number of electrodes depends on the tumor size and that the electrodes inserted at the base perpendicular to the tumor long axis increased the antitumor effectiveness respect to other electrode configurations used.
In spite of these results, the efficacy of DEC treatment has been controversial since an optimum electrode array and a threshold amount of electrical charge for each type of tumor have not been established. We believe that the procedure to determine the amount of electrical charge for each type of tumor is completely destroyed is more feasible to implement than that for the optimum electrode array, which involves several variables, such as polarity, number, and orientation of the electrodes. The knowledge of the optimum values of these parameters may lead to maximize the antitumor effectiveness of DEC and minimize their adverse effects in the organism. This allows the establishment of a therapeutic procedure for the tumor treatment in animals and in clinical oncology.
The aim of this study is to test the hypothesis that the responses of the tumors treated with DEC is dependent on dose. Ehrlich and fibrosarcoma Sa-37 tumors were used. The survival rates of the mice bearing of these two types of tumor were determined. The antitumor effects of DEC were also evaluated through the peritumoral and tumoral findings in Ehrlich tumor.
Methods
Animals
The experiment was run in accordance with Good Laboratory Practice rules and animals protection laws. The experiment was approved by the ethical committee of Oriente University, which follows the guideline from the Cuban Animal Ethical Committee. BALB/c male mice, 7–8 week old and 20–22 g weight were used. They were supplied from the National Center for Production of Laboratory Animals (CENPALAB), Havana City, Cuba, and were kept in standard laboratory conditions with water and food ad libitum. Animals were healthy (without signs of fungal or other infections) and were maintained in plastic cages inside a room at a constant temperature of 23 ± 2°C and relative humidity of 65 %, and a natural day-night cycle. During therapy the animals were firmly fixed on wooden boards, so all treatments were performed in the absence of anesthesia. All treated animals showed uneasy and quick breathing during fixation.
Tumor cell lines
Ehrlich and fibrosarcoma Sa-37 cell lines, growing in BALB/c mice, were received from the Center for Molecular Immunology, Havana City, Cuba. Both cell lines are being maintained in the Cell Culture Collection of the Department of Pathologic Anatomy, Hospital "Conrado Benítez", Santiago de Cuba, Cuba.
The Ehrlich and fibrosarcoma Sa-37 ascitic tumor cell suspensions, transplanted to the BALB/c mouse, were prepared from the ascitic forms of the tumors. Ehrlich solid and subcutaneous tumors, located dorsolaterally in animals, were initiated by the inoculation of 5 × 106 viable tumor cells in 0.2 ml of 0.9 % NaCl, while fibrosarcoma Sa-37 solid and subcutaneous tumors located dorsolaterally in animals, were initiated by the inoculation of 1 × 105 viable tumor cells in 0.2 ml of 0.9 % NaCl. For both tumors, the viability of the cells was determined by Trypan blue dye exclusion test and it was over 95 %. Cell count was made using an hematocytometer.
Tumor growth was followed by measuring three perpendicular tumor diameters (a, b and c, where a > b > c) with a vernier caliper. The tumor volume was estimated using the equation . The mean tumor volume with the corresponding standard deviation of three determinations was calculated in each experimental group. Mice with non-palpable tumor at day 60 after the treatment were designated as cured.
Tumor doubling time (DT, in days) was determined for each individual tumor as the time needed to double the initial tumor volume. For each experimental group the mean DT and its standard deviation were calculated.
Histopathological study of the Ehrlich tumor
The histologic cuts from each tumor were made according to the largest diameter. They were fixed in a 10 % formol solution and processed by the paraffin method.
Hematoxylin and eosin stained slides were used to evaluate the presence of necrosis. Hematoxylin and eosin stained slides were examined under an Olympus light microscope. The extent of necrosis was defined as the percentage of necrotic region compared with the whole area of the tumor section.
The peritumoral alterations were evaluated as none (-), slight (+), moderate (++) and severe (+++).
Electrochemical treatment
To supply electrochemical treatment, a high stability and low noise DEC source was built at the National Center for Applied Electromagnetism (CNEA). The electrode configuration consisted of a multi-electrode array formed by two anodes and two cathodes inserted into the base perpendicular to the tumor long axis keeping about 3 mm distance between them. Cathode and anode were connected in alternate sequence. This multi-electrodes array was proposed taking into account the results reported by Chou et al. [5]. All electrodes were cleaned and sterilized in alcohol prior to use. Platinum electrodes of 0.7 mm diameter and 20 mm long were used. After the electrodes were inserted, they were connected to the DEC source.
In order to find the thresholds of the electrical charge for which Ehrlich and fibrosarcoma Sa-37 tumors are completely destroyed, different amounts of electrical charge in the range of 5.5 to 110 C/cm3 were used. From this range of electrical charge three values were chosen to show the DEC effectiveness in both types of tumors. When the tumors reached approximately 0.5 cm3 in BALB/c mice, a single shot electrotherapy was supplied (zero day). For each type of tumor four groups consisting of 10 mice each randomly divided were formed. For Ehrlich tumor the groups formed were: control group (CG1), treated group with electrical charge of 36 C/cm3 (18.0 C in 0.5 cm3) and 6.7 mA for 45 min (TG1-1), treated group with 63 C/cm3 (31.5 C in 0.5 cm3) and 11.7 mA for 45 min (TG1-2), and treated group with 92 C/cm3 (46.0 C in 0.5 cm3) and 17 mA for 45 min (TG1-3). For fibrosarcoma Sa-37 tumor the groups formed were: control group (CG2), treated group with 36 C/cm3 (18 C in 0.5 cm3) and 6.7 mA for 45 min (TG2-1), treated group with 63 C/cm3 (31.5 C in 0.5 cm3) and 11.7 mA for 45 min (TG2-2), and treated group with 80 C/cm3 (40.0 C in 0.5 cm3) and 14.8 mA for 45 min (TG2-3).
The dose of 105 C/cm3 (52.5 C in 0.5 cm3) and 19.4 mA for 45 min was supplied to 10 mice (5 mice bearing Ehrlich tumor and 5 mice bearing fibrosarcoma Sa-37 tumor). Also the dose of 110 C/cm3 (55 C in 0.5 cm3) and 20.3 mA for 45 min was supplied to 10 mice (5 mice bearing Ehrlich tumor and 5 mice bearing fibrosarcoma Sa-37 tumor). These doses were used to evaluate the therapeutic and adverse effects of the DEC above 100 C/cm3. For each type of tumor was formed a control group of 10 mice.
In order to examine from the histolological point of view the effects of direct electric current in Ehrlich tumor two experimental groups were formed: a control group (CG-A) and a treated group with 77 C/cm3 (27.0 C in 0.35 cm3) and 10 mA for 45 min (TG-A). This treated group was divided in three subgroups TG1-A, TG2-A and TG3-A to show the tumor and peritumoral findings at 1, 2 and 4 days after DEC treatment. Each experimental group was formed by 6 mice. When the Ehrlich tumor volumes reached 0.35 cm3, two anodes and two cathodes were inserted into the base perpendicular to the tumor long axis and a single shot electrotherapy was supplied (zero day).
In all experiments, before treatment the DEC was increased gradually step by step for two minutes until the desired intensity. During treatment it was constant and continually monitored. The voltage was also continually monitored. It varied, in accordance with the change of tissue resistance during the current application, between 5 and 25 V. The total electrical charge was calculated in real time. After a single application of the intended dose, the treatment was stopped. In this case, the current was decreased step by step for two minutes until its intensity was 0 mA. During electrotherapy, mice were firmly restrained, without obvious discomfort; therefore no anesthesia was necessary.
In the control groups, four electrodes were placed into the base perpendicular to the tumor long axis without applying any direct current (0 mA). The animals of this group were firmly fixed but without DEC and showed uneasy and quick breathing during their fixation.
Survival rates of the mice bearing both types of the tumors were determined for each experimental group. The survival rate (in %) was defined as the ratio between the number of live animals and the total number of animals, multiplied by 100 %. Survival checks mortality were made daily.
Histopathological study of the tumor
The histologic cuts from each tumor were made according to the largest diameter. They were fixed in a 10 % formal solution and processed by the paraffin method.
Hematoxylin and eosin staining was used. Each cut was divided into four microscopic fields in order to calculate the necrosis percentage through panoramic lens. This percentage was calculated as the ratio between the necrosis area and the tumor total area, multiplied by 100 %.
Statistical criteria
The nonparametric statistical criterion of one-tailed Wilcoxon-Mann-Whitney rank sum was used to compare volumes between the treated groups with DEC and their respective control groups. Survival curves for the three different mice treatment groups for each tumor type were estimated by using the Kaplan-Meier product limit estimator [6].
McNemar's statistical criterion was used for comparing the main histopathological findings in peritumoral zones in animals from CG-A and TG-A. P values of less than 0.05 were considered significant. The mean value and its mean standard error were reported for each experimental group.
Results
As it is shown in Table 1 and Figure 1, Ehrlich tumors in DEC-treated mice were significantly inhibited as compared with tumors of untreated mice (P < 0.02). This tumor growth inhibition following DEC treatment was observed in every individual mouse. Also there are significant differences between the treated groups being more evident for TG1-3 (P < 0.05). Similar effect of DEC treatment was observed in fibrosarcoma Sa-37 bearing mice (Table 1 and Fig. 2). In these mice DEC treatment also resulted in significant inhibition of tumor growth (P < 0.02). For this type of tumor also were observed significant differences between the treated groups being more evident for TG2-3 (P < 0.05).
The results shown in this study revealed that the sensitivity of the Ehrlich and fibrosarcoma Sa-37 tumors was dose dependent. The sensitivity to DEC of both types of tumors increased with the increase of the amount of electrical charge (Table 1 and Figs. 1 and 2). These results also made evident that fibrosarcoma Sa-37 tumor were more sensitive to DEC than Ehrlich tumor under the same amount of electrical charge (TG1-1 compared with TG2-1 and TG1-2 compared with TG2-2). For these doses there were significant differences (P < 0.05). It was also observed on Ehrlich tumor for doses of 36 and 63 C/cm3 that the tumors partially regressed for 2 and 4 days, respectively. For these same doses the fibrosarcoma Sa-37 tumor reached their respective partial regressions for 4 and 5 days. Both eventually outgrew again.
The complete regression of the Ehrlich tumor was observed 25 days after treatment with 92 C/cm3 (Table 1 and Fig. 1); however, for the fibrosarcoma Sa-37 tumor it was observed 15 days post-treatment with 80 C/cm3 (Table 1 and Fig. 2). After 60 days post-treatment the tumors were non palpable in TG1-3 and TG2-3. For these doses there were no significant differences (P > 0.05) in the growth of these two types of tumor after treatment; however, there were significant differences in the time for which each type of tumor was completely destroyed (P < 0.05).
In the case of the untreated tumors, fibrosarcoma Sa-37 tumor showed a quicker growth than that of the Ehrlich tumor. Also, the DT of fibrosarcoma Sa-37 was 0.7 times smaller than that of the Ehrlich tumor (Table 1).
The overall survival curves of the mice bearing Ehrlich and fibrosarcoma Sa-37 tumors are shown in figures 3 and 4, respectively. These figures show that for both types of tumors the survival rate of the mice treated with DEC was significantly greater when compared with that of their respective untreated mice (P < 0.001). In this figure it was also observed that the cure rates were 80 % (8/10) for Ehrlich tumor (TG1-3) and 90 % (9/10) for fibrosarcoma Sa-37 tumor (TG2-3). Significant differences between the survival rates of the mice treated with different amounts of electrical charge (P < 0.05) were also found, being more marked for TG1-3 and TG2-3 for Ehrlich and fibrosarcoma Sa-37 tumors, respectively. For the dose of 36 C/cm3 there were no significant differences between both types of tumor (P > 0.05); however, for the other doses there were significant differences (P < 0.05).
The cured mice were sacrificed at 100 days post-treatment. Before sacrifice, the animals were active and in good physical condition with adequate body weight. They had good posture and coats of hair. After sacrifice, the histopathological findings in each of these mice showed complete disappearance of the tumor and evidence of healing. In the treated mice a very little necrotic tissue remained within a fibrous scar. Serology and histological finding of the organs did reveal neither abnormalities nor metastases (results not shown).
The death of a mouse 1-day after DEC treatment was observed in TG1-3. The histological findings revealed damages in the lungs due to hemorrhage and a small circular necrosis. Metastases were not observed in this mouse. It was also observed the death of a mouse 25 days post-treatment in TG1-3 and 50 days in TG2-3 due to the cannibalism shown by the mice, probably because of the blood present in the tumors after DEC treatment. All the mice died for amounts of electrical charge above 100 C/cm3, during the first 24 hours after DEC treatment. The histological findings showed both severe alterations in liver and kidney and an increase in the weight of these organs. Metastases were not observed in any of these mice.
The histopathological findings revealed that in the Ehrlich untreated tumors (CG-A) the necrotic area was mainly central and it constituted approximately from 20 % of the tumor total area (Fig. 5). However, in tumors treated with DEC, a wide necrotic area was observed. The tumor necrosis percentages of treated groups at 1, 2 and 4 days after treatment were approximately 2.7, 3.9 (Fig. 6) and 4.7 (Fig. 7) times higher than that of the CG-A, respectively. These differences were significant (P < 0.02). Also there were significant differences between the necrosis percentages of treated tumors at 1, 2 and 4 days (P < 0.02).
There was a lack of well defined necrosis zones surrounding the electrodes. The morphologic pattern of the necrotic cell mass observed is the coagulative necrosis. The dead tissue becomes both swollen and firm in consistency. Preservation of the basic profile of the coagulated cancerous cell and nuclear karyolysis were also observed. The lysed erythrocytes was also observed. This type of necrosis was accompanied by accumulation of neutrophil polymorphonuclear leucocytes.
Lymphocytes (L) and plasmatic cells, named CP, were observed in all the tumors in CG-A and TG-A but there were no significant differences (Table 2, Fig. 8). Neutrophil infiltration (N) and vascular congestion, named CV, were observed in all animals from the TG-A (Figs. 9 and 10). The intensity grades of these peritumoral findings were severe; however, the intensity grade of the monocyte infiltration (M) was slight to moderate in this TG-A. Edema and acute inflammatory response were observed 1, 2 and 4 days after treatment (Figs. 9 and 10). These peritumoral findings were not present in any of the animals from the CG-A (Table 2). There were significant differences (P = 0.008) between the peritumoral findings of the CG-A and TG-A.
In this experiment no mouse died from intercurrent disease during or after the treatment. Before sacrifice, the animals were active and in good physical condition with adequate body weight. They had good posture and coats of hair.
Discussion
The results of this study demonstrated that DEC has a marked antitumor effect because a single-shot electrotherapy delivered via four platinum electrodes inserted into the base of the fibrosarcoma Sa-37 and Ehrlich murine tumors significantly retarded their growths when compared with their respective control groups. The fact that tumor regression increases with the increase of the amount of electrical charge may be explained because the induced necrosis by DEC into the tumor depends directly on its intensity, a matter that is in agreement with the results of Robertson et al. [7]. In an additional experiment was corroborated that the decrease of each treated tumor volume is due to the higher necrosis percentage induced into the tumor by DEC action. The histopathological findings made to mice 100 days post-treatment may suggest that an increase of the dose bring about an increase of the percentage of the tumor necrosis and the necrotic overlap. Also these findings confirm that the results of the pathology study were consistent with the survival study.
We believe that the necrosis is the predominant mechanism of cell death, by the cellular tumefaction (or cellular swelling), cell rupture, breakdown of organelles and acute inflammatory response observed during the first 4 days post-treatment in all treated tumors, result that agrees with that previously reported by Dodd et al. [8] and Holandino et al. [9]. Von Euler et al. [10] demonstrated that the appearance of the necrosis depends on the polarity of the electrode. The findings of necrosis observed by these researchers around anode and cathode electrodes were also observed in all treated tumors (coagulative necrosis, extravasation of blood cells, nuclear karyolysis and edema), fact that was explained because both electrodes were inserted into the tumors. On the other hand, Von Euler [11] observed both apoptosis and necrosis around the anode but only necrosis around cathode.
The necrosis may be due to the ischemia observed in all tumors treated with DEC, which could lead to an irreversible cell injury of the tumor cells and therefore to cellular death. This fact could be related with other experimental findings found after DEC treatment, such as: degradation of phospholipids, lost of high energy phosphate and increase of the intracellular calcium [7], membrane damage [5], ionic imbalance [2,12], mitochondrial alterations [9] and ischemia/reperfusion injury [13].
The prolonged acute inflammation observed during 4 days after DEC treatment may be explained by the persistent leukocyte infiltrate also observed in the peritumoral findings. This persistent leukocyte infiltrate (essential feature of the inflammatory response) becomes a harmful agent because during the chemotaxis they amplify the effects of the initial inflammatory stimulus through the liberation of potent mediators (enzymes, chemical mediators and toxic radical of oxygen) that lead to both endothelial and tissue damages. This leukocyte infiltrate may also activate the immune system [14]. In all these processes the reactive oxygen species have been shown to have an important role. In addition to these species are essential elements in the emergence of an inflammatory process [14,15]. Therefore we speculate that the oxidative burst may be the immediate cause of cell death in both tumors, although not investigated in this study.
These facts and the high necrosis percentages shown in this study may lead to the complete destruction of the solid tumor treated with DEC. The complete disappearance of the Ehrlich and fibrosarcoma Sa-37 tumors achieved for 92 and 80 C/cm3, respectively, may suggest that each tumor model has its threshold of electric charge from which it is completely destroyed. This threshold depends on the electric nature of the tumor and their physiological characteristics (stage, volume and histogenic characteristics). This fact explains the cure of the mice and why the tumors do not duplicate their initial volumes during the observation time (infinite DT, represented in Table 1 by ∞ symbol).
The experimental data revealed that the fibrosarcoma Sa-37 showed the higher sensitivity and curability to DEC than Ehrlich tumor and that both tumor response and survival rate of mice were DEC dependent. However, in the untreated tumors Fibrosarcoma Sa-37 showed a DT shorter than that of the Ehrlich tumor. This fact indicates the higher agressiveness of Fibrosarcoma Sa-37.
The mortality observed in all animals treated with amounts of electrical charge above 100 C/cm3 could be explained by the severe damages induced by DEC in kidney and liver. Griffin et al. [12] explained this result by the induced serum electrolyte imbalance resulting from a metabolic load due to the breakdown products of the tumors.
The hemorrhage observed in the lungs of the mouse death 1 day after DEC treatment in TG1-3 may be explained by the vascular rupture and/or perforation of blood vessels due to a mechanic effect by the insertion of an electrode. The small circular necrosis also observed in this organ's mouse may be consequence of the cytotoxic action of DEC.
The uneasy and quick breathing observed in both control and treated groups, during the fixation of the mice did not have any influence in the results obtained in this study.
Conclusions
The data presented indicate that electrotherapy with low-level DEC is feasible and effective in the treatment of the Ehrlich and fibrosarcoma Sa-37 tumors. Our results demonstrate that the sensitivity of these tumors to direct electric current and survival rates of the mice depended on both the amount of electrical charge and the type of tumor. Also the complete regression of each type of tumor is obtained for a threshold amount of electrical charge.
Competing interests
The author(s) declare that they have no competing interests
Authors' contributions
HCC conceived the study, and participated in its design and coordination. Also, he carried out the inoculation of the tumor cells in the mice, the measure of the tumor volumes and the survival rate of mice as well as elaborated the manuscript. MCSQ participated in the design of the study and participated in the measure of the histological findings of the organs and tumor and peritumoral findings and as well as elaborated the manuscript. LEBC carried out the inoculation of the tumor cells in the mice, conceived and participated in the design of the study and performed the statistical analysis as well as elaborated the manuscript. RNPB and DSL participated in the design of the study and contributed to elaboration of this manuscript. MFS participated in the design of the study and carried out the serology. All authors read and approved the final manuscript. OGP and TRG participated in the design of the study and contributed to elaboration of this manuscript. All authors read and approved the final manuscript. JLMF participated in the design of the study and performed the statistical analysis.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to thank Yarindra Mesa Mariño, Kenia Caballero Bordeloy, and Emilio Suárez for their technical assistance.
This research was supported by the Ministry of Science and Technology of Santiago de Cuba and Ministry of Superior Education, Republic of Cuba.
Figures and Tables
Figure 1 Effect of DEC on the growth curve of Ehrlich tumor. Data are means ± mean standard error (vertical bars). The experimental groups formed for the Ehrlich tumor were control group, CG1 (-■-); treated group with 36 C/cm3, TG1-1 (-●-); treated group with 63 C/cm3, TG1-2 (-▲-); treated group with 92 C/cm3, TG1-3 (-▼-). Each experimental group is formed by 10 mice.
Figure 2 Effect of DEC on the growth curve of fibrosarcoma Sa-37 tumor. Data are means ± mean standard error (vertical bars). The experimental groups formed for the fibrosarcoma Sa-37 tumor were control group, CG2 (-■-); treated group with 36 C/cm3, TG2-1 (-●-); treated group with 63 C/cm3, TG2-2 (-▲-); treated group with 80 C/cm3, TG2-3 (-▼-). Each experimental group is formed by 10 mice.
Figure 3 Survival rates in BALB/c mice bearing Ehrlich tumor after electrochemical treatment. Data are means ± mean standard error (vertical bars). The experimental groups formed for the Ehrlich tumor were control group, CG1 (-■-); treated group with 36 C/cm3, TG1-1 (-●-); treated group with 63 C/cm3, TG1-2 (-▲-); treated group with 92 C/cm3, TG1-3 (-▼-). Each experimental group is formed by 10 mice.
Figure 4 Survival rates in BALB/c mice bearing fibrosarcoma Sa-37 tumor after electrochemical treatment. Data are means ± mean standard error (vertical bars). The experimental groups formed for the fibrosarcoma Sa-37 tumor were control group, CG2 (-■-); treated group with 36 C/cm3, TG2-1 (-●-); treated group with 63 C/cm3, TG2-2 (-▲-); treated group with 80 C/cm3, TG2-3 (-▼-). Each experimental group is formed by 10 mice.
Figure 5 Central necrosis area in an Ehrlich untreated tumor (+). HE. × 32.
Figure 6 Necrosis area in the Ehrlich treated tumor (+) 2 days after DEC treatment. HE. × 32.
Figure 7 Necrosis area in the Ehrlich treated tumor (+) 4 days after DEC treatment. HE. × 32.
Figure 8 Lymphoplasmocytic infiltrate in both untreated and DEC treated tumors: plasmatic cells (CP) and lymphocytes (L). HE. × 400.
Figure 9 Peritumoral findings of treated tumors at 1 day after DEC treatment: leucocytes neutrophil infiltration (N), monocytes (M) and lymphocytes (L). HE. × 400.
Figure 10 Pattern of acute inflammatory response observed during 1, 2 and 4 days after treatment: vascular congestion (CV) and neutrophils (N). HE. × 100.
Table 1 Mean doubling time
Ehrlich Tumor Fibrosarcoma Sa-37 tumor
CG1 TG1-1 TG1-2 TG1-3 CG2 TG2-1 TG2-2 TG2-3
DT1 2.4 ± 0.3 6.8 ± 0.7 16.9 ± 2.4 ∞3 1.6 ± 0.2 11.2 ± 1.3 23.6 ± 3.8 ∞3
- 2.9 7.1 ∞3 - 7.0 14.9 ∞3
1 DT (in days) is the double time of the tumors. Data are means ± standard deviation of tumors.
2 DTTG/CG is a variable that characterizes the increase of DT in each treated group. (DTTG) in respect to its control group (DTCG) for both types of tumors.
3 The symbol ∞ means infinite tumor doubling time (see Discussion).
Table 2 Peritumoral pathological findings The number of the mice in each experimental group is specified by n. CGA is the control group and TG1-A, TG2-A and TG3-A are the experimental subgroups of the group treated with 77 C/cm3at 1, 2 and 4 days after DEC treatment, respectively. Mc Nemar Test shows statistically significant differences in peritumoral findings at 1, 2 and 4 days.
Alterations found Experimental Groups Number of mice (% of the total) in different degrees of alterationb
- + ++ +++
Lymphoplasmocytic infiltrate: Lymphocytes (L) and plasmatic cells (CP) CGA (n = 6) 0 (0.0) 0 (0.0) 6 (100.0) 0 (0.0)
TGA-1 day (n = 6) 0 (0.0) 0 (0.0) 6 (100.0) 0 (0.0)
TGA-2 days (n = 6) 0 (0.0) 0 (0.0) 6 (100.0) 0 (0.0)
TGA-4 days (n = 6) 0 (0.0) 0 (0.0) 6 (100.0) 0 (0.0)
Neutrophilic infiltrate (N) CGA (n = 6) 6 (100.0) 0 (0.0) 0 (0.0) 0 (0.0)
TGA-1 day(n = 6) 0 (0.0) 0 (0.0) 0 (0.0) 6 (100.0)a
TGA-2 days (n = 6) 0 (0.0) 0 (0.0) 0 (0.0) 6 (100.0)a
TGA-4 days (n = 6) 0 (0.0) 0 (0.0) 0 (0.0) 6 (100.0)a
Monocytic infiltrate (M) CGA (n = 6) 6 (100.0) 0 (0.0) 0 (0.0) 0 (0.0)
TGA-1 day (n = 6) 0 (0.0) 6 (100.0)a 0 (0.0) 0 (0.0)
TGA-2 days (n = 6) 0 (0.0) 0 (0.0) 6 (100.0)a 0 (0.0)
TGA-4 days (n = 6) 0 (0.0) 0 (0.0) 6 (100.0)a 0 (0.0)
Vascular congestion (CV) CGA (n = 6) 6 (100.0) 0 (0.0) 0 (0.0) 0 (0.0)
TGA-1 day (n = 6) 0 (0.0) 0 (0.0) 0 (0.0) 6 (100.0)a
TGA-2 days (n = 6) 0 (0.0) 0 (0.0) 0 (0.0) 6 (100.0)a
TGA-4 days (n = 6) 0 (0.0) 0 (0.0) 0 (0.0) 6 (100.0)a
a P = 0.008.
b Note: The signs " -, +, ++ and +++ ", represent: none, slight, moderate and severe intensity grades of alterations found, respectively.
==== Refs
Nordenström BE Biologically Closed Electric Circuits Clinical, Experimental and Theoretical Evidence for an Additional Circulatory System 1983 Nordic Medical Publications
Bergues LC Camué HC Pérez RB Suárez MCQ Hinojosa RA Montes De Oca LG Fariñas MS De la Guardia OP Electrochemical Treatment of Mouse Ehrlich Tumor with Direct Electric Current Bioelectromagnetics 2001 22 316 322 11424154 10.1002/bem.56
WemyssHolden SA Dennison AR Finch GJ Hall P Maddern GJ Electrolytic ablation as an adjunct to liver resection experimental studies of predictability and safety British Journal of Surgery 2002 89 579 585 11972547 10.1046/j.1365-2168.2002.02064.x
Ren RL Vora N Yang F Longmate J Wang W Sun H Li JR Weiss L Staud C McDougall JA Chou CK Variations of dose and electrode spacing for rat breast cancer electrochemical treatment Bioelectromagnetics 2001 22 205 211 11255217 10.1002/bem.40
Chou CK Mcdougall JA Ahn C Vora N Electrochemical treatment of mouse and rat Fibrosarcomas with direct current Bioelectromagnetics 1997 18 18 24
Kaplan EL Meier P Nonparametric estimation from incomplete observations J Am Stat Assn 1958 53 457 481
Robertson GS Wemyss-Holden SA Denniso AR Hall P Baxter P Maddern GJ Experimental study of electrolysis-induced hepatic necrosis Br J Surg 1998 85 1212 1216 9752862 10.1046/j.1365-2168.1998.00806.x
Dodd NJ Moore JV Taylor TV Zhoo S Preliminary evaluation of low level direct current therapy using magnetic resonance imaging and spectroscopy Phys Med 1993 4 2 8
Holandino C Veiga VF Capella MM Menezez S Alviano CS Damage induction by direct electric current in tumoural targets cells (P815) Ind J Exp Biol 2000 38 554 558
von Euler H Nilsson E Olsson JM Lagerstedt AS Electrochemical treatment EChT effects in rat mammary and liver tissue. In vivo optimizing of a dose-planning model for EChT of tumours Bioelectrochemistry 2001 54 117 124 11694391 10.1016/S1567-5394(01)00118-9
von Euler H Electrochemical Treatment of Tumours Department of small animals Clinical Science Uppsala, Swedish University of Agricultural Sciences, Doctor's thesis 2002 1 32 ISSN 1401-6257
Griffin DT Dodd NJ Moore JV Pullan BR Taylor TV The effects of low level direct current therapy on a preclinical mammary carcinoma: tumor regression and systemic biochemical sequelae Br J Cancer 1994 69 875 878 8180017
Jarm T Čemažar M Serša G Miklavčič D Blood Perfusion in a murine fibrosarcoma tumor model after direct current electrotheraphy: a study with 86Rb extraction technique Electr Magnetobiol 1998 17 271 280
Cotran RS Kumar V Collins T Patología Estructural y Funcional 1999 Sexta McGraw-Hill-Interamericana de España, S A U: Madrid 277 347
Ghisham MB Reactive Metabolites of oxygen and nitrogen in Biology and Medicine 1995 3 RG Landes Company Austin/Georgetown: New York
| 15566572 | PMC539271 | CC BY | 2021-01-04 16:03:02 | no | BMC Cancer. 2004 Nov 26; 4:87 | utf-8 | BMC Cancer | 2,004 | 10.1186/1471-2407-4-87 | oa_comm |
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-4-301558506110.1186/1472-6920-4-30Research ArticleCritical appraisal skills training for health care professionals: a randomized controlled trial [ISRCTN46272378] Taylor Rod S [email protected] Barnaby C [email protected] Paul E [email protected] Rebecca J [email protected] Department of Public Health & Epidemiology, University of Birmingham; Birmingham, UK2 Health Services Research Unit, School of Hygiene and Tropical Medicine; London, UK3 Somerset Research and Development Support Unit, Taunton & Somerset NHS Trust, Taunton; UK4 Health Economics Facility, University of Birmingham, Birmingham, UK2004 7 12 2004 4 30 30 16 8 2004 7 12 2004 Copyright © 2004 Taylor et al; licensee BioMed Central Ltd.2004Taylor 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
Critical appraisal skills are believed to play a central role in an evidence-based approach to health practice. The aim of this study was to evaluate the effectiveness and costs of a critical appraisal skills educational intervention aimed at health care professionals.
Methods
This prospective controlled trial randomized 145 self-selected general practitioners, hospital physicians, professions allied to medicine, and healthcare managers/administrators from the South West of England to a half-day critical appraisal skills training workshop (based on the model of problem-based small group learning) or waiting list control. The following outcomes were assessed at 6-months follow up: knowledge of the principles necessary for appraising evidence; attitudes towards the use of evidence about healthcare; evidence seeking behaviour; perceived confidence in appraising evidence; and ability to critically appraise a systematic review article.
Results
At follow up overall knowledge score [mean difference: 2.6 (95% CI: 0.6 to 4.6)] and ability to appraise the results of a systematic review [mean difference: 1.2 (95% CI: 0.01 to 2.4)] were higher in the critical skills training group compared to control. No statistical significant differences in overall attitude towards evidence, evidence seeking behaviour, perceived confidence, and other areas of critical appraisal skills ability (methodology or generalizability) were observed between groups. Taking into account the workshop provision costs and costs of participants time and expenses of participants, the average cost of providing the critical appraisal workshops was approximately £250 per person.
Conclusions
The findings of this study challenge the policy of funding 'one-off' educational interventions aimed at enhancing the evidence-based practice of health care professionals. Future evaluations of evidence-based practice interventions need to take in account this trial's negative findings and methodological difficulties.
==== Body
Introduction
For clinicians to make sense of scientific evidence and follow an evidence-based approach to their practice it has been stated they should be able to: (1) turn problems of their clinical practice into focused questions; (2) comprehensively search for literature to address these questions; (3) critically appraise this literature for its usefulness and scientific validity; and, (4) apply the results of this appraisal to their practice [1].
McColl and colleagues undertook one of the few studies of the prevalence of critical appraisal skills (CAS). In a sample of family practitioners, it was reported that only about one third claimed they "understood and could explain to others" terms which are intimately associated with an ability to critically appraise research [2].
A number of approaches have been developed to help clinicians enhance their CAS, including the publication of a number of critical appraisal checklists and the introduction of CAS teaching into undergraduate and postgraduate education in UK and abroad [3,4]. In UK and abroad, the Critical Appraisal Skills Programme (CASP) has become one of the most widely been disseminated forms of CAS training [5].
Four systematic reviews have been published that explore the effectiveness of CAS training [6-9]. These reviews observed marked heterogeneity in the nature of education intervention across individual studies, particularly in terms of duration (which varied across studies from 1 hour or less to 10 hours or more). However, these reviews consistently reported that CAS training results in small improvements in participants' knowledge of methodological and statistical issues in clinical research and enhances their attitudes towards the use of medical literature in clinical decision making. Nevertheless these findings need to be interpreted with considerable caution as most of the studies had poor internal validity. Only one randomized controlled trial was identified [10] and, in general, studies failed to blind outcome assessment. A focus on classroom-based interventions delivered to either medical students or medical residents, also limits the generalisability of the current evidence base. The aim of this study was to undertake a randomized controlled trial to assess the effectiveness and cost of CAS training in a range of practising healthcare professionals using a range of validated outcomes. Given its wide dissemination, the CASP model of CAS was evaluated in this trial.
Methods
Study design
The study was a prospective randomized controlled trial. Study outcomes were not assessed at baseline to avoid a pre-test effect. The possibility of a pre-assessment leading to a higher post assessment score due to an item-practice effect is well recognised in the educational evaluative literature [11]. However, trial participants' characteristics (i.e. gender, age, attitude towards the use of evidence about healthcare research, and details of previous training in research, epidemiology, or statistics) were collected by questionnaire prior to randomization and used as covariates to reduce variation from individual differences. Ethical approval for the study was obtained from all of the local district ethics committees from which the participants were drawn.
Selection of subjects & setting
Over a three-month period, 1,305 practitioners, working within the South and West Regional Health Authority in England, were sent an invitation to participate in one of a number of CAS workshops being run across the region. Invitations were sent to the health authority offices and all general practices in the geographical area. The letters of invitation included an explanation that agreement to take part in the workshops would include a formal evaluation. Applying to attend, which involved completion of a questionnaire with baseline questions, was taken as consent to enter the study. On receipt of a completed questionnaire, participants were randomized to either intervention or control. The intervention group were given a date to attend a CAS workshop and the control participants assigned to a waiting list to attend a workshop. The only exclusion criterion for entry into the study was attendance at a previous CAS workshop.
Sample size determination
The target sample size was 200, 100 in each group, which was chosen to allow the study to detect a 'moderate' effect size difference of 0.4 standard deviation units (in any outcome) at 80% power and a 5% significance level (2-tailed) [12].
Randomization and blinding
An independent researcher used computer generated codes to allocate applicants randomly to intervention (attend a critical appraisal workshop) or control group ('waiting list'), stratified by occupation: manager/administrator; medically qualified practising physician; nurse/profession allied to medicine and 'other' professions. The researchers who scored study outcomes were blinded to the allocation of participants at all times.
Intervention group
The teaching programme used in this study was based on the Critical Appraisal Skills Programme (CASP). The half-day workshop centres upon facilitating the process by which research evidence is systematically examined to assess study validity, the results and relevance to a particular clinical scenario. Participants practise these skills, during the workshop, by critically appraising a systematic review article and then receive follow up materials following the workshop (see Appendix 1 for details of intervention).
Development of outcomes
Given the absence of suitable validated outcomes measures, the outcomes were developed for use in trial. A questionnaire was developed and validated (reliability and internal consistency) to assess the following outcomes – knowledge of the principles necessary for appraising evidence; attitudes towards the use of evidence about healthcare; evidence seeking behaviour; perceived confidence in appraising evidence; and, knowledge of the principles necessary for appraising evidence; attitudes towards the use of evidence about healthcare; evidence seeking behaviour; perceived confidence in appraising evidence. A copy of the outcome questionnaire can be found in Appendix 2 (see Additional file 1). Full details of the validation process can be found elsewhere [13].
The questionnaire included 18 multiple-choice knowledge questions, 7 attitude statements and 6 confidence statements. Possible response categories to the knowledge questions were 'true', 'false' or 'don't know'. Correct, incorrect and don't know responses were awarded scores of 1, -1 and 0 respectively. Knowledge scores across question were summed giving a possible range of scores from -18 to +18. Attitude statements were scored on a five-point Likert scale. A 'strongly agree' to a positive attitude statement or 'strongly disagree' to a negative attitude statement was given a score of 5. Conversely, a 'strongly disagree' with a positive attitude statement and 'strongly agree' with a negative attitude statement was give a score of 1. Attitude scores were summed giving a possible range of scores from 7 to 35. The 6 statements of confidence in critical appraisal skills statements were scored using a 1 to 5 Likert scale and summed. A minimum overall score of 5 indicated 'little or no confidence' while a maximum total score of 30 indicated 'complete confidence'.
Critical appraisal ability was assessed through the appraisal of a systematic review article. Participants' critiques were independently assessed by two of the authors (BR & PE) using a 5-point visual analogue scale, a high score indicating a superior level of appraisal skill. A framework for scoring the reviews was developed and agreement assessed; a random sample of 20 appraisals (10 control and 10 intervention) was assessed using this framework. Intra-class correlation coefficients were calculated for each of the three aspects of critical appraisal skills assessed: 'methodology' (0.86), 'results' (0.84) and 'relevance/generalisability' (0.70), indicating satisfactory inter-assessor agreement.
Assessment of outcomes
Six months after the CAS workshop, the intervention group were asked to complete the outcome questionnaire and undertake the critique of a systematic review article (different to article used in the workshop). Five to six months after randomisation, and about one month prior to attending the workshop, controls were asked to complete the same outcomes. Thus, outcomes were obtained from both groups at about the same time after randomisation.
Statistical analysis
Primary analysis of the difference between CAS training and control groups was performed on an intention-to-treat basis, adjusting for baseline characteristics. Given that not all participants in the intervention group attended a CASP workshop, a secondary explanatory analysis was also conducted, i.e. according to whether participants received the intervention or not (see Figure 1). For continuous outcomes, multiple linear regression modeling was used to adjust for potential confounding arising from baseline differences in prognostic variables between groups. Regression model goodness of fit was checked by examining model residuals. Ordinal outcomes were compared by Mann-Whitney U tests, and binary outcomes were compared by Chi-squared analyses. Percentages and time variables were analysed as continuous variables. All analyses were carried out using STATA. All statistical tests used a level of significance of 0.05 and two-sided hypothesis testing. 95% confidence intervals (95% CI's) were calculated for differences between the two groups. No adjustment for multiple comparisons was made. However, all analyses were planned a priori and reported in full. Costs were analysed using recognized methods [14].
Figure 1 Flow diagram summarising participant recruitment and receipt of outcomes
Cost analysis
A detailed analysis of the costs of setting up and delivering the program of CAS workshops was undertaken. This cost analysis was carried out from the perspective of the NHS. Based on information about the resources and associated costs of providing the workshops, the following items were considered – costs of inviting and processing applications to attend a workshop, time of workshop organizers in the Regional R&D Office, hire of workshop venue and catering, time and expenses of workshop tutors associated with preparing and delivering the workshops, time and expenses (including locum cover) of workshop participants associated with attending the workshops. Published health and social care costs [15], local costs (e.g. NHS trust costs) and Whitley Council pay scale were used to estimate the value of staff time.
Results
Subject enrolment
Despite intensive efforts, the trial failed to recruit the target number of individuals. A revised power calculation estimated that, at 5% significance and 80% power, the 145 participants actually recruited would enable the trial to detect a difference of 0.47 standard deviation units (~20% larger than the originally powered difference). 72 were randomized to the control group and 73 to the intervention group. A total of 61 (85%) and 44 (60%) questionnaires and 43 (60%) and 21 (29%) appraisals were returned by the control and CAS training participants respectively (see Figure 1).
The two groups were well balanced for baseline demographic characteristics (see Table 1).
Table 1 Distribution of baseline characteristics of health care practitioners randomized to two groups. Values are numbers (percentages) unless otherwise stated.
Characteristics CAS training N = 73 Control N = 72
Sex, male 48 (65.8) 46 (63.9)
Age (years)
<30 2 (2.7) 4 (5.5)
30–39 20 (27.4) 20 (27.8)
40–49 37 (50.6) 32 (44.4)
50–59 12 (16.4) 13 (18.0)
60 + 2 (2.7) 3 (4.2)
Access to medical library 71 (97.3) 68 (97.1)
Prior experience of searching literature 47 (64.4) 45 (64.3)
Received formal education* in research methods 31 (42.5) 33 (47.1)
Received formal education* in epidemiology 24 (32.9) 22 (31.4)
Received formal education* in statistics 36 (49.3) 39 (55.7)
Prior involvement in research 50 (68.5) 41 (58.6)
*: postgraduate education
Study outcomes
1. Knowledge of the principles necessary for appraising evidence
Participants were asked to answer six knowledge questions, each of which had three parts. The frequency of correct answers to 4 of the 6 questions was higher in the CAS training group than the control. Total knowledge score was significantly higher for the CAS training group than controls [ITT mean difference: 2.6 (95% CI: 0.6 to 4.6); explanatory analysis mean difference 3.1 (95% CI: 1.1 to 5.2)] (see Table 2). A difference in total knowledge score of 2.0 and 3.0 corresponds to difference of 0.2 to 0.3 standard deviation units respectively i.e. below the cut off of 0.4 standard deviations units corresponding to a 'moderate' effect size [12].
Table 2 CAS training and control groups total score for knowledge of the principles necessary for appraising evidence, attitude towards the use of evidence, perceived confidence and appraisal skill.
CAS training Mean (SD) Control Mean (SD) Intention to treat analysis Mean difference+ (95% CI) Explanatory analysis Mean difference+ (95% CI)
Knowledge [range -18 to 18] 9.7 (5.3) 8.0 (5.1) 2.6 (0.6 to 4.6)* 3.2 (1.1 to 5.2)*
Attitude [range 7 to 35] 25.0 (3.8) 24.8 (4.0) 0.04 (-1.5 to 1.6) -0.04 (-1.7 to 1.6)
Confidence [range 6 to 30] 15.0 (5.3) 13.8 (5.1) 1.4 (-0.5 to 3.3) 1.13 (-0.8 to 3.1)
Appraisal skill [all range 1 to 5]
Methodology 2.4 (2.5) 2.0 (2.1) 0.6 (-0.8 to 1.9) 0.6 (-0.9 to 2.1)
Results 2.6 (2.8) 1.7 (1.8) 1.2 (0.01 to 2.4)* 1.1 (-0.2 to 2.4)
Relevance/Generalisability 2.7 (2.2) 2.4 (1.7) 0.3 (-0.8 to 1.4) 0.6 (-0.6 to 1.8)
+ Adjusted for sex, age, attendance at previous educational activity, access to medical library, prior experience of searching literature, formal education in research methods and/or epidemiology and or statistics, prior involvement in research
* Statistically significant at P ≤ 0.05
2. Attitudes towards the use of evidence about healthcare
With the exception of a more positive response to one attitude statement ('systematic reviews play a key role in informing evidence-based decisions'), in the CAS training group compared to control there were no other significant differences between groups in attitude statements. There was no evidence of difference in overall attitude score between groups (see Table 2).
3. Perceived confidence in appraising a published paper
There was no evidence of a statistically significant difference between groups in total confidence score (see Table 2).
4. Ability to appraise a systematic review
There was some evidence of the ability of participants in the CAS training group to appraise 'results' of the systematic review article [ITT mean difference: 1.2 (95% CI: 0.01 to 2.4)]. However, the difference was not significant when assessed using explanatory analysis. No difference between groups was observed in the ability to appraise 'methodology' or 'relevance/generalisability' of evidence (see Table 2).
5. Reading and evidence seeking behaviour
A comparison of various aspects of evidence seeking behaviour is detailed in Tables 3 and 4. The participants in the CAS training group self reported to: (1) read more articles, both for keeping up-to-date and for solving healthcare problems; (2) spend less time reading professional literature for keeping up-to-date, but spend more time reading professional literature for solving healthcare problems; (3) read 'thoroughly' a higher proportion of articles; and (4) use of the Cochrane library more frequently and, (5) read research reports, textbooks and other resources less frequently for solving healthcare problems. However, with the exception of (4), none of these differences were statistically significant in comparison to control
Table 3 CAS training and control groups reported number of articles read, and number of hours spent reading.
CAS training Mean (SD) Control Mean (SD) Intention to treat Mean difference+ (95% CI) Explanatory analysis Mean difference+ (95% CI)
No. articles looked at or read thoroughly each week for keeping up-to-date 5.7 (6.4) 5.1 (4.3) 0.9 (-0.6 to 1.2) 0.5 (-0.7 to 1.3)
No. hours spent reading professional literature each week for keeping up-to-date 2.2 (1.9) 2.5 (3.9) 0.9 (-0.6 to 1.2) 0.9 (-0.6 to 1.3)
No. articles looked at or read thoroughly each week to solve a health care problem 1.1 (0.8) 0.9 (0.8) 1.5 (-0.8 to 2.7) 1.4 (-0.8 to 2.7)
No. hours spent reading professional literature to solve a health care problem 0.9 (0.7) 0.9 (0.6) -0.02 (-0.4 to 0.3) -0.1 (-0.5 to 0.2)
Proportion of articles read thoroughly 21.9 (23.6) 19.2 (19.9) 1.3 (-0.8 to 2.0) 2.6 (-0.7 to 1.8)
Proportion of articles skim read 37.0 (20.8) 42.3 (24.9) -5.7 (-15.4 to 4.1) -8.2 (-18.1 to 1.6)
Proportion of articles for which only abstracts read 49.7 (23.4) 40.8 (26.7) 7.9 (-3.3 to 19.1) 12.0 (1.0 to 23.0)*
+ Adjusted for sex, age, attendance at previous educational activity, access to medical library, prior experience of searching literature, formal education in research methods and/or epidemiology and or statistics, prior involvement in research
* Statistically significant at P ≤ 0.05
Table 4 CAS training and control groups use of the resources for solving a health care problem
CAS training Median (LQ, UQ) Control Median (LQ, UQ) Median Difference (p-value) †
Review articles 2.0 (1.0, 3.0) 2.0 (1.25, 2.0) 0 (0.66)
Research reports 1.0 (1.0, 2.0) 2.0 (1.0, 2.0) 1.0 (0.97)
Secondary journals 2.0 (1.0, 3.0) 2.0 (1.0, 2.0) 0 (0.22)
Textbooks 2.00 (2.0, 3.0) 3.0 (1.0, 3.0) 1.0 (0.77)
Worldwide Web 1.0 (0, 2.0) 1.0 (0, 2.0) 0 (0.98)
Guidelines 2.0 (2.0, 3.0) 2.0 (2.0, 3.0) 0 (0.64)
Cochrane Library 1.0 (0, 2.0) 0 (0, 1.0) 1.0 (0.05)
Colleagues 3.0 (2.75, 3.0) 3.0 (2.0, 3.0) 0 (0.55)
Other resources 2.0 (0, 3.0) 3.0 (1.5, 3.75) 1.0 (0.41)
† Mann Whitney test; Likert Scale: '0': 'never'; '1': 'rarely'; '2': 'occasionally'; '3': 'often' & '4': 'very often'; UQ: upper quartile; LQ: lower quartile
Costs
The mean cost to the NHS of conducting the CAS workshops was £250 per person (see Table 5). The majority of this cost (approximately £140) resulted from salary costs associated with the time of the participants attending the workshop. The remaining costs of the workshops were associated with the administration (approximately £25 per person), venue hire (approximately £42 per person), and tutors' time and travel (approximately £49 per person). There was some variation in the cost (from approximately £240 – £340 per person) across the 7 workshops, due to the attendance level, i.e. workshops with the most participants tended to have the lower cost.
Table 5 Summary of costs of CAS training
Workshop I II III IV V VI VII Total Total per Head
No. Attendees 16 19 9 14 18 5 7 88
Administration* £322 £329 £313 £318 £316 £295 £300 £2,193 £25
Venue £1028 £670 £285 £916 £475 £215 £74 £3,663 £42
Participants' costs† £2,074 £2,825 £1,567 £1,878 £2,179 £793 £992 £12,310 £140
Tutors' costs £709 £719 £890 £712 £555 £636 £73 £4,294 £49
Total £4,132 £4,542 £3,057 £3,824 £3,525 £1,940 £1,439 £22,460. £255
Total per head £258 £239 £340 £273 £196 £388 £206 £255
* Costs of initial invitations, copyright permission for use of paper, invitations to participants, pre workshop & post workshop pack production, postage costs, preparation for workshop, and travel bookings.
† Costs included participants' time at workshop with the exception of GPs, in these cases the cost of locum cover was applied.
Discussion
The results of this prospective randomized controlled trial demonstrates that a half-day CAS workshop can elicit small improvements in healthcare professionals' knowledge of the principles and theory of evidence-based practice and some improvement in aspects of their critical appraisal skills ability. Nevertheless, we found little evidence of any improvement, as a result of CAS training, in the other study outcomes, i.e. participants' attitude towards evidence or their evidence seeking behaviour. Taking into account the set up costs and of time and locum expenses of participants, the mean cost of conducting these CAS training workshops was about £250 per person. The lack of substantive improvements in knowledge, skills and attitudes outcome observed in this trial are consistent with previous studies of CAS training [6-9].
Potential limitations of this study
The number of participants recruited was less than that intended, not all participants provided outcomes and the trial was about 20 percent under the desired power. Nevertheless this study remains the largest randomized controlled trial to date and some statistically significant differences were observed.
The educational context in which this randomized trial was undertaken imposed certain constraints on its conduction and execution. As a result, poor recruitment, loss to follow up and poor uptake of the CAS training experienced by this trial may have threatened both its internal validity and generalisability. However, efforts were made in the analysis of the findings of this trial to overcome these limitations. The return of outcomes in this trial could not be mandatory. Despite considerable efforts by the project team (reminders and personal telephone calls from the trial principle investigator to participants), we failed to obtain a substantial proportion of outcomes in the trial participants – 60% and 85% of the knowledge, attitude and behaviour outcomes were obtained for CAS training and control groups respectively, and even less for the critique of the published systematic review. It is plausible that respondents may have differed in some way to non-respondents, such as in their level of motivation, and may therefore responded more positively to this educational intervention. However, this was not supported by the poor outcome response rate. Moreover there was no evidence of a difference in the baseline characteristics of participants who returned their outcomes, and those who did not. A differential response rate across the two study groups possibly reflects a greater reluctance in those individuals who had undertaken the educational intervention to return their outcomes (i.e. 'more to lose') compared to those in the control group. If true, the direction, in terms of over- or underestimating the impact of the intervention, is uncertain. An interview-administered assessment, rather than a mail based one, may have enhanced outcome response rate.
Of the 73 participants allocated to receive CAS training only 52 (71%) actually attended. The reasons for this were unclear, and were not formally addressed within this study. In addition to conventional intention-to-treat analyses, secondary explanatory analyses, i.e. based upon the participants who actually did attend the workshop, were undertaken. That there were no differences between groups for most outcomes, irrespective of whether an intention-to-treat or explanatory analysis, was used (see Tables 2 and 3) suggests that the poor intervention uptake was not important source of bias.
Implications of findings
With the drive to evidence-based practice in recent years, considerable efforts have been made in providing CAS training as part of healthcare professionals' undergraduate and postgraduate activities in many countries. The findings of this study, the largest randomized controlled trial to date, provide only limited support for such training. However, it is important to put this finding in the appropriate educational context. The half-day CASP workshop evaluated in this trial has been widely disseminated and its duration and format is consistent with many previous CAS interventions [9]. Nevertheless it is probably unrealistic to expect that the half-day workshop evaluated in this trial would in itself result in changes in professional behaviour. This is supported by a large body of evidence and theory on changing professional practice [17]. Therefore it is important to see, and assess, CAS training, not in isolation, but as one part of education approach towards evidence-based practice or as a part of the undergraduate and postgraduate curriculum. It is also important to reassess the objective of CAS training. With increasing availability of carefully appraised evidence such as secondary journals (e.g. Evidence Based Medicine) and on-line critically appraised topics ('CATs'), the most important role of CAS training may be simply be to sensitise participants to the availability of high quality evidence. Further debate is therefore needed about refocusing critical appraisals skills training towards finding such evidence and the role of healthcare librarians and the new initiatives such as the National Electronic Library for Health. A number of commentators have criticised previous evaluations of CAS training for not using experimental designs [6-9]. However, the experience of this study has demonstrated some of the difficulties in implementing an evaluation of 'real life' educational intervention using such an experimental design. The difficulty of employing randomized controlled trials in the evaluation of educational interventions has been highlighted by others [18]. Future evaluations of CAS and other educational interventions aimed at promoting evidence-based practice need to take into account both these perspectives.
Conclusions
This prospective randomized controlled found small improvements in self-selected healthcare professionals' knowledge and understanding of the medical literature and appraisal skills with critical appraisal skills training. No improvement was observed in attitudes towards the use evidence and evidence-seeking behaviour. The findings of this study challenge the policy of funding in isolation 'one-off' educational interventions aimed at enhancing the evidence-based practice of health care professionals. Future evaluations of evidence-based practice interventions need to take in account both this trials' negative findings and methodological difficulties.
List of abbreviations
CAS – critical appraisal skills
CASP – Critical Appraisals Skills Programme
95% CI – 95 percent confidence interval
ITT – intention to treat
NHS – National Health Service
R&D – research and development
Competing interests
The author(s) declare that they have no competing interests.
Author's contributions
RT, BR and PE conceived, designed and secured funding for the trial. RST drafted the paper. RJT collected the study outcomes and undertook the data analysis. RST is a guarantor for the study.
Funding
NHS R&D Executive: Evaluating methods to practice the implementation of R&D [project no. IMP 12-9]
Appendix 1. Objectives, syllabus, and delivery methods of critical appraisal skills workshop for health care decision makers
Workshop objectives (taken from workshop materials)
• To critically appraised a published review article.
• To understand the terms systematic review and meta-analysis.
• To be able to explain why critical appraisal skills are important for provision of health care.
• To have greater confidence in your ability to make sense of the research evidence.
Workshop format
3 hours attendance (also advised to undertake at least 1 hour preparation reading the article to be appraised in the workshop and address a written 'clinical scenario')
• Introductory talk: overview of the importance of evidence based health care practice, the theoretical basis of the appraisal of a systematic review, and orientation to the JAMA appraisal guideline (~60 mins).
• Small group work: appraisal of a published systematic review (~60 mins).
• Plenary session: feedback from the small group, general discussion of the relevance of the appraisal to clinical scenario and ballot of opinions on the clinical scenario. (~60 mins)
All workshops were run by 3 to 4 individuals each of whom had a formal training in health services research methods and were experienced in delivering CASP workshops.
Workshop materials
One to two weeks prior to the workshop, a pre-workshop pack was sent to participants.
• Workshop objectives.
• Orientation guide.
• Clinical scenario and questions
• Systematic review paper.
• Glossary.
One to two weeks post workshop, a post workshop pack was sent to participants:
• Introductory talk slides.
• Systematic review checklist.
• JAMA guidelines for systematic review [15].
Educational rationale
The workshop is based on the Critical Appraisal Skills Programme (CASP) developed by Oxford Regional Health Authority and developed from the educational methods of McMaster University in Canada [5]. The 'McMaster model' key features include, self-directed learning, small group teaching methods and the importance of grounding education within the clinical decision making process.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
MS-Word format, contains study outcome questionnaire
Click here for file
Acknowledgements
We thank all the health practitioners who participated; the former South and West NHS R&D Regional Office, for assistance in undertaking the trial workshops; the trial advisory group – Dr Amanda Burls, Prof. Martin Eccles, Dr Ruairdih Milne and Prof. David Sackett; and Mr John Keast, Dr Sarah Binns and Dr Joanna Hartland for their contributions to the delivery of the trial.
==== Refs
Sackett DL Richardson WS Rosenberg WMC Haynes RB Evidence based medicine: how to practice and teach Evidence Based Medicine 1997 London, BMJ Publishing Group
McColl A Smith H White P Field J General practitioner's perceptions of the route to evidence-based medicine: a questionnaire survey BMJ 1998 316 361 5 9487174
Oxman AD Sackett DL Guyatt GH Evidence Based Medicine Working Group Users guide to the medical literature: I How to get started JAMA 1993 270 2093 2095 8411577 10.1001/jama.270.17.2093
Greenhalgh T How to read a paper: Getting your bearings (deciding what the paper is about) BMJ 1998 315 243 246 9253275
Milne R Donald A Chambers L Piloting short workshops on the critical appraisal of reviews Health Trends 1995 27 120 123 10162322
Norman GR Shannon SI Effectiveness of instruction in critical appraisal (evidence-based medicine): a critical appraisal CMAJ 1998 158 177 81 9469138
Taylor R Reeves B Ewings P Binns S Keast J Mears RA Systematic review of the effectiveness of critical appraisal skills training for clinicians Med Educ 2000 34 120 5 10652064 10.1046/j.1365-2923.2000.00574.x
Parkes J Hyde C Deeks J Milne R Teaching critical appraisal skills in health care settings (Cochrane Review) The Cochrane Library 2004 Chichester, UK: John Wiley & Sons, Ltd
Coomarasamy A Taylor R Khan KS A systematic review of postgraduate teaching in evidence-based medicine and critical appraisal Med Teach 2003 25 77 81 14741863 10.1080/0142159021000061468
Linzer MJ Brown JT Frazier LM DeLong ER Siegel WC Impact of a medical journal club on house-staff reading habits, knowledge, and critical appraisal skills. A randomized controlled trial JAMA 1988 260 2537 41 3050179 10.1001/jama.260.17.2537
Welch WW Walberg HJ Pretest and sensitisation effects in curriculum evaluation Amer Educ Res J 1970 7 605 614
Cohen C The statistical power of abnormal-social psychological research: A review J Abnor Soc Psychol 1962 65 145 153
Taylor RS Mears R Reeves B Ewings E Binns S Keast J Khan K Developing and validating a questionnaire to evaluate the effectiveness of evidence-based practice teaching Med Educ 2001 35 544 7 11380856 10.1046/j.1365-2923.2001.00916.x
Barber JA Thompson SG Analysis and interpretation of cost data in randomized controlled trials: review of published studies BMJ 1998 317 1195 1200 9794854
Netten A Dennett J Unit costs of health and social care 1996 University of Kent, Personal Social Services Research Unit
Oxman AD Cook D Guyatt GH Evidence Based Medicine Working Group Users guide to the medical literature: How to use an overview JAMA 1994 272 1367 1371 7933399 10.1001/jama.272.17.1367
Bero LA Grilli R Grimshaw JM Harvey E Oxman A Thomson MA Getting research findings into practice: Closing the gap between research and practice: an overview of systematic reviews of interventions to promote the implementation of research findings BMJ 1998 317 465 468 9703533
Prideaux D Researching the outcomes of education interventions: a matter of design BMJ 2002 324 126 127 11799017 10.1136/bmj.324.7330.126
| 15585061 | PMC539272 | CC BY | 2021-01-04 16:30:54 | no | BMC Med Educ. 2004 Dec 7; 4:30 | utf-8 | BMC Med Educ | 2,004 | 10.1186/1472-6920-4-30 | oa_comm |
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-4-311558831810.1186/1472-6920-4-31Research ArticleTeaching statistics to medical students using problem-based learning: the Australian experience Bland J Martin [email protected] Department of Health Sciences, University of York, Heslington, York YO10 5DD, United Kingdom2004 10 12 2004 4 31 31 12 8 2004 10 12 2004 Copyright © 2004 Bland; licensee BioMed Central Ltd.2004Bland; 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
Problem-based learning (PBL) is gaining popularity as a teaching method in UK medical schools, but statistics and research methods are not being included in this teaching. There are great disadvantages in omitting statistics and research methods from the main teaching. PBL is well established in Australian medical schools. The Australian experience in teaching statistics and research methods in curricula based on problem-based learning may provide guidance for other countries, such as the UK, where this method is being introduced.
Methods
All Australian medical schools using PBL were visited, with two exceptions. Teachers of statistics and medical education specialists were interviewed. For schools which were not visited, information was obtained by email.
Results
No Australian medical school taught statistics and research methods in a totally integrated way, as part of general PBL teaching. In some schools, statistical material was integrated but taught separately, using different tutors. In one school, PBL was used only for 'public health' related subjects. In some, a parallel course using more traditional techniques was given alongside the PBL teaching of other material. This model was less successful than the others.
Conclusions
There are several difficulties in implementing an integrated approach. However, not integrating is detrimental to statistics and research methods teaching, which is of particular concern in the age of evidence-based medicine. Some possible ways forward are suggested.
==== Body
Background
Problem-based learning (PBL) is a method of teaching and learning which is widely used in the education of medical students [1,2]. In problem-based learning, students working in a small group are presented with a problem, typically a description of a patient presentation. They decide what features of the problem are outside their present knowledge and divide these topics between them. They then research their topics using library and internet material and report back to the next small group tutorial with their findings. In a problem-based learning curriculum, this is the principal method of learning. More traditional methods, such as lectures and practical exercises, provide background and support material. An example of a PBL teaching problem is given in the Appendix.
This project was conceived at a meeting of statisticians from UK medical schools, which is held every year at the Burwalls conference centre, University of Bristol. In 2002, one of the topics for discussion was problem-based learning (PBL). Several things became apparent during this discussion. First, most of the statisticians present had no idea what problem-based learning was. Second, in the few UK medical schools where PBL was in use, statistics and research methods were not being taught through this medium. Statisticians in some of these schools felt excluded. Third, the new medical schools which were starting up in the UK were adopting PBL as their method of teaching.
As far as I know, only four of the older UK medical schools have adopted a PBL curriculum at the time of writing: Liverpool, Manchester, Glasgow, and St. George's. St. George's is a special case, as it has two medical courses. The five-year undergraduate entry course is taught in the traditional way, with a small 'case-based learning' element added on. The new four-year Graduate Entry Programme (GEP) is based on the McMaster model, a PBL course which takes graduates in any subject [2]. In none of these courses are statistics and research methods taught as an integral part of the PBL. At St. George's, for example, there is a parallel, non-PBL, seminar-based course. This attempts to match the illustrative material to the case of the week, but is taught separately from that case.
New medical schools, such as Anglia, Peninsular, and Hull-York, are being set up as PBL courses [3]. This is particularly suitable for courses based in more than one university, as in Peninsular (Plymouth and Exeter) and Hull-York. Problems can be set by teachers in either centre, and presented to students in both.
The adoption of PBL in new medical schools may indicate the possibility that PBL will become more widely adopted in UK medical schools. If this were to happen, there would be a real danger of statistics and research methods being marginalised in the medical curriculum. It is difficult enough to persuade medical students of the importance of these topics and it would be even more so if it were taught outside the mainstream of the course. More importantly, in the era of evidence-based medicine these topics should be central to the medical curriculum, not on the sidelines.
Methods
During a sabbatical visit to Australia, I spent one to three weeks each at most of the medical schools. I interviewed educationalists and statisticians at each of these to ask how statistics and research methods were taught to medical students in their problem-based learning curricula. I had no prior knowledge of how this might be done, so the interviews were open in nature, rather than structured. I enquired about statistics as a subject in its own right and the wider principles of research, critical reading of research, and evidence-based medicine. I then synthesised the information collected to provide a picture of the current situation.
I visited the following universities with medical schools: the University of Western Australia, Perth, Flinders University, Adelaide, Monash University, Melbourne, the University of Melbourne, the Australian National University, Canberra (medical school about to commence), the University of Sydney, the University of Newcastle, and the University of Queensland, Brisbane. I omitted the University of Adelaide, the University of New South Wales, and James Cook University, Townsville, but was able to get information from them by email. I identified potential informants from the university website and emailed them as follows: 'I am interested in the teaching of statistics to medical students. Last year I visited several medical schools in Australia, but unfortunately I did not manage to visit [your university] on this trip, a serious omission. I have written a report on my experiences, which is available on one of my websites at . I am particularly interested in how statistics is taught in problem-based learning programmes. I am presenting this material at the forthcoming International Biometrics Conference in Cairns and preparing a paper for possible publication. I would be very interested to learn how these matters are ordered in [your university] . . . Are you the right person to ask? If not, could you suggest someone who would be?' I obtained helpful replies from all three universities.
Results
I went to Australia in search of the fully integrated teaching of statistics and research methods as part of the PBL tutorials. I did not find it anywhere. I did find three different models:
• Material integrated but separately taught,
• A parallel course,
• PBL used only for 'public health' related subjects.
In addition, one school (New South Wales) did not use problem-based learning and one (James Cook) taught virtually no statistics. More information about the individual medical schools is available [see Additional file 1].
Material integrated but separately taught
This approach was used or planned at the University of Sydney, the University of Melbourne, and the Australian National University. However, of these only the University of Sydney had actually put this into practice. The University of Melbourne and the Australian National University were about to implement what was essentially the Sydney model.
In this model, statistics and research methods are taught by PBL and the PBL triggers are integrated with the PBL problems for other parts of the course, but the material is not taught in the same tutorials or by the same tutors as anatomy, biochemistry and physiology. There will be a separate set of triggers for the statistics, etc., presented in a separate tutorial, and by separate tutors.
I asked why the main PBL tutors could not do this. As I understood it, the function of a PBL tutor is to facilitate and guide the group, not to impart knowledge. I had no problem, at least that I was aware of, in acting as PBL tutor when students were working with triggers designed to elicit questions about anatomy, biochemistry, and physiology, subjects of which I know virtually nothing. Besides, many of these tutors must routinely read journals which bristle with P values, t tests, correlation coefficients, etc. They must be familiar with the terms, if nothing else. Answers to this included:
• the tutors themselves refused to do it,
• in the early years of the course we need expert tutors,
• many tutors are still rooted in the old paradigm and are reluctant to embrace EBM and it was the view of my informants that experience round the world shows that it is difficult to teach EBM principles.
The team at the University of Sydney were on the whole positive about their course.
A parallel course
This approach was used at Monash University, the University of Queensland, Flinders University, the University of Newcastle, and the University of Adelaide.
In this approach, a non-PBL course is given separately from the main PBL course. This may consist of any combination of lectures, seminars, practicals, web pages, or text handouts. Usually there is an attempt to link this to the PBL cases by using examples related to the case of the week. For example, the case of the week might be asthma and the parallel course could include a critique of a paper reporting a trial of a treatment for asthma.
There are several problems with this approach to teaching statistics and research methods. As noted in the Background, the subject may seem peripheral to the main thrust of the medicine course. Student feedback tends to give a much lower approval to parallel courses than to the main PBL teaching. Finally, teaching is dependent on the cases chosen by the PBL teachers, who may change the cases or reorder them at little or no notice. This can make statistics teaching, which is much more dependent on the order of presentation than most subjects in the medical curriculum, extremely difficult.
Most people involved in these courses were unhappy with them, the exception being a group project in the Flinders course.
PBL used only for 'public health' related subjects
This approach was used at the University of Western Australia.
This was an unusual model, found at only one university. PBL teaching had been initiated by an enthusiast, after a period spent at McMaster University. She was a member of the public health group and persuaded her colleagues to introduce PBL. However, only about one third of the course is taught this way, anatomy, biochemistry and physiology are taught traditionally. The consequence of this approach is that the tutors are drawn from the population medicine area and so are quite happy to teach statistics, research methods, and EBM. The triggers can be chosen as population-oriented problems, rather than being restricted to the patient case.
People I spoke to were very positive about this course, not surprisingly as they saw themselves as the educational leaders in their institution.
Discussion
I did not to find anywhere in Australia a single instance of truly integrated teaching of statistics and research methods through PBL. I have to conclude from this that there are considerable difficulties. I did, however, find an instance of unintegrated PBL teaching, at the University of Western Australia, where anatomy, biochemistry and physiology have continued to be taught in a more traditional manner.
What are these difficulties?
• If tutors are drawn mainly from the laboratory disciplines, they may be unsympathetic to the population and clinical foci of EBM and its core subjects. We need strong advocates to persuade them of the importance of these topics.
• If tutors are drawn from the clinical staff, they may be unsympathetic to the idea of EBM as a core activity. Converting the existing clinical teachers to the new paradigm of EBM may take a generation, but this will be a problem whatever teaching method we use.
• Tutors may be ignorant of the principles and details of statistics. This may be true. My own experience as a PBL tutor has been that lack of knowledge has seldom been a problem. It is not the function of a PBL tutor either to impart knowledge or to explain concepts. Persuading tutors of this should be part of their general PBL training.
• The changing patterns of cases as the course develops is a particular problem for statistics. This is undoubtedly true.
• The nature of the subject does not lend itself to PBL. It would certainly be difficult to deduce the calculations required for a t test from a problem based on a patient with asthma. However, we must consider what we actually want to teach. I do not think there is much point in teaching undergraduate medical students to analyse data, even using computers. What we need to teach them is how to understand research publications, the evidence on which we hope their future evidence-based practice will be based. We can certainly do this using triggers such as published papers, as the University of Western Australia has demonstrated.
• The patient case is not suited to teaching these subjects. This is true, though I do not think it is beyond us to devise cases which lead to research and evidence questions. However, we must persuade our colleagues that the patient vignette is not the only type of problem which we can use.
We should not see only the difficulties, however, but also the opportunities. If we can integrate statistics and research methods into PBL, there may be great advantages. One consequence of integration would be that the subject would not be marginalised or seen as separate. It would be just one aspect of medicine. Indeed, one of the features of a PBL course is that the distinctions between the different subjects should blur.
A second advantage would be that students would be learning statistics and research methods in the contexts in which we hope that they will apply them: the interpretation of clinical data and the assessment of research evidence. The relevance of the subject should be very clear and during their professional careers they would be more likely to be able to recall and use this material when needed.
So how can we do it? There must be an advocate for statistics and research methods at the start of the preparations for PBL, who is able to argue convincingly for the inclusion of these subjects. They must form part of the matrix of topics and problems. This is not an easy task and I have been told of great difficulty experienced at this level, as other teachers sought to exclude these subjects so that their own discipline could have more time. This is a natural human reaction, of course. I think that statistics and research methods are central to medical education, and would argue for more of them, at the expense of some anatomy if necessary.
We must get away from the idea that the problem must be a patient vignette. There are some statistical topics which can be covered quite conveniently in this way, such as measurement error, coefficients of variation, reference intervals, sensitivity and specificity, etc. After all, when my GP looks at my serum cholesterol on his practice computer, it has by the side of it a 95% reference interval. However, problems could equally be a published paper. We use these routinely in the teaching of statistics, research methods, and critical appraisal.
We could use a paper as a trigger for non-research methods topics. For example, a paper on an asthma trial could trigger questions about asthma as well as about randomisation. This may lead to too many questions being raised by the trigger. Another possibility would be to have such a trigger immediately following a case vignette problem on the disease in question. The patient vignette would raise the questions on anatomy, biochemistry, pharmacology, etc., associated with the disease. The following problem using a research paper would then raise only the research methodological questions. The fixed resource sessions in this week would then be devoted to these.
We may also, for variety, link research publications to the patient vignette by devices such as newspaper articles which the patient presents to the physician (e.g. reporting a trial), or say that in a patient problem the clinician has already found a Cochrane review.
Fixed resource sessions could include lectures, but I am very reluctant to suggest formal lectures in statistics and research methods for medical students. I would prefer to offer open question and answer sessions, where students can ask the statistician lecturer to explain anything they are unsure of. For example, if the PBL trigger is a randomised controlled clinical trial with results presented in terms of P values, we might be expecting this trigger to lead to questions such as 'why randomize?', 'what does P < 0.05 mean?', and 'should patients be told they are in a trial?'. Students should have made some attempt to answer these before the fixed resource session.
Conclusions
There are several difficulties in implementing an integrated approach. However, not integrating is detrimental to statistics and research methods teaching, which is of particular concern in the age of evidence-based medicine. I remain optimistic that we can incorporate statistics and research methods into a fully problem-based curriculum. I think that if we do not, medicine will be the poorer for it.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
J M Bland is the sole author.
Appendix
The following is a typical example drawn from the teaching at St. George's Hospital Medical School. The students work in a small group with a tutor to act as facilitator. The students are given the following information:
'You are a member of the on-duty Trauma Team in the Accident and Emergency department one Monday evening, when you are alerted to the imminent arrival of an "RTA (road traffic accident) patient". As she is being wheeled in to the Emergency Room, the paramedic accompanying her reports on the circumstances:
'The patient is called Janet Phillips, she appears to be a medical student in her early 20s, and was on a Pelican crossing when she was struck by a car which had gone out of control. Janet had been thrown some distance from the car by the force of the impact. She is confused, in shock, has superficial lacerations to her face and severe pain in her pelvis and legs.
'At the accident site her airway was cleared, oxygen administered, a traction splint applied to her right leg, and she was immobilised on a long spinal board.
'The driver had only minor injuries but his breath smelled strongly of alcohol, and he was now in police custody.
'In the Accident and Emergency Department Janet is given a pelvic clamp to arrest her internal bleeding.'
The students are told that they should address four general themes in their discussion: basic and clinical sciences, patient and doctor, community and public health, and professional development. In the first tutorial, they are asked to work through the following steps:
1. Clarify any terms and concepts in the scenario with which they are not familiar.
2. Define the problem(s) and issue(s) raised by the scenario.
3. Analyse the problems and issues, seeking explanations or hypotheses.
4. Agree on specific questions (learning objectives) for each of the four general themes to which you need answers.
They then work individually on tackling the questions agreed upon. In thesubsequent tutorial they discuss their findings and decide how the problems and issues raised by the scenario could be resolved. They may then be given further information about the patient, which might raise further topics for their research. When a problem has raised all the points required, a fresh problem is presented to the students.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
'The Australian Universities visited'. This provides a descriprtion of teaching at the universities visited as part of this project.
Click here for file
Acknowledgements
I thank the many people in Australia who provided me with places to work and stay, particularly Max Bulsara, David Prideaux, Jenny McCulloch, Andrew Forbes, Judy Simpson, Caro Badcock, Kerrie Mengersen, Denise Schultz, Paul Glasziou, Susan Button, Marilyn Chalkley, and Hugh Sadler. I must thank the medical educators and statisticians who provided me with information about their courses: Sandra Carr, Judith Flynn, Ian Jacobs, Matthew Knuiman, Sally Reagan, Neil Piller, David Prideaux, David Goddard, Rory Wolfe, John Carlin, Sue Elliott, Steve Farish, Terry Nolan, Bruce Shadbolt, Alex Barrett, Jill Gordon, Greg Ryan, Lindel Travino, John Attia, Julia Byles, Bob Gibbard, Michael Hensley, Alison Koschel, Paul Glasziou, Marie Louise Dick, Luc Betbeder, Eilean Watson, Philip Ryan, and Reinhold Muller. Next, I thank those at St. George's Hospital Medical School who enabled this trip to take place, Sir Robert Boyd, Principal, and Ross Anderson, my head of department, both of whom supported it even though I then resigned my post to go to York. Finally, I am greatly indebted to Janet Peacock, who took over my teaching and leadership of our statistical group while I was away.
==== Refs
Bligh J Problem based, small group learning: an idea whose time has come Brit Med J 1995 311 342 343 7640532
Neufeld VR Woodward CA MacLeod SM The McMaster M.D. program: a case study of renewal in medical education Academic Medicine 1989 64 423 432 2751777
Howe A Campion P Searle J Smith H New perspectives–approaches to medical education at four new UK medical schools Brit Med J 2004 329 327 331 15297339
| 15588318 | PMC539273 | CC BY | 2021-01-04 16:30:53 | no | BMC Med Educ. 2004 Dec 10; 4:31 | utf-8 | BMC Med Educ | 2,004 | 10.1186/1472-6920-4-31 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-5-1771554117010.1186/1471-2105-5-177Research ArticleCalibration and assessment of channel-specific biases in microarray data with extended dynamical range Bengtsson Henrik [email protected]önsson Göran [email protected] Johan [email protected] Mathematical Statistics, Centre for Mathematical Sciences, Lund University, Box 118, SE-221 00 Lund, Sweden2 Department of Oncology, Lund University, Barngatan 2, SE-221 85 Lund, Sweden2004 12 11 2004 5 177 177 21 6 2004 12 11 2004 Copyright © 2004 Bengtsson 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
Non-linearities in observed log-ratios of gene expressions, also known as intensity dependent log-ratios, can often be accounted for by global biases in the two channels being compared. Any step in a microarray process may introduce such offsets and in this article we study the biases introduced by the microarray scanner and the image analysis software.
Results
By scanning the same spotted oligonucleotide microarray at different photomultiplier tube (PMT) gains, we have identified a channel-specific bias present in two-channel microarray data. For the scanners analyzed it was in the range of 15–25 (out of 65,535). The observed bias was very stable between subsequent scans of the same array although the PMT gain was greatly adjusted. This indicates that the bias does not originate from a step preceding the scanner detector parts. The bias varies slightly between arrays. When comparing estimates based on data from the same array, but from different scanners, we have found that different scanners introduce different amounts of bias. So do various image analysis methods. We propose a scanning protocol and a constrained affine model that allows us to identify and estimate the bias in each channel. Backward transformation removes the bias and brings the channels to the same scale. The result is that systematic effects such as intensity dependent log-ratios are removed, but also that signal densities become much more similar. The average scan, which has a larger dynamical range and greater signal-to-noise ratio than individual scans, can then be obtained.
Conclusions
The study shows that microarray scanners may introduce a significant bias in each channel. Such biases have to be calibrated for, otherwise systematic effects such as intensity dependent log-ratios will be observed. The proposed scanning protocol and calibration method is simple to use and is useful for evaluating scanner biases or for obtaining calibrated measurements with extended dynamical range and better precision. The cross-platform R package aroma, which implements all described methods, is available for free from .
==== Body
Background
The microarray technology provides a way of simultaneously measuring transcript abundances of 103 – 105 genes from one or more cell or tissue samples. A microarray, also known as a gene chip, has well defined regions that each consists of immobilized sequences of DNA, which each is unique to a specific gene. These regions are referred to as probes [1]. When fluorophore labeled cDNA, referred to as targets, obtained by reverse transcription of mRNA extracted from the samples of interest is let to hybridize to the probes for a few hours, each region on the microarray will specifically bind a certain amount of hybridized DNA unique to the corresponding gene. Depending on if a two-channel or single-channel microarray platform is used, either several and differentially labeled targets are hybridized to the same array, or different targets are each hybridized to separate arrays using identical labels. Next, the array is scanned at different wavelengths to excite the fluorescent molecules using a light source, for instance a laser. Shortly after the fluorophores have been excited they emit photons, which are registered and quantified in each position by the scanner, which results in a high-resolution digitized image for each channel. Using image analysis methods, the pixels that belong to the regions that contain the probes are identified and averaged, and an estimate of the transcript abundance for each gene is obtained.
Since these estimates are obtained from a complex measurement process of several steps, it is likely that the observed signals contain not only measurement noise, but also systematic variations of different kinds [2].
In this report, we show the existence of a channel-specific bias introduced by the scanner and most likely its detector parts. Our results indicate that the image analysis may also contribute with a small bias. The effects channel-specific biases have on the downstream microarray analysis are many [2,3]. We suggest a scan protocol and a model that will allow us to estimate the biases and calibrate the observed signals accordingly. The result will be that the intensity dependent effects are removed, but also that the effective dynamical range of the scanner is increased several times.
Model
General model
Consider a microarray experiment involving genes i = 1 ,..., I from RNA extracts c = 1 ,..., C. In single-channel microarrays each array measures the gene expression levels in one RNA extract, whereas in two-color microarrays each array measures two RNA extracts, one in each channel. We will refer to each set of signals from each RNA extract as channels. Let μc,i be the true gene expression (transcription) level of gene i in channel c. Ideally, statistical analysis can then be done on these quantities. For instance, by comparing the relative abundances in two channels, that is ri = μ1,i/μ2,i for all genes i, it is possible to identify genes that are significantly differentially expressed (ri ≠ 1). However, in reality we do not observe the true expression levels, but only the quantified spot intensities yc,i. The general relation between the observed and the true expression levels can be written as
yc,i = fc(μc,i) + εc,i, (1)
where fc(·) is an unknown channel-specific function, which we refer to as the measurement function, that includes all steps in the microarray process. Moreover, we assume independent intensity dependent error terms εc,i such that E[εc,i] = 0. Because we want to do inference based on μc,i, it must be possible to find the inverse of fc(·), which (at least in theory) is possible if it is strictly increasing.
To be able to find the form of fc(·), high quality calibration data from several stages along the microarray process is required. Here we will consider a simpler case. Split the overall measurement function into two parts. The first part xc,i = gc(μc,i) models, in channel c, the amount of light from spot i that enters the photomultiplier tube (PMT) [4] as a function of the transcription level of clone i. The second part, which is studied in this report, is yc,i = hc(xc,i) and models the observed signal as a function of the amount of photons in channel c and spot i that enters the PMT. That is, it captures the characteristics of the scanner's light detector, but also of the image analysis methods. We want to emphasize that the light from one spot does not necessarily originate solely from the fluorescent molecules that are attached to the hybridized target DNA. Light from other sources such as cross-hybridized target, intrinsic fluorescence from spot buffer, and scatter light may also contribute with photons of similar wavelengths.
Next we will show that hc(·) is almost perfectly affine. This measurement function also depends on the scanner settings, especially the scanner sensitivity, which is indicated below with the super index (k). In other words,
where for each fix scanner setting k, and are channel-specific bias and scale parameters, respectively. Assume that xc,i is fix for all PMT voltages.
Note that the above relationship is not necessarily linear. Here we refer to linear in the strict sense where ac = 0 so that the output is proportional to the input. Lack of linearity has important implications on downstream analysis. For instance, when spotted as well as in-situ synthesized microarrays are used it is common to do statistical analysis on the log-ratios Mi = log2(yR,i/yG,i) and the log-intensities for all genes i [5], where we for convenience have denoted the two channels to be compared by R and G although such comparisons are not limited to within-array measurements. One of the rationales for this bijective transform is that under ideal conditions the main measure of interest, the fold change, is contained in one variable only. However, a channel specific bias introduced by fc(·) will introduce a so called intensity dependent bias in the observed log-ratios. Commonly observed intensity dependent effects in the log-ratios [6] can partly be explained by the fact that the logarithm is taken on affinely transformed signals [2,3,7].
Constrained model
The model in equation (2) is not identifiable. We address this problem as follows. Consider the case where the same array has been scanned at K different PMT settings. Let be the vector of the K quantified signals for gene i and channel c. In the noise-free case it follows from (2) that lies on the line L(ac, bc) in K, which has direction and goes through the point . The 2K parameters of ac and bc are not identifiable, since L has only 2K - 2 degrees of freedom. In fact, any transformation bc ← k·bc and ac ← ac + l·bc, where k and l are scalars, will leave L intact. In this paper, we make ac and bc identifiable by choosing k and l so that = 1 and ac is the point on L closest to the (diagonal) line L' = {ec(1,..., 1); ec∈ }. The choice of ac can be motivated by looking at observed data. By inspection, we observe that the bias in Model (2) is not varying much when the PMT gain is changed. To demonstrate this, have been plotted for each of the six possible PMT pairs in Figure 1. First, the close fit of the lines to data is evidence that the scanner is linear in its dynamical range. Second, all lines go through approximately the same point, lets call it (ec, ec), suggesting that there is a common PMT-independent bias ec. More precisely, we split the bias term into two parts, one dependent and one independent on the PMT gain according to
and define ∈ K. For this split, data indicates that ||dc|| ≈ 0, where ||·|| is the norm in, say, L2 (Euclidean distance). Let d = y - e(1,..., 1) where y ∈ L and e ∈ . The constraint that ac is the point of L closest to L' can then be formulated as
where the minimization is with respect to y and e. Equivalently, this means that dc is orthogonal to bc and (1,..., 1). The above can be interpreted geometrically as follows. By definition, ac is a point on the line L(ac, bc). Similarly, ec = ec(1,..., 1) is a point on the diagonal line that goes through (0,..., 0) and (1,..., 1) in K, i.e. L'. Minimizing dc according to (4) is the same as finding the shortest distance between the line L and the diagonal line, which is also the distance between the two points ac and ec. From this geometrical interpretation it is also clear that in order for the parameters to be uniquely identifiable the line L must not be parallel to the diagonal line, that is, must be different from for some k.
A robust estimate of L was proposed in [2], using iteratively re-weighted principal component analysis (IWPCA). This estimate of L, together with the above parametrization of ac and bc, give us estimates and of all 2K - 2 parameters of ac and bc, as well as an estimate of ec.
Let us illustrate the parametrization and estimation procedure for K = 2. Since two (non-parallel) lines will always intersect, constraint (4) degenerates to the assumption that dc = 0 or, equivalently, that = ec In the noise-free case the line L is described by
where and . By setting in (5) and applying the constraint , we get that ac = (ec, ec) and bc = (1, βc) where ec = αc/(1 - βc). To further illustrate the stability of the PMT independency, the parameters (ec, βc) have been estimated for each of the six PMT pairs independently based on data from array A scanned by the Axon scanner and quantified by GenePix. The various estimates for both channels are listed in Table 1. The average estimate of the bias across all PMT pairs in the red channel was = 18.0 (with standard deviation 1.12). For the green channel the average bias estimate was = 20.3 (with standard deviation 0.80). The small standard deviations confirm that dc is indeed small.
Results
This analysis was done with eight arrays (A-H) that were scanned on two different scanners at four different PMT settings. Two different image analysis applications were used. See Methods for details.
Parameter estimates
For every combination of array, scanner and signal quantification method (image analysis or raw pixel intensities), we estimated the parameters ac (including ec and dc), and bc in Model (2)-(4) for both channels (see Methods). To get a better understanding of the properties of the estimates, we used a non-parametric bootstrap approach to obtain not only bias corrected estimates, but also their standard deviations. Data was resampled over gene index in a way such that the same bootstrap data sets were used whenever pairwise comparison was done, e.g. when comparing bias estimates in red and green channels. For GenePix and Spot quantified signals a bootstrap sample of size 100 was used.
For the estimates based on the raw pixel intensities a different approach was taken. Because the number of pixels for one scan is about 107 (per channel) and we had four scans, our computer system limited us to estimate the model based on a subset of 106 pixel intensities. This was done for 100 random subsets and the mean and standard deviation of the parameter estimates were calculated, much like the bootstrap method above. The mean and the standard deviation of and for all possible setups are listed in Tables 2 &3. The mean and standard deviation of over all arrays are shown in Table 4.
Comparison of arrays
The bias estimates for all bootstrap replicates in Tables 2 &3 have been depicted as box plots in Figure 2. Considered that the signals are in [0, 65535], the bias estimates are very stable between different arrays. The biases span 9.8 units (0.15‰) in the red channel and 7.8 units (0.12‰) in the green channel.
Comparison of scanners
For the two scanners, we found that the estimated bias based on signals obtained by the Agilent scanner are consequently higher than the estimates from the Axon scanner. The box plots of their differences in the common bias ec (for each bootstrap sample) between the Agilent and the Axon scanner in Figure 3 confirm this divergence. See also Table 4. This significant difference could be an effect of scan order, that is, all arrays were first scanned on the Agilent scanner and then on the Axon scanner. The arrays in hand were part of a much bigger project based solely on Agilent scanned data. To keep a consistent scan protocol and to minimize bleaching, we could not balance the experimental design by letting some arrays be scanned in the reverse order. Instead, to test for scan order trends we scanned one array first on Agilent (H-1), then on Axon (H-1) and then again on Agilent (H-2). No apparent trend was found.
Comparison of image analysis methods
Estimates of the common bias ec based on GenePix quantified signals are consistently greater than the corresponding ones based on Spot signals, cf. Tables 2,3,4. The box plots in Figure 4 show differences in estimates of the common bias (for each bootstrap sample) between GenePix and Spot. The difference may be explained by the fact that the two applications use different spot segmentation algorithms [8,9]. Because the concentration of fluorophores is not homogeneous across a spot, the result is that the distribution of pixel intensities will vary with the segmentation method. This effect can be more profound for spots with strong donut effects. Robust estimates such as the median pixel value will to some extent protect against this variance, but not completely. It has been suggested [10] that the median of (pixel) ratios is a better estimate of the ratio of hybridized cDNA than the ratio of median (pixels). However, the former requires that the images are perfectly aligned with respect to shift, rotation, shear and so on. Also, it applies exclusively to two sample comparisons. Because of this, we do not believe that pixel-ratio signals are useful in practice.
Pixel-based estimates
To better understand the underlying reasons for the observed channel biases, the proposed affine model was also applied to pixel intensities (instead of spot signals). The estimated biases for the two channels for different arrays using IWPCA based on pixel values are shown in Tables 2 &3. Except for the green channel in the second scan round of Array H, the pixel-based estimates are consistently higher than the estimates based on GenePix and the Spot foreground signals. As noted above, pixel-based estimates are very sensitive to image distortions. This is especially a concern for the Agilent scanner since it reloads the arrays between subsequent scans. To investigate the effect of image distortion, we did a test where a person with experience in microarray analysis was asked to subjectively rank how badly aligned the four images in the red channel with different PMT gains from the Agilent scanner were for each of the (unlabeled) nine arrays. The person rated the images from Arrays A, B, D, and H-1 to be "extremely" misaligned. The images from Array E were considered to be "quite" misaligned, and the images from Array C to be "slightly" misaligned. For the rest of the arrays the images were considered to be aligned (less than a pixel off). This is perfectly in line with the discrepancies in Table 2, which confirms our hypothesis. Another disadvantages with pixel-based methods is that they are extremely memory and time consuming. For instance, estimation with 106 pixels is approximately 50 times slower than with 55,488 signals.
Comparison of channels
As Figure 5 shows, the common bias ec is greater in the green channel than in the red channel, especially for GenePix quantified signals, when estimated based on data from the Axon scanner. For the Agilent scanner this trend is less clear although the Spot quantified signals seem to give higher bias in the green channel than in the red channel. Furthermore, the biases in the red and the green channels are stable between arrays, which give further evidence to our hypothesis that the bias originates from the scanner (and/or the image analysis methods).
Deviation in bias estimates between PMT gains
In Figure 6 the distribution of the "bias residuals" are depicted for different scans k and channels c, for each separate array, but also for all arrays together, and for both scanners and both image analysis methods. Most apparent is that the estimates based on signals from the Axon scanner and especially those quantified by the Spot software are greater than for the others, cf. Tables 2 &3. The reason for this difference is not clear to us. For some arrays the estimates from the red and the green channels are strongly correlated, but it is not clear to us when this occurs. Although not in general, for some combinations of scanner and image analysis method, there is a trend in the PMT order (or possibly scan order). Again, we do not know why. To summarize, we have by means of exploratory data analysis (not shown) tried to understand what sometimes looks like patterns in the :s, but we found no apparent relationships. However, systematic effects indicate that may be modeled further.
Calibration
When data was calibrated according to the backward transformation in (8)-(10) estimates (up to a scale factor) of all xc,i:s were obtained. Since we do not know the true values we can not verify the estimates directly. However, partly we can do it indirectly by looking for remaining systematic effects in the log-ratios, but also by comparing the empirical densities of the calibrated scans. For a detailed study on systematic effects introduced by affine transformations, see [2]. For instance, the amount of intensity-dependent curvature in the log-ratios is related to the bias and the relative scale factor via the product assuming ||dc|| = 0. To demonstrate this relationship, we have for different PMT pairs compared the within-channel log-ratios and log-intensities
respectively, with the corresponding ones for the backward transformed data, which we denote by and . The log-ratios versus the log-intensities for the raw signals of all six PMT pairs are shown in the left scatter plot in Figure 7. The corresponding plot for the backward transformed signals is shown to the very right. For each of the six data clouds, the curvature, but also the overall bias, in the log-ratios is removed. To further underline the effect that a channel-specific bias has, we have calculated the log-ratios for the bias-subtracted signals (no rescaling), which makes Model (2) linear. As seen in the middle scatter plot, the curvature introduced by the bias and the logarithm is removed. The overall bias in the log-ratios which remains is and is removed when the signals are rescaled. It is not correct to shift only the log-ratios towards zero, because then the log-intensities will be incorrect.
The various M versus A plots become very similar and so do the four empirical density functions of the signals as seen in Figure 8. The small bumps at high intensities are due to the saturated signals, cf. Figure 7.
Extended dynamical range
For the Agilent scanner the effective scale parameters / were estimated to be in the order of approximately 1 : 3.5. For the Axon scanner they were in the order of approximately 1 : 40, cf. Table 1. Thus, the calibration method extends the effective dynamical range, with preserving linearity, by a factor of 3.5 for signals from the Agilent scanner and a factor of 40 for signals from the Axon scanner.
Discussion
Sources of the bias
Because bias introduced before the PMT would be amplified differently at different gains, we suspect that the observed bias is due to the scanner and most likely its detector parts such as the analogue-to-digital converter (ADC) after the PMT, but possibly also due to the image analysis method. The observed differences between the channels can be explained by the fact that there is one PMT and one ADC per channel, which may have slightly different properties. Although there are differences in bias between the two scanners, they are still of the same order, which we find remarkable. Another lab with a GenePix scanner reported biases also around 15–20 (personal communication). A possible reason for this is that the scanners consist of similar parts.
Other estimates
To rule out the obvious situation where all pixel intensities are biased we compared the above estimates with the minimum pixel intensities. For example, for Array A (scanned on Axon and analyzed with GenePix Pro), the minimum pixel intensities in the red channel were 9, 0, 8, and 9 for PMT 500, 600, 700 and 800 volts, respectively. In the green channel the minimum pixel intensity is 0 for all scans. It is not useful to use the minimum spot signals, , either. For example, for the above scan the average minimum signal across all scans in the red channel is 19.8 (median 19.5, std. dev. 0.96), but in the green channel it is 34.8 (median 28.0, std. dev. 19.6), cf. Table 3.
On background subtraction
If the scanner is the main source for the observed bias, then the background estimates should be affected by this bias as well and subtracting the background from the foreground estimates will therefore not only correct for physical background noise from the array itself, but also for the scanner bias. The strong intensity dependent effects of the log-ratios that are due to the bias, are much less apparent if we apply background subtraction (not shown), giving more evidence to our hypothesis that the observed systematic effects originate from the scanner. Thus doing background correction might correct for the bias, but it will also introduce more noise at any given intensity. Also, for the data set in hand background subtraction results in 4050 (7.3%), 6237 (11%), 7015 (13%) and 7349 (13%) negative values (in either channel), respectively, whereas bias subtraction results in no negative values. If we assume that the noise is additive such that the background is added to the foreground signals, then for probes with few or no fluorescent molecules the true foreground signal should be close or identical to the true background signal. As both are estimates, approximately half of the foreground signals for non-signal spots are less or equal to the corresponding background signals. Thus, about half of such spots results in negative signals. However, the different numbers of negative signals for different PMT voltages suggest that this can not be the full explanation. One reason could be that the background estimates are likely to be biased [9]. An error model that incorporates different noise sources, but also different scan parameters, might give some answers to this problem. Some models in this context have already been suggested [3,7,11], as well as models based on empirical Bayesian methods [12]. Another way to put it is that the background estimate is local and based on individual spots/pixels whereas the bias estimate is global, that is, there is one estimate for the whole array (although local estimation of bias is possible). Therefore, the background subtracted intensity estimates are noisier, resulting in more negative estimates for low intensity spots.
The problem of non-positive estimates, but also high variance close to zero, are limitations of the logarithmic transform and alternatives such as the generalized logarithmic transform etc. have been suggested [7,13,14].
Photo bleaching
We estimated the red dye (Cy5™) to bleach about 2% and the green dye (Cy3™) about 1% in a typical microarray experiment (not shown). Because the amount of bleaching is relatively small, but also because it is a very complex phenomenon, we decided to not try to incorporate it in the above model. Some of the systematic variation seen in the bias estimates for the different PMT settings may be due to bleaching.
Signal density normalization
As the results show, the empirical distributions of signals match each other remarkably well after calibration. It is interesting to compare this method with the quantile normalization methods proposed by [15-17]. The latter is based on the "statistical" assumption that the signals in all channels (scans) should be equal whereas the former is based on a "physical" assumption that the signals should be linear in the dynamical range. For a further discussion on this see [2].
Incremental robust estimates
It turned out to be infeasible to estimate the model parameters based on all pixel intensities, which limited us to use only on a 10% subset of data. As argued above, pixel-based estimates are not reliable and therefore not of interest. However, for spot-based estimates the same limitations may apply as larger data sets are made available. We wish to overcome such memory constraints. For this reason, we investigate the possibility to use (approximative) incremental re-weighted PCA methods [18,19].
Related work
Another method that combines multiple scans is the masliner (Microarray Spot LINEar Regression) algorithm [20]. It works by combining one low-PMT scan and one high-PMT scan into a new virtual scan. If a signal in the high-PMT scan is within a specified linear range its value is used, otherwise the corresponding signal from the low-intensity range is used after being transformed affinely to fit the high-PMT scan. To combine three or more scans, the new virtual scan can be combined with another PMT scan and so on. The result is that the effective dynamical range is extended. However, there are several unnecessary drawbacks. First, although several observations of the same spot concentration exist, which all may be within the dynamical range of the scanner, only one observation is used. Statistically, the average (calibrated) scan would be a more precise estimate. Second, since the scans are combined pairwise the estimate of the affine relationship between the scans is less robust. Third, although a sensitivity discussion is carried out in the supplementary materials, masliner fits the affine models in a non-robust fashion (in L2). Also, classical linear regression is used, which assumes no error in the explanatory variable. Since masliner makes the signals from different PMT settings proportional to each other it will indeed remove for instance curvature in within-channel M versus A scatter plots. However, masliner does not model the possibility of a PMT-independent bias and will therefore not correct for it. We believe this is the reason why the authors observe a "curvilinear effect" [[20], supplementary material]. For these reasons, we believe that the robust multiscan calibration method presented in this paper is superior to the masliner algorithm and should be used instead.
Conclusions
By scanning the same microarray at various PMT settings we have shown that there exists a bias in the measurement of the concentration of fluorescent molecules in the spots on the microarray. Our analysis indicates that this bias is mainly due to the scanner, but also due to the image analysis methods. By using a constrained affine model for the relationship between the obtained fluorescent intensities and fluorophore concentrations in the spots, we have been able to estimate the aforementioned bias. With estimates of the bias and scale parameters in each channel back transformation gave estimates of the amount (up to a scale factor) of photons from each spot that enters the PMT. Although not all photons originate solely from fluorophores in the target DNA, this is still a far better estimate of the amount of hybridized target DNA in each spot than the corresponding signal quantified by the scanner and the image analysis.
Before calibration, our data show a strong intensity dependent effect in the log-ratios, whereas after calibration there is no apparent intensity dependent trend. Furthermore, the distributions of signals from subsequent PMT scans are almost identical after calibration. In addition, the signal-to-noise ratio is increased with multiple scans. Finally, scanning at both low and high PMT settings extends the dynamical range of data, which gives higher resolution at low intensities without having to pay the price of saturated signals.
The proposed method can be applied to other microarray technologies such as single-channel oligonucleotide arrays or nylon arrays, and possibly to other gene expression technologies such as quantitative real-time polymerase chain reaction (QRT-PCR).
To conclude, we suggest that hybridized microarrays are scanned at two (preferably more) PMT gain levels to identify channel dependent bias terms. Knowing the exact PMT settings is not important, but the larger the differences are, the more precise the estimates will be. We recommend that the scans are done in decreasing PMT-gain order (although we did not do so here). Given estimates, data can then be calibrated easily. For practical reasons it might, however, be sufficient to estimate bias terms for a specific scanner once and then use estimates for calibration of subsequent microarrays. The small inter-array variation observed for channel specific bias in our data suggests that this would be possible. On the other hand, without multiple scans, afore mentioned increase in signal-to-noise and dynamical range will be lost. Also, not investigated within the scope of this study, bias terms for a specific scanner might change over time. For these reasons, we suggest that microarrays are scanned multiple times.
For two-channel microarrays, after calibrating each channel separately, a similar strategy can be applied once more to bring differently labeled channels to the same scale as suggested in [2]. This would rely on the assumption that the amounts of hybridized DNA in all channels are approximately equal for the majority of the spots, which in turn is based on the commonly used assumption that most genes are non-differentially expressed. This also applies to normalization between arrays.
All necessary methods are made available in a free R package named aroma [21]. A typical usage is calibrateMultiscan(rg) where rg is the object containing the red and green signals. In addition, we are currently implementing the methods as a plug-in module for the BASE system [22].
Methods
Arrays and hybridization
The analysis was based on eight different hybridizations of spotted oligonucleotide microarrays (A-H). Arrays A and B were hybridized in October 2003. Arrays C-G were hybridized the following day and Array H was hybridized seven weeks later. All arrays contain the same human oligonucleotide set (QIA GEN) and all have an identical layout of 12-by-4 print-tip groups each containing 34-by-34 (1156) spots. In total there are 55488 spots on each array. The average (GenePix) spot area is 45–50 pixels and the average center-to-center distance between the spots is approximately 12–13 pixels (120–130 μm). Arrays were produced by the SWEGENE DNA Microarray Resource Centre, Department of Oncology at Lund University using a MicroGrid II 600R arrayer fitted with MicroSpot 10 K pins (BioRobotics). All arrays except Array H were spotted in the same print batch on UltraGAPS™ coated slides (Corning Incorporated) during August 2003. Array H was spotted in October the same year. Printing was performed in a temperature (18–20°C) and humidity (44–49% RH) controlled area. After printing was completed, arrays were left in a desiccator to dry for 48 hours, rehydrated for 1 second over steaming water, snap dried on a hot plate (98°C), UV-cross-linked (800 mJ/cm2) and subsequently hybridized with various test and reference RNA samples. Samples were labeled, purified and hybridized using Pronto!™ Plus System 6 (Corning Incorporated) according to manufacturer's instructions.
Scanning
Each array was scanned at four different PMT settings on two different types of scanners. First the arrays were scanned on an Agilent G2505A DNA microarray scanner (Agilent Technologies) at PMT gains 100%, 30%, 50%, and 80% (in that order). The so called dark offset intentionally added to all signals by the Agilent scanner [[23], p. 18] has been uninstalled. Arrays were then re-scanned on an Axon GenePix 4000 A scanner (Axon Instruments) at PMT gains 600, 700, 800, and 500 volts (in that order), except for Array A, which was scanned at 700, 800, 500 and 600 volts, and Array H, which was scanned at 600, 400, 500 and 700 volts. Thus, the images obtained by the Axon scanner were bleached more than the preceding ones obtained by the Agilent scanner. For both scanners the power of the 532 nm and the 635 nm lasers was set to 100% and the scan resolution to 10 μm/pixel. Moreover, a one-pass (both channels scanned simultaneously) and one-sample-per-pixel ("lines to average" equals one) procedure was used. The Agilent scanner has a special loading mechanism for microarrays, which allows automatic scanning of subsequent arrays without human intervention. However, due to limitations in the software or the scanner, each batch of arrays can only be scanned at a single PMT gain. To scan at more PMT gains with the Agilent scanner, it was therefore necessary to eject and reload the arrays between different settings, which means that the alignment between the scanned images may not be perfect. Contrary, for the Axon scanner the arrays were put in the scanner one by one, then scanned at all PMT settings without being moved.
Image analysis – spot segmentation and registration
To quantify the foreground and the background signals, the scanned images (65536 gray scales and approximately 2000-by-5600 pixels) were analyzed using both the Axon GenePix Pro v4.1.1 software (Axon Instruments) and the Spot v2 software [8,24]. We first analyzed each image with GenePix. For each of them, the grid and spot positions were manually set and then the alignment was optimized by GenePix. These positions were then re-entered and re-optimized by Spot with visual inspection to verify the correctness. Moreover, for each individual scan the image analysis software was let to find the optimal spot segmentation. Thus, what is defined as a foreground pixel may vary with PMT setting although the images are from the same array. We decided on this schema for various reasons. The first reason was that the Agilent arrays are loaded and unloaded between subsequent scans and therefore require a separate spot segmentation. To be able to compare the results from the Axon and the Agilent scanner we choose the same procedure for the images scanned on the Axon scanner, even though, the optimized segmentation for the strongest image could have been reused. We further believe that this allows us compare Spot and GenePix more fairly.
For both Spot and GenePix the median spot pixel intensity was used as foreground signal. Background estimates were not considered in this analysis. No spot signals were discarded.
Calibration
Given estimates of and data can be calibrated using backward transformation. Let
be the backward transformed observed signal and the rescaled error terms, respectively. The affine Model (2) can then be rewritten as
Moreover, let
be the average backward transformed signal for gene i in channel c. Now, if , then
when all and are known. Thus, if (8) is applied with estimates of and that are consistent as I → ∞, and the error terms have zero mean, the mean of the backward transformed signals will converge to xc,i as I grows. Even though is not observable, we can estimate it consistently by increasing the number of scans K. Inspection of the residuals of calibrated signals (not shown) indicates that the variance of the calibrated noise is independent of PMT setting, that is . Assuming independent noise terms, the variance of the sample mean decrease with K as
In summary, we obtain consistent estimates (up to a multiplicative constant) of all xc,i with increasing I and K.
Finally, signals that are saturated by the scanner have to be excluded before calculating the average. If the quantified signal for a spot happens to be saturated in all scans, then that spot is marked as saturated, which still may be informative when compared to other non-saturated signals.
Data analysis
All further analysis was carried out using R [25,26] and the aroma package (f.k.a. com.braju.sma) [21]. All methods used can be found in the latter.
Authors' contributions
GJ and JVC carried out the practical microarray laboratory work and the scanning of hybridized arrays. Image analysis using GenePix software was carried out by JVC whereas HB carried out image analysis using Spot software. HB performed all the statistical analysis and conceived the constrained affine model used to identify and estimate channel-specific bias in microarray data. All authors participated in the design of the study and approved the final manuscript.
Acknowledgments
We thank Professor Ola Hössjer at Mathematical Statistics at Stockholm University and Halfdan Grage at The Centre for Mathematical Sciences, Lund University for their valuable feedback on the methods and the manuscript. We also thank Professor Åke Borg at Department of Oncology, Lund University for providing microarray data and Assistant Professor Markus Ringnér at Department of Theoretical Physics Lund University for proof reading the manuscript. This work was supported by grants from the Swedish Foundation for International Cooperation in Research and Higher Education (STINT), Fulbright Commission, Foundation Blanceflor Boncompagni-Ludovisi née Bildt, Royal Swedish Academy of Sciences, Royal Physiographic Society in Lund, Swedish Cancer Society, Ingabritt and Arne Lundberg's Research Foundation, and American Cancer Society. Microarrays were produced by the SWEGENE DNA Microarray Resource Center at Lund University, supported by the Knut and Alice Wallenberg foundation through the SWEGENE consortium.
Figures and Tables
Figure 1 Affine relationship between quantified fluorescent intensities and concentration of fluorophores. Scanning the arrays at different PMT gains indicates that there is an affine relationship between quantified fluorescent intensities and concentration of fluorophores. Left: Observed signals in the green channel for PMT pairs (800, 500), (700, 500), (800, 600), (600, 500), (700, 600), and (800, 700) are shown in green, red, cyan, black, blue and magenta, respectively. An affine model is estimated for each pair and displayed as a line. Data points at the very top are saturated and ignored. Right: A zoom-in of the same data. For each pair the estimated , which corresponds to the true origin, has been highlighted with a circle. All lines seem to intersect at the same point on the line. Shown are signals from the green channel on Array A quantified by GenePix from Axon scanner images.
Figure 2 Bias estimated for each array and channel on both scanners Estimated biases ec for each array and for all arrays together from the Agilent scanner (left column) and the Axon scanner (right column). The top four graphs are for the red channel and the bottom four are for the green channel. For each channel the top two are estimates based on signals quantified by GenePix and the bottom two on signals quantified by Spot. See also Table 2 & 3.
Figure 3 Comparison of bias estimates between scanners. Differences in the common biases ec between the Agilent and the Axon scanners for each separate array and for all arrays together. The bias is significantly larger for Agilent, regardless of channel (top and bottom) and image analysis method (left and right).
Figure 4 Comparison of bias estimates between image analysis methods. Differences in the common biases ec between the GenePix and the Spot image analysis method for each separate array and for all arrays together. The bias is significantly larger for GenePix, regardless of channel (top and bottom) and scanner (left and right).
Figure 5 Comparison of bias estimates between channels. Differences in the common biases ec between the red and the green channel for each separate array and for all arrays together. At the top is data from the Agilent scanner and at the bottom from the Axon scanner. Data in the left column by GenePix and Data in the right column was quantified by Spot. For data from the Axon scanner the common bias is greater in the green channel than the red channel, especially for GenePix quantified signals. For the Agilent scanner this trend is less clear although the Spot quantified signals clearly seem to have a higher bias in the green than the red channel.
Figure 6 Bias residual estimates for all arrays. Estimated bias residuals for each array and for all arrays together from the Agilent scanner (left column) and the Axon scanner (right column). For each array the distribution of the four of are shown in increasing PMT order. The top four graphs are for the red channel and the bottom four are for the green channel. For each channel the top two are estimates based on signals quantified by GenePix and the bottom two on signals quantified by Spot.
Figure 7 Log-ratio and log-intensity plots of raw, translated, and calibrate signals. The affine transformation gives curvature in the M versus A plots, which is corrected for by the affine normalization method. The three scatter plots show the within-channel log-ratio versus the log-intensity for each of the six PMT pairs based on the same data as in Figure 1. Left: Observed signal for different PMT pairs. For each pair the estimated has been marked with a circle, cf. Figure 1. Mid: Bias corrected signals. Right: Bias and scale calibrated signals. The range of the M axis is twice the range of the A axis so that (6)-(7) appear as a rotation in the plots.
Figure 8 Densities of the logarithm of the raw, translated, and calibrate signals. The affine normalization method makes the signal densities much more similar. Left: Density plots of the logarithm of the raw signals for each of the fours scans. Mid: Bias corrected signals. Right: Bias and scale calibrated signals. Data and colors are the same as in Figure 1.
Table 1 Pairwise parameter estimates. Red and green channel estimates of the bias and the slope for each PMT voltage pair together with the mean and the standard deviation of all estimates. The small standard deviation is evidence that the bias to a large extent is independent of the PMT settings. Data shown originates from Array A scanned on Axon and analyzed with GenePix.
PMT pair
(800, 500) 18.9 32.4 20.7 32.5
(700, 500) 19.0 13.0 20.9 12.8
(800, 600) 17.3 7.1 19.8 7.9
(600, 500) 19.0 4.5 20.8 4.1
(700, 600) 17.6 2.9 20.9 3.1
(800, 700) 16.3 2.5 18.9 2.6
mean 18.0 ± 1.12 20.3 ± 0.80
Table 2 Red channel parameters estimates. Bootstrapped parameter estimates for the red channel with standard deviations for each combination of array, image method (or pixel intensities) and scanner.
array image Agilent Axon
method
A Spot 17.9 ± 0.289 1.14 ± 0.148 15.8 ± 0.090 4.40 ± 0.145
A GenePix 19.8 ± 0.253 0.82 ± 0.162 18.1 ± 0.058 1.27 ± 0.095
A pixels 44.9 ± 0.036 22.85 ± 0.043 19.0 ± 0.032 0.68 ± 0.052
B Spot 19.2 ± 0.255 1.05 ± 0.165 15.5 ± 0.121 4.01 ± 0.255
B GenePix 21.2 ± 0.225 1.83 ± 0.203 17.3 ± 0.114 1.72 ± 0.253
B pixels 42.8 ± 0.049 20.71 ± 0.051 17.8 ± 0.034 2.06 ± 0.079
C Spot 17.6 ± 0.366 2.00 ± 0.192 16.3 ± 0.076 4.83 ± 0.191
C GenePix 19.8 ± 0.284 0.94 ± 0.209 17.9 ± 0.034 2.11 ± 0.135
C pixels 21.0 ± 0.138 1.48 ± 0.079 18.8 ± 0.022 1.24 ± 0.061
D Spot 18.2 ± 0.301 1.56 ± 0.103 15.3 ± 0.093 4.10 ± 0.149
D GenePix 21.1 ± 0.274 0.95 ± 0.139 17.2 ± 0.049 1.74 ± 0.090
D pixels 25.0 ± 0.522 4.10 ± 0.358 18.3 ± 0.025 1.02 ± 0.044
E Spot 18.0 ± 0.428 1.15 ± 0.109 16.1 ± 0.101 3.01 ± 0.131
E GenePix 22.1 ± 0.268 1.77 ± 0.223 17.7 ± 0.064 1.02 ± 0.096
E pixels 24.7 ± 0.144 5.51 ± 0.169 18.7 ± 0.024 0.39 ± 0.025
F Spot 19.9 ± 0.423 0.48 ± 0.163 16.4 ± 0.087 3.76 ± 0.138
F GenePix 23.8 ± 0.316 1.47 ± 0.258 18.5 ± 0.060 1.07 ± 0.106
F pixels 24.2 ± 0.131 1.37 ± 0.101 19.3 ± 0.026 0.43 ± 0.038
G Spot 16.0 ± 0.300 2.30 ± 0.166 15.5 ± 0.077 3.54 ± 0.114
G GenePix 18.3 ± 0.208 1.88 ± 0.198 17.3 ± 0.070 1.57 ± 0.134
G pixels 19.1 ± 0.096 1.48 ± 0.080 18.3 ± 0.026 0.66 ± 0.038
H-1 Spot 20.4 ± 0.161 2.12 ± 0.073 17.9 ± 0.025 1.35 ± 0.043
H-1 GenePix 22.1 ± 0.124 2.41 ± 0.097 18.7 ± 0.024 0.81 ± 0.041
H-1 pixels 44.9 ± 0.513 17.68 ± 0.528 19.1 ± 0.013 0.78 ± 0.027
H-2 Spot 20.1 ± 0.141 0.33 ± 0.040 N/A N/A
H-2 GenePix 21.8 ± 0.087 0.22 ± 0.084 N/A N/A
H-2 pixels 21.8 ± 0.047 0.26 ± 0.042 N/A N/A
Table 3 Green channel parameters estimates. Bootstrapped parameter estimates for the green channel with standard deviations for each combination of array, image method (or pixel intensities) and scanner.
array image Agilent Axon
method
A Spot 20.2 ± 0.106 0.51 ± 0.058 16.6 ± 0.165 6.31 ± 0.302
A GenePix 21.8 ± 0.068 0.20 ± 0.045 20.4 ± 0.123 0.88 ± 0.242
A pixels 33.6 ± 0.050 10.52 ± 0.048 22.8 ± 0.048 1.57 ± 0.093
B Spot 19.6 ± 0.094 0.90 ± 0.055 17.2 ± 0.127 2.20 ± 0.231
B GenePix 21.3 ± 0.046 0.94 ± 0.086 20.1 ± 0.152 1.20 ± 0.326
B pixels 35.7 ± 0.153 12.85 ± 0.146 21.9 ± 0.041 2.04 ± 0.087
C Spot 19.8 ± 0.050 1.21 ± 0.073 17.0 ± 0.127 2.75 ± 0.254
C GenePix 20.8 ± 0.039 0.99 ± 0.086 19.4 ± 0.100 1.11 ± 0.218
C pixels 21.3 ± 0.024 0.78 ± 0.035 22.0 ± 0.040 2.39 ± 0.082
D Spot 19.4 ± 0.097 1.49 ± 0.073 17.0 ± 0.194 1.24 ± 0.376
D GenePix 20.9 ± 0.059 0.80 ± 0.095 20.3 ± 0.126 2.31 ± 0.286
D pixels 21.6 ± 0.105 0.21 ± 0.124 22.1 ± 0.046 3.10 ± 0.092
E Spot 18.2 ± 0.121 3.29 ± 0.129 16.5 ± 0.132 2.88 ± 0.240
E GenePix 20.2 ± 0.081 1.92 ± 0.128 19.2 ± 0.137 0.74 ± 0.264
E pixels 21.2 ± 0.017 0.13 ± 0.046 21.0 ± 0.032 0.94 ± 0.058
F Spot 21.4 ± 0.125 0.58 ± 0.103 16.5 ± 0.181 4.79 ± 0.327
F GenePix 23.7 ± 0.084 1.32 ± 0.116 20.2 ± 0.132 0.34 ± 0.199
F pixels 24.1 ± 0.019 1.57 ± 0.026 22.4 ± 0.039 1.87 ± 0.071
G Spot 18.9 ± 0.088 1.62 ± 0.088 16.9 ± 0.149 3.72 ± 0.258
G GenePix 20.4 ± 0.049 1.10 ± 0.088 20.0 ± 0.164 0.84 ± 0.262
G pixels 21.0 ± 0.017 0.75 ± 0.024 22.2 ± 0.045 2.60 ± 0.085
H-1 Spot 21.2 ± 0.134 3.25 ± 0.080 18.5 ± 0.090 2.00 ± 0.170
H-1 GenePix 22.6 ± 0.100 3.47 ± 0.107 20.3 ± 0.043 0.63 ± 0.100
H-1 pixels 42.8 ± 1.417 16.68 ± 1.843 21.3 ± 0.033 1.82 ± 0.070
H-2 Spot 19.6 ± 0.229 0.16 ± 0.077 N/A N/A
H-2 GenePix 21.7 ± 0.104 0.55 ± 0.076 N/A N/A
H-2 pixels 21.5 ± 0.086 0.29 ± 0.021 N/A N/A
Table 4 Mean and standard deviation of bias estimates. Mean and standard deviation of the bias estimates of all arrays and for each signal quantification method. The median and MAD (median absolute deviation) of ditto gives very similar results expect for Agilent's pixel estimates, which become smaller but are still greater than the others.
image
method Agilent Axon Agilent Axon
Spot 18.6 ± 1.40 16.1 ± 0.792 19.8 ± 0.958 17.0 ± 0.633
GenePix 21.1 ± 1.55 17.8 ± 0.529 21.5 ± 1.05 20.0 ± 0.433
pixels 29.8 ± 10.4 18.7 ± 0.446 27.0 ± 7.76 22.0 ± 0.537
==== Refs
Duggan DJ Bittner M Chen Y Meltzer P Trent JM Expression profiling using cDNA microarrays Nature Genetics 1999 21 10 14 9915494 10.1038/4434
Bengtsson H Hössjer O Methodological study of affine transformations of gene expression data with proposed normalization method. Preprints in Mathematical Sciences 2003:38, Mathematical Statistics, Centre for Mathematical Sciences, Lund University, Sweden 2003
Cui X Kerr MK Churchill GA Data Transformations for cDNA Microarray Data Tech rep, The Jackson Laboratory, USA 2002
Burle Industries Inc Photomultiplier Handbook Lancaster, PA, U S A 1980
Dudoit S Yang YH Callow MJ Speed TP Statistical methods for identifying differentially expressed genes in replicated cDNA microarray experiments Technical Report 578, Department of Statistics, University of California at Berkeley 2000
Yang YH Dudoit S Luu P Speed TP Bittner ML, Chen Y, Dorsel AN, Dougherty ER, San Jose Normalization for cDNA microarray data In Proceedings of SPiE, Volume 4266 of Microarrays: Optical Technologies and Informatics 2001 California: The International Society for Optical Engineering 141 152
Huber W von Heydebreck A Sültmann H Poustka A Vingron M Variance stabilization applied to microarray data calibration and to the quantification of differential expression Bioinformatics 2002 18 S96 104 12169536
Yang YH Buckley M Dudoit S Speed T Comparison of methods for image analysis on cDNA microarray data Technical Report 584, Department of Statistics, University of California at Berkeley 2000
Bengtsson A Microarray Image Analysis: Background Estimation using Region and Filtering Techniques Master's Theses in Mathematical Sciences, Mathematical Statistics, Centre for Mathematical Sciences, Lund Institute of Technology, Sweden 2003
Brody JP Williams BA Wold BJ Quake SR Significance and statistical errors in the analysis of DNA microarray data Proc Natl Acad Sci U S A 2002 99 12975 12978 12235357 10.1073/pnas.162468199
Rocke DM Durbin B A Model for Measurement Error for Gene Expression Arrays Journal of Computational Biology 2001 8 557 569 11747612 10.1089/106652701753307485
Kooperberg C Fazzio TG Delrow JJ Tsukiyama T Improved background correction for spotted DNA microarrays Journal of Computational Biology 2002 9 55 66 11911795 10.1089/10665270252833190
Hawkins DM Diagnostics for conformity of paired quantitative measurements Statistics in Medicine 2002 21 1913 1935 12111897 10.1002/sim.1013
Durbin B Hardin J Hawkins D Rocke D A variance-stabilizing transformation for gene-expression microarray data Bioinformatics 2002 18 S105 S110 12169537
Bolstad B Irizarry R Åstrand M Speed T A comparison of normalization methods for high density oligonucleotide array data based on variance and bias Bioinformatics 2003 19 185 93 12538238 10.1093/bioinformatics/19.2.185
Yang YH Thorne NP Goldstein DR Normalization for Two-color cDNA Microarray Data In Science and Statistics: A Festschrift for Terry Speed, Volume 40 of Monograph Series 2003 IMS Lecture Notes 403 418
Thorne NP Single-channel normalisation and analysis of two-colour cDNA microarray data PhD thesis, Department of Medical Biology and Department of Mathematics and Statistics, The Walter and Eliza Hall Institute of Medical Research 2004
Skocaj D Leonardis A Weighted and robust incremental method for subspace learning In ICCV03 2003 1494 1501
Li Y Xu LQ Morphett J Jacobs R An Integrated Algorithm of Incremental and Robust PCA In IEEE International Conference on Image Processing (ICIP2003) 2003
Dudley AM Aach J Steffen MA Church GM Measuring absolute expression with microarrays with a calibrated reference sample and an extended signal intensity range Proc Natl Acad Sci U S A 2002 99 7554 9 12032321 10.1073/pnas.112683499
Bengtsson H aroma – An R Object-oriented Microarray Analysis environment Preprint in Mathematical Sciences 2004:18, Mathematical Statistics, Centre for Mathematical Sciences, Lund University, Sweden 2004
Saal LH Troein C Vallon-Christersson J Gruvberger S Borg Å Peterson C BioArray Software Environment (BASE): a platform for comprehensive management and analysis of microarray data Genome Biol 2002 3 SOFTWARE0003 12186655 10.1186/gb-2002-3-8-software0003
Agilent Technologies, Inc. Agilent G2565AA and Agilent G2565BA Microarray Scanner System – User Manual third, Palo Alto, CA 2002 [G2566-90007]
CSIRO Australia Spot user's manual CSIRO Mathematical and Information Sciences, Image Analysis Group 2003
R Development Core Team R: A language and environment for statistical computing R Foundation for Statistical Computing, Vienna, Austria 2003
Ihaka R Gentleman R R: A Language for Data Analysis and Graphics Journal of Computational and Graphical Statistics 1996 5 299 314
| 15541170 | PMC539274 | CC BY | 2021-01-04 16:02:47 | no | BMC Bioinformatics. 2004 Nov 12; 5:177 | utf-8 | BMC Bioinformatics | 2,004 | 10.1186/1471-2105-5-177 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-5-1801555506010.1186/1471-2105-5-180Research ArticlePerformance of a genetic algorithm for mass spectrometry proteomics Jeffries Neal O [email protected] Office of the Clinical Director, National Institute of Neurological Disorders and Stroke, Bethesda MD, USA2004 19 11 2004 5 180 180 29 7 2004 19 11 2004 Copyright © 2004 Jeffries; 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
Recently, mass spectrometry data have been mined using a genetic algorithm to produce discriminatory models that distinguish healthy individuals from those with cancer. This algorithm is the basis for claims of 100% sensitivity and specificity in two related publicly available datasets. To date, no detailed attempts have been made to explore the properties of this genetic algorithm within proteomic applications. Here the algorithm's performance on these datasets is evaluated relative to other methods.
Results
In reproducing the method, some modifications of the algorithm as it is described are necessary to get good performance. After modification, a cross-validation approach to model selection is used. The overall classification accuracy is comparable though not superior to other approaches considered. Also, some aspects of the process rely upon random sampling and thus for a fixed dataset the algorithm can produce many different models. This raises questions about how to choose among competing models. How this choice is made is important for interpreting sensitivity and specificity results as merely choosing the model with lowest test set error rate leads to overestimates of model performance.
Conclusions
The algorithm needs to be modified to reduce variability and care must be taken in how to choose among competing models. Results derived from this algorithm must be accompanied by a full description of model selection procedures to give confidence that the reported accuracy is not overstated.
==== Body
Background
When Petricoin et al. [1] published their analysis using serum to distinguish individuals with ovarian cancer from individuals with benign conditions, it suggested great promise in using high throughput mass spectrometry to improve upon existing biomarkers for patient groups that could greatly benefit from accurate and early diagnosis. Results from the analyses using this algorithm have, along with early findings from other groups (e.g. [2,3]), fueled an explosion of interest in using mass spectrometry techniques for quick and accurate diagnosis. Many other investigators have since used classification techniques to achieve impressive results in correctly categorizing unlabeled mass spectrometry samples as either diseased or healthy. Among the more common classification methods used are classification trees [2], boosting [4], stepwise discrimination methods [5], and wavelet discrimination [6], though few have used a genetic algorithm. Baggerly et al. [7] do use a genetic algorithm though its properties are substantially different than that used in the earlier studies and is generally not subject to the conclusions drawn below. Here we evaluate the performance of a genetic algorithm described in the original Lancet paper and subsequent studies [8,9].
Description of the algorithm
The specifics of the genetic algorithm used here are based on webpages at the NCI-FDA website housing the data [10] and [1,9]. As the descriptions are limited, this attempt to reproduce the algorithm may differ from the implementation supporting the published results.
• As described, the genetic algorithm (GA) seeks to find a collection of markers that separate cases and controls. Here the markers correspond to a biological sample's measurements at a given set of m/z values. In the GA framework such a collection of markers is called a chromosome. Each chromosome is evaluated by a fitness function in the following way. Suppose a given chromosome is composed of N "genes" (i.e. N m/z values in this case). Each sample's intensity values at the N genes are linearly scaled to lie between 0 and 1; the smallest of the N intensities is assigned 0, the largest assigned 1, and intermediate values interpolated in a linear way. The first sample is assigned its own cluster with centroid (i.e. mean values) given by its N values. The next sample is compared to the first and if the Euclidean distance between the two samples exceeds .1· then the second sample is assigned a new cluster with centroid given by its values. If the distance is less than this limit the second sample is assigned to the first cluster and the centroid values recalculated as the mean of both cases. Subsequent cases are handled similarly – if a sample's representation as an N dimensional point lies within the .1· limit of a cluster the case is assigned to the closest cluster and centroid values are recalculated. If the smallest distance to any cluster centroid exceeds the limit then the case is assigned a new cluster. When all the cases are clustered a cluster's type is designated by majority vote – those clusters composed mostly of cancer cases are deemed cancer clusters and those composed mostly of nondiseased are likewise defined. The fitness function then computes the chromosome's fitness as its classification accuracy, i.e. the proportion of cases assigned to a cluster of the appropriate type.
• The selection process starts with 1500 randomly chosen chromosomes, i.e. sets of markers. The documentation indicates that chromosomes with length between 5 and 20 markers are used. In this implementation, different chromosomes may have different lengths – the 1500 are chosen to have a length between 5 and 20 with a uniform probability of 1/16 governing the choice. It is not clear how the original GA treated chromosomes of different length.
Each of the 1500 chromosomes is evaluated by the fitness function as described above. Chromosome pairs are then produced with the likelihood of being selected for a pair related to the fitness function. The available documentation does not make clear how this probability is explicitly related to the fitness. Here the choice was made by ranking the fitness scores and setting parameters α and β such that
Prob [ selecting kth ranked chromosome ] = α + βk
where α and β were chosen such that
This approach is described in section 2.1 of [11] – ranking is preferred to absolute fitness values as it avoids some potential problems with scaling. An alternative method based on absolute fitness values produced qualitatively similar findings (results not shown).
• A new generation of chromosomes is produced by first creating 750 parent pairs from the set of 1500 chromosomes. A parent pair is created by choosing two of the 1500 chromosomes using the above probabilities. A given chromosome can be chosen to be in more than one pair. For a given pair each chromosome is broken to produce two sub-chromosomes. The location of the break is random with uniform probabilities. The two sets of sub-chromosomes are then crossed-over to produce two new chromosomes. As an example suppose chromosome 1 has genes = (3001, 5500, 7800, 11011, 13059) and chromosome 2 is composed of m/z locations = (2500, 4200, 909, 15002) and the first chromosome breaks between its second and third elements while the second chromosome breaks after its third. Then the resulting new chromosomes are (3001, 5500, 15002) and (2500, 4200, 909, 7800, 11011, 13059). In this way the 750 pairs of chosen chromosomes produce a next generation of 1500 chromosomes. At this point each gene in each new chromosome may be randomly changed to any other gene in the entire spectrum range with probability .0002 (this corresponds to genetic mutation). In our implementation it is possible to match sub-chromosomes that would merge to be longer than 20 units. In this event the chromosome is truncated to 20. Further, here it is allowed to have chromosomes composed of as few as 2 markers as there seemed no compelling reason to impose a lower limit of 5 m/z values as described in the documentation. It is unclear if the original GA allowed chromosomes of different lengths to reproduce or instead restricted cross-overs to pairs with the same length.
• After mutation this new generation of 1500 chromosomes is then evaluated by the fitness test and then another round of selection, cross-over, and mutation processes produce the next generation. Typically, the average fitness of generations increases over time. According to the documentation, the process stops 1) after 250 generations, or 2) when a perfectly discriminating chromosome is found. In the first instance, that model that has produced the highest fitness score is chosen.
• Given a chosen model derived from a training set, an unlabeled spectrum (e.g. test set spectrum) is classified by determining that cluster with the centroid that lies closest to the unlabeled case and assigning the label of the cluster. The documentation describes a second, related approach that is to make this assignment only if the nearest centroid is with .1·; otherwise assign the case as of a third, unknown/new type. In this current work, classification errors for a test set correspond to the first criteria of nearest centroid, without the .1· requirement. Empirically, this led to greater classification accuracy.
Some concerns about this algorithm have been raised by others [7]; a few issues will be examined below in more detail.
• Each chromosome is evaluated by a fitness function that measures how well the chromosome classifies the training set. However it is clear that the order in which the cases are considered may make a difference in what cluster a case is assigned as well as the clusters' centroid values. Consequently, different results may arise in attempts to replicate findings.
• The GA algorithm starts with a random selection of 1500 chromosomes and then letting these evolve through a random mating process. As the initial selection and evolutionary process is random it is again the case that different ultimate models may be chosen, depending on the seed of a random number generator.
• In this application, each chromosome partitions the samples of the training set into clusters defined as groups of cases with centroids that are at least .1· apart from one another. It is not uncommon to find chromosomes that partition perfectly, but rely upon a large number of clusters (e.g. > 30). This suggests overfitting of data. As described, this algorithm does not penalize or otherwise take into account the number of clusters or length of chromosomes.
Largely because of insufficient information, this implementation of the algorithm likely differs from that supporting the published results. However, some elements should be the same. In particular, the evaluation of the fitness function should yield the same results except for the issue of how the order of the cases can change the clusters' attributes. Therefore, we should be able to match or come close to verifying the published results for a given model. However, our results are likely to be different as far as generating best models. This is in part due to the inherent randomness the process employs as well as possible differences in how the fitness function scores generate members of the next generation.
Datasets
Two publicly available datasets were used to evaluate the algorithm; information regarding them is available from an NCI-FDA website [10]. Both datasets consist of ovarian cancer patients and healthy controls. The first dataset, hereafter referred to as DS1, contains "low resolution" mass spectrometry data from a Ciphergen instrument and is identified on the NCI-FDA website as the 8-7-02 data. The data consist of 162 ovarian cancer samples and 91 control samples. The second dataset, DS2, contains "high resolution" data from a hybrid quadrupole time-of-flight spectrometer. Description and analysis of these data are available in [12]. The dataset contains spectra from 121 cancer samples and 95 controls.
For both datasets GA-produced models are presented on the NCI-FDA website that were developed from a training set and perfectly discriminate a test set. There is no designation as to which individuals were used for training and which for testing.
Results
The NCI-FDA website shows the chromosome consisting of m/z values {435.46, 465.57, 2760.67, 3497.55, 6631.70, 14051.98, 19643.41} was able to perfectly discriminate a test set drawn from the low resolution dataset, DSl. The 253 samples were randomly split into a training set of 81 cancer and 46 control individuals with the remainder forming a test set. Here we illustrate how the test set looks for these seven markers. Figure 1 shows the ratio of the second marker (molecular weight of 465.6) to the first (weight of 435.5) does an excellent job in separating the two types of samples in the test dataset. In the training set only two clusters were determined – one composed completely of cancer cases and the other solely of controls. In the test dataset one sample is misclassified (essentially because of its values on the remaining 5 markers) though it should be again pointed out that the number of misclassifications does vary by the order in which samples are processed and how the cases are split into test and training sets. Ten consecutive trials in which different test/training splits and ordering decisions were randomly made produced 1, 0, 0, 0, 0, 0, 0, 0, 5, and 1 errors (all misclassifications of normal as cancer) for this 7 marker model.
This exercise verifies that the model does quite well though it establishes that results do change with ordering and test/training set splits. Also, it confirms that results of 100% accuracy should be understood to depend upon the particular split and ordering. This may be of great importance when the goal is to develop tests with sensitivity and specificity exceeding 99% [12].
Next, the GA developed for this paper was then applied to these data with the expectation it should produce something like the 7 marker chromosome given above. After 7 generations, a chromosome was found that perfectly split the training set, but 7 clusters were required and 10 markers were used.
The graph of test set classification, Figure 2, shows a less compelling picture of discrimination; the third marker is perhaps best (based on t-test p-values) at distinguishing samples. This marker corresponds to a molecular weight of 831.1 Daltons. To examine robustness the algorithm was run 9 more instances using different initial sets of 1500 chromosomes to try to get a sense of the variation in the algorithm's chosen models. The same ordering, test samples, and training samples were maintained.
In each case a perfectly discriminating (i.e. training set error of 0%) chromosome was found within a few generations. Table 1 shows that there is considerable variation in the test set accuracies given they were all produced by the same data. Further, the best discriminating single marker within the chosen chromosome shows little consistency. Also, the set of 10 generally shows a large number of clusters and markers – nothing very similar to the published model that contained only two clusters.
The results in Table 1 suggest there are many markers that are different in these data and it is easy to find classifiers that performs well – at least in the training set. However, it is also clear this creates a kind of algorithmic instability in that considerable variation in results can arise from the same training set data. Even if one uses a fixed training/test set division there is now a question of how does one decide which results to use? One could run the algorithm just once, but run the risk of choosing a not very good model (e.g. the model with 24 clusters and 16 markers). However, if the algorithm is run many times in the search for a good model, the reported sensitivity and specificity in the test set are likely to be biased. This question will be pursued further in the Discussion section below.
Table 1 and the preceding discussion of variation suggest this implementation of the algorithm might be improved by changing the procedure to favor models with fewer clusters/markers. In this way, those models that may overfit the training data are penalized and the number of perfect discriminators of the training set consequently reduced. A simple way of doing this is to alter the fitness function to penalize large numbers of clusters and/or markers. In the analysis above the fitness function was given as
Fitness = Accuracy
= % Correctly classified cases.
This could now be modified to
Fitness = % Correctly classified cases
- p1·# of clusters
- p2·# of markers
where p1 and p2 are non-negative penalization weights.
A resampling method was used to determine the performance of different parameter combinations of p1 among {0, .002, .005, .008} and p2 in {0, .001, .002}. Specifically, random training samples (chosen without replacement) were selected from the entire set of samples so the original training sample size of 127 was maintained with 81 cancer and 46 control spectra. Then, for a given pair of p1 and p2 values the GA was trained on this pseudo-random training set and a model chosen. The remaining cases that were omitted from the training set (81 cancer and 45 control) were then treated as a test set. We repeat this procedure for 50 randomly chosen training sets and examine the distribution of the test set classification accuracy for the different parameter combinations. The same 50 training and test sets were used for each set of p1 and p2 combinations. In addition to illustrating the test set accuracy, the number of clusters, the number of markers associated with the different GA models, and the proportion of times (out of 50) a perfectly discriminating chromosome was found (for the training set) are also indicated.
This procedure gives a sense for the performance of the algorithm for different parameter combinations. Another question concerns performance if model selection were incorporated into the procedure. This was assessed in the following way. For each of the 50 training sets, an additional 5-fold cross-validation determined which of the 12 p1 and p2 combinations performed best in terms of predicting the omitted cases (in the event of ties the model associated with the most restrictive p1 and p2 was chosen with p1 the first tie-breaker). The chosen parameters were then used with the entire training set to develop a model that was evaluated on the associated test set. As before, this procedure was performed on the same 50 training and test sets. In the tables that follow, the results for this model are labeled as "Best GA". These results are perhaps most representative of overall performance for the algorithm developed here.
As a means of comparison two other classification schemes were applied to the same bootstrap samples. Boosting is a general method of combining a weighted set of classifiers that each "vote" on the class of a sample in question with majority vote dictating the set's aggregate classification. It has been successfully used in classification of mass spectrometry data [4,13]. Here, the base classifier is a simple threshold classifier, e.g. if intensity at mass 245.8 ≤ 47.5 then classify as cancer, otherwise classify as normal. The general process by which the set of base classifiers is chosen is discussed at length in [4] and [14]. Here 150 was chosen as the number of base classifiers for the aggregate classifier and the algorithm generally followed that outlined in section 10.1 of [14]. The second algorithm used was PAM (Predictive Analysis for Microarrays), a shrunken centroid method of classification [15]. This method has been used for high dimensional microarray studies and is relatively easy to implement. Both methods require little operator assisted tuning to obtain a small feature set – an important consideration when conducting so many resamplings. For these methods the data were normalized (test and training sets normalized separately for each resampling) so each spectra had the same average intensity. Also, attention was restricted to those m/z values showing Bonferroni-corrected differential expression (calculated anew for each resampling). Computer code and information regarding the parameters and details of these methods are available on a webpage [16] with supporting documentation.
Table 2 shows the 25th and 75th percentiles for test set accuracy among 50 samples as described above. The two penalty parameters have the desired effect in reducing the number of clusters/markers but there is relatively small variation over the different parameter combinations. The GA models apparently perform a bit better than PAM and a little worse than the boosting method though all models have high accuracy. Some reviewers of these data [5,17], have questioned why the groups are so easy to classify and whether the entire m/z range should be used. The criticism centers around the strong signals that are present in very low m/z values (e.g. 2.79 and 245.54 Daltons) that are speculated to be products of experimental procedures rather than reflective of biological differences. Other investigators [2] routinely exclude the lower end of the spectrum (less than 1500 or 2000 Daltons) as they feel it too contaminated by matrix and other effects to be clearly interpretable. As a result of these concerns the experiment was rerun with the m/z range restricted to be greater than 1500 Daltons.
Table 3 is based upon the m/z restricted dataset and shows evidence of greater spread among the different GA models – those with p1 = 0 or .008 do not appear to do as well as p1 = .002 or .005. With no penalty on the number of clusters one sees very high dimensional models (median number of clusters > 90), perfect training set performance every time, and relatively poor test set performance indicating some type of penalization is necessary. Increasing the value of p2 has the desired effect of yielding more parsimonious models without an obvious decline in performance. As before, the GA models seem to perform better than PAM but less well relative to the boosting model.
Next, results from the high resolution dataset are presented. The data require preprocessing. Some samples contain raw data from approximately 370,000 m/z values in the 700 – 12,000 Dalton range while other samples have about 330,000 data points. This discrepancy is particularly worrisome as the cancer samples appear more likely to have fewer datapoints. The information presented at the NCI-FDA website and in [12] includes some discussion of how the data were aggregated. The implementation in this work is similar to that described at the NCI-FDA website – details are available at a webpage containing supporting material [16]. After aggregation, the resulting spectra containing 7106 points were normalized to have the same average intensity. We note (data not presented) that while the models for the high resolution dataset on the NCI-FDA website have relatively good test set performance (accuracy of about 95%) they entail a large number of clusters – typically between 30 and 50. This is in contrast to the model reported for the low resolution DS1 Ciphergen data that had two clusters.
The results in Table 4 are quite similar to those reported for DS1 in the m/z >1500 range in that poorly performing high dimensional models are associated with p1 = 0 and the GA appears to again perform at an intermediate level.
Discussion
As implemented here, the genetic algorithm without penalties produces a large number of chromosomes that can perfectly discriminate a training set of the type considered here. For the last two analyses (DS1 with m/z > 1500 and DS2) those models produced with p1 = 0 are associated with a large number of clusters (median ≥ 90), indicating that many clusters have only 1 or 2 individuals. As demonstrated above, a resampling approach shows that models with large numbers of clusters will generally not perform as well as more parsimonious chromosomes and the use of penalization parameters greatly improves performance.
While this modification results in better models it does not address the other fundamental question of how to choose a final model. Because of their reliance on random choices GA models can and, for these data, will present very different solutions from a given dataset. Therefore, even if one uses a supplemental cross-validation process to choose some ideal set of penalization parameters many different models can be chosen by running the GA repeatedly. This is in contrast to many other means (e.g. boosting, discrimination methods) that have no such reliance on random processes and will produce the same answer given a fixed dataset and parameters. Some methods such as decision trees and PAM employ cross-validation as part of their fitting process and therefore do have a random component, but the results are not nearly so variable, at least for the data considered here. Next we explore some consequences when the GA is repeatedly applied to a fixed training and test set with the goal of finding "best" or superior models. Such repeated examination of test set performance violates the principal of evaluating the test set only after the model has been selected [14].
Consequences of repeated model fitting
Given a model developed on a training set, the performance of such a model on an independent test set is an unbiased estimate of its performance when exposed to a subsequent group of unlabeled cases that are generated by the same process. However, the situation becomes more complicated when a collection of models is considered. It is generally not true that the best performing model (judged by which model attains highest test set accuracy) will reproduce similar results on a yet another group of cases. Essentially, while every model has a true error rate, its performance on a particular test set is a function of both the true error and random variation. The best performing model is likely the beneficiary of positive random variation that is unlikely to be repeated in application to yet another set of data. In this sense the best performing model has an underestimated error rate when the selection of the best model is performed via repeated examination of a test set. We present a final set of bootstrap based analyses to illustrate the degree of bias.
For DS1, 50 runs of the following type of experiment were performed. First, a bootstrap sample of size 253 with 162 cancer and 91 controls is drawn (the cancer and control individuals were bootstrapped separately from their respective cohorts). This is denoted as Xb while the original cohort is X. This bootstrap sample is then split into training (81 cancer, 46 control) and test sets (81 cancer, 45 control). On the training part of the bootstrap sample the GA is run with p1 = .005 and p2 = .001. These parameters were chosen as they seemed to perform relatively well in Tables 2, 3, and 4 and they generally employed a smaller number of clusters and markers. The GA produced a model associated with this particular bootstrap sample, denoted . The performance of that model was then evaluated by the error rate in the test set portion of the bootstrap sample, denoted . Because the GA process produces different estimates when run on the same data due to randomness as described above, the model-fitting process is then repeated 19 more times on the same bootstrap sample to obtain 20 different models and 20 different measures of performance . The order of the training set and random sampling decisions made by the GA were allowed to vary though the composition of the test and training set were fixed for a given bootstrap sample. The best model, denoted , was chosen as that among the 20 with lowest classification error, . In the event of a tie, the number of clusters served as a tie-breaker (smaller is better). This procedure is meant to mimic the idea of applying 20 models to the test set and settling upon the best one. To get an idea of the bias in estimation error we then examine how the chosen, best model performs for the original cohort of 253 cases – this error rate is denoted as and the bias estimated by . This procedure was performed 50 times and one obtains an estimate of the distribution of the bias from
in 1) DS1 using the whole m/z range, 2) DS1 restricting the range to m/z > 1500, and 3) DS2 with 700 <m/z < 12000 (using different corresponding sample sizes). The use of the bootstrap to assess bias in this way conforms to the notion of treating the full sample distribution like a population distribution and the bootstrap sample distribution like the full sample distribution – see chapter 10 of [18].
The results in Table 5 show the degree of bias is relatively modest in the first dataset – on the order of 2% and somewhat higher (median of 4–5%) in the other data under consideration. This may not be of great practical import unless one is particularly concerned that the specificity be near 100% to justify using such tests on the basis of widespread diagnostic testing [12]. The degree of bias is influenced by, among other things, the number of times the test set is interrogated – here the figure used was 20 and it may be that greater bias is associated with increased searching. This analysis could also have been performed by splitting the data into 3 datasets (training, test, and bias assessment groups) though these datasets are small enough that the bootstrap approach was preferred in that it makes more efficient use of the data.
Generalizing results
There is considerable controversy regarding these ovarian cancer datasets – particularly with respect to whether the multitude of models with high or perfect sensitivity and specificity are more the result of rich complexity reflecting true biological variation [19,20] or flaws in experimental design [5,17]. While it is of paramount importance to know if true biological difference or flaws in experimental design are primarily responsible for the ease with which classification algorithms can separate the cancer and normal spectra (especially the low resolution dataset with 0 <m/z <20000 Daltons), the algorithms' performances will not change regardless of the answer to this question. Therefore, in the limited context of algorithmic performance considered here this critical issue is of secondary importance and not addressed. This observation indicates that these algorithmic analyses may still be valuable even if one believes the datasets to be flawed.
It is interesting to note how the GA performed on these three different datasets and speculate on its performance in other circumstances. The results in Table 5 regarding bias arising from multiple applications of the GA do vary somewhat among the three datasets under consideration. Because the spectra in the full, low resolution dataset (0 <m/z < 20000 Daltons) are easiest to correctly classify (as seen in Table 2) this dataset shows the smallest degree of bias arising from multiple applications of the GA – the median bias is about 2% in Table 5. Essentially the bias is low because the normal and cancer samples are so distinct and many models do very well. This is the case even though the multiple models may look very different from one another and use different primary m/z values to discriminate; in this case the bias is low because virtually all the dissimilar models do quite well.
The truncated, low resolution dataset (m/z > 1500 Daltons) was used to exclude the lower mass values that some [21] believe are difficult to interpret. Exclusion of these masses made the spectra harder to classify (see Table 3) and the associated bias in Table 5 was greater. The GA's performance in the high resolution dataset showed perhaps an intermediate level of difficulty in correctly classifying spectra (Table 4) and a corresponding intermediate degree of bias. The results suggest that as the spectra become easier to classify, the degree of bias due to repeated model fitting declines. This generalization is speculative in that it is based solely on these three related datasets and should be investigated in other datasets. Also, it should be pointed out that bias may be quite low in situations where there are only a few m/z values that can distinguish spectra. In this case one could speculate that repeated model fittings may identify primarily the same chromosome and therefore lead to very little bias.
Conclusions
This paper presents a genetic algorithm based on descriptions in earlier work. It was difficult to exactly reproduce performance of the original algorithm because important aspects were not well described and questions directed to the associated website were unacknowledged. Consequently, the GA's implementation described here is likely different than that made to produce the published findings. In particular, there are ambiguities concerning the manner in which more "fit" chromosomes are chosen to produce the subsequent generation, how chromosomes of different lengths may be produced and combined, and the possible use of penalization or other means to obtain parsimonious models. Despite these potential differences some aspects of the original algorithm's performance are likely shared with those of the model developed here.
Some modification of the algorithm to guard against overfitting seems necessary to obtain good performance. In particular, defining the fitness function simply as training set classification accuracy produces models with too many clusters. Here, a penalization based on the number of clusters and markers was imposed that improved algorithmic performance. A cross-validation procedure was incorporated to choose the penalization parameters – this resulted in algorithmic performance similar to other classification schemes. Results based on this type of algorithm should be accompanied by a clear description of how individual models are generated, e.g. what penalization parameters or other means of reducing the number of clusters are included and how they were chosen.
There is randomness and lack of reproducibility in model performance that depends on order of cases, random choice of initial chromosomes, and how the fitness function determines the subsequent chromosomes. Consequently, for a fixed training dataset, the algorithm can produce many chromosomes that perform well simply by repeatedly running the algorithm. There may be a temptation to use the algorithm repeatedly and evaluate test set performance to select the final model(s). While this can be a problem for classification algorithms in general, the random characteristics of this procedure may make it especially hard to resist. Here we saw some sense of the bias resulting from, such an approach. As the discussion regarding bias demonstrates, the reported sensitivities and specificities cannot be adequately assessed without very detailed description of the models' discovery. In this sense, those who employ such a scheme must supply complete information regarding the entire process used to choose the given models and users of algorithms that have this property of producing multiple models from a fixed dataset must be aware of this potential bias.
Overall, once modifications have been incorporated to address the overfitting concerns, the algorithm's performance seems comparable to other methods. It should be noted that the final models produced by this GA are of a simple to interpret form that may be based on a small number of markers and clusters. This simplicity is not necessarily present for other algorithms (e.g. boosting, neural nets, support vector machines). This algorithm seems a reasonable option for creating discrimination models though it does have disadvantages that might guide analysts to choose a different approach.
Methods
Data sources
The low resolution dataset, DS1, was obtained from a ProteinChip Biomarker System-II (PBS-II) surface-enhanced laser desorption ionization time-of-flight (SELDI-TOF) instrument produced by Ciphergen Biosystems, Inc. of Fremont, CA, USA using a WCX2 ProteinChip array, also produced by Ciphergen. Further details regarding sample handling and preparation are not readily available from the NCI-FDA website. While data from earlier low resolution datasets were available from the NCI-FDA website, these data (labeled 8-7-02) were chosen because the baseline does not appear to have been subtracted. As discussed by others [17] it does not appear possible to reproduce the original results of the genetic algorithm after baseline subtraction has been performed.
The high resolution dataset, DS2, was obtained from a hybrid quadrupole time-of-flight mass spectrometer (QSTAR pulsar I, Applied Biosystems, Inc. Framingham, MA, USA) modified to read the WCX2 ProteinChip. Additional information regarding handling and preparation of samples is available in [12].
Normalization of the two datasets was performed differently. For the low resolution data the spectra were rescaled linearly so the smallest value was 0 and the largest was 1. This was described in an earlier document on the NCI-FDA website (since removed) and was the approach described in [17]. This transformation has no effect on the genetic algorithm since additional rescaling is done within each individual spectrum on a chromosome by chromosome basis. This may have some effect (relative to performing no normalization) on the boosting and PAM algorithms but it is likely to be quite small as the maximum value for each spectrum was 100 (except one which reported a max value of 99.75) and the minima lay between 3.75 and 3.95 – so the effect was nearly one of applying the same transformation to each spectrum. The PAM and boosting algorithms were implemented after an additional normalization step that equalized the average intensity for each spectrum.
For the DS2 data, once the raw values were aggregated into 7106 bins the spectra were normalized to have the same average intensity. Here it seemed necessary to try to address the fact that the intensities for samples processed later were generally less than those processed earlier – see the QC document on the NCI-FDA website and [12]. Again, the normalization has no effect for the genetic algorithm. For the other algorithms it seemed important to try to address this temporal effect.
Data processing
Computing for all the classification algorithms (GA, boosting, and PAM) was done using the R programming language. Results for the PAM algorithm were obtained using the pamr package available from the R website [22]. On the website housing supporting information for this paper [16], full details are available showing the code and steps necessary to reproduce the findings presented in this paper.
Acknowledgements
This study utilized the high-performance computational capabilities of the Biowulf PC/Linux cluster [23] at the National Institutes of Health, Bethesda, MD, USA.
Figures and Tables
Figure 1 Test set classification of DS1 via 7 markers
Figure 2 Test set classification of DS1 via 10 markers
Table 1 Variation in results of GA applied 10 times
Generations Test Set Errors Clusters Markers Primary m/z
7 6(5%) 7 10 831.1
14 10(8%) 24 16 617.7
9 2(2%) 8 19 246.7
6 4(3%) 21 7 632.6
13 8(6%) 13 14 226.9
5 9(7%) 7 20 42.6
9 4(3%) 5 11 831.1
9 7(6%) 30 10 617.7
9 1(1%) 7 11 786.5
7 1(1%) 5 17 42.6
Table 2 Accuracy percentiles and characteristics of models from 50 cross-validation samples of DS1, m/z > 0
Algorithm Test Set Accuracy 25th, 75th Percentile Median # Clusters Median # Markers Proportion of Perfect Chromosomes
GA(P1 = 0, p2 = 0) .96, .98 12 15.5 1.0
GA(0, .001) .95, .99 12.5 11 1.0
GA(0, .002) .95, .98 12.5 10 1.0
GA(.002, 0) .96, .98 7 16 .98
GA(.002, .001) .97, .98 7 12.5 1.0
GA(.002, .002) .96, .99 8.5 9 .98
GA(.005, 0) .96, .98 6 15 .92
GA(.005, .001) .96, .98 5 11 .92
GA(.005, .002) .96, .99 6 8 .94
GA(.008, 0) .96, .98 3 15 .72
GA(.008, .001) .97, .99 4 9 .88
GA(.008, .002) .97, .99 4 8 .84
Best GA .97, .99 6 12 .94
Boosting .99, 1.0 NA NA NA
PAM .93, .97 NA NA NA
Table 3 Accuracy percentiles and characteristics of models from 50 cross-validation samples of DS1, m/z > 1500
Algorithm Test Set Accuracy 25th, 75th Percentile Median # Clusters Median # Markers Proportion of Perfect Chromosomes
GA(p1 = 0, p2 = 0) .83, .87 90 11 1.0
GA(0, .001) .80, .87 93 11 1.0
GA(0, .002) .82, .87 92.5 9 1.0
GA(.002, 0) .87, .92 12 19 .26
GA(.002, .001) .90, .93 14 13 .38
GA(.002, .002) .88, .92 13 7.5 .20
GA(.005, 0) .87, .92 5 20 .02
GA(.005, .001) .87, .90 5 10 .02
GA(.005, .002) .87, .92 5 6.5 0
GA(.008, 0) .85, .90 3.5 18.5 0
GA(.008, .001) .85, .89 4 8.5 0
GA(.008, .002) .85, .90 4 6 0
Best GA .87, .91 7 10 .12
Boosting .93, .97 NA NA NA
PAM .80, .84 NA NA NA
Table 4 Accuracy percentiles and characteristics of models from 50 cross-validation samples of DS2, 700 <m/z < 12000
Algorithm Test Set Accuracy 25th, 75th Median # Clusters Median # Markers Proportion of Perfect Chromosomes
GA(p1 = 0, p2 = 0) .70, .80 91.5 17 1.0
GA(0, .001) .63, .75 100 11 1.0
GA(0, .002) .61, .74 100 10 1.0
GA(.002, 0) .89, .93 22.5 19.5 .86
GA(.002, .001) .88, .93 22 14.5 .90
GA(.002, .002) .87, .91 22 8 .86
GA(.005, 0) .88, .93 7 19 .20
GA(.005, .001) .88, .93 7 11 .20
GA(.005, .002) .87, .93 7 6 .16
GA(.008, 0) .86, .91 5 18 .02
GA(.008, .001) .86, .91 5 10 .04
GA(.008, .002) .86, .91 5 4 .02
Best GA .88, .93 7 12 .24
Boosting .92, .95 NA NA NA
PAM .83, .86 NA NA NA
Table 5 Percentiles relating to bias of repeated examinations of a test set
Percentile
10th 25th 50th 75th 90th
DS1, m/z > 0
Bootstrap smallest error: 0.0 0.0 0.0 0.0 0.0
Original cohort error: .01 .01 .02 .03 .05
Estimated bias: .003 .012 .020, .032 .051
DS1, m/z > 1500
Bootstrap smallest error: .02 .02 .03 .05 .06
Original cohort error: .07 .07 .08 .11 .13
Estimated bias: .02 .04 .05, .07 .10
DS2, 700 <m/z < 12000
Bootstrap smallest error: .01 .02 .03 .04 .05
Original cohort error: .04 .05 .06 .08 .10
Estimated bias: .01 .03 .04, .06 .07
==== Refs
Petricoin E Ardekani A Hitt B Levine P Fusaro V Steinberg S Mills G Simone C Fishman D Kohn E Liotta L Use of proteomic patterns in serum to identify ovarian cancer Lancet 2002 359 572 577 11867112 10.1016/S0140-6736(02)07746-2
Adam BL Qu Y Davis J Ward M Clements M Cazares L Semmes O Schellhammer P Yasui Y Feng Z Wright GJ Serum protein fingerprinting coupled with a pattern-matching algorithm distinguishes prostate cancer from benign prostate hyperplasia and healthy men Cancer Research 2002 62 3609 3614 12097261
Li J Zhang Z Rosenzweig J Yang Y Chan D Proteomics and bioinformatics approaches for identification of serum biomarkers to detect breast cancer Clinical Chemistry 2002 48 1296 1304 12142387
Qu Y Adam BL Yasui Y Ward M Cazares L Schellhammer P Feng Z Semmes O Wright GJ Boosted decision tree analysis of surface-enhanced laser desorption/ionization mass spectral serum profiles discriminates prostate cancer from noncancer patients Clinical Chemistry 2002 48 1835 1843 12324514
Sorace J Zhan M A data review and re-assessment of ovarian cancer serum proteomic profiling BMC Bioinformatics 2003 4 24 12795817 10.1186/1471-2105-4-24
Qu Y Adam B Thornquist M Potter J Thompson M Yasui Y Davis J Schellhammer P Cazares L Clements M Wright G Feng Z Data reduction using a discrete wavelet transform in discriminant analysis of very high dimensionality data Biometrics 2003 59 143 151 12762451
Baggerly K Morris J Wang J Gold D Xiao L Coombes K A comprehensive approach to the analysis of matrix-assisted laser desorption/ionization-time of flight proteornics spectra from serum samples Proteomics 2003 3 1667 1672 12973722 10.1002/pmic.200300522
Petricoin EF Ornstein DK Paweletz C Ardekani A Hackett P Hitt B Velassco A Trucco C Wiegand L Wood K Simone C Levine P Linehan W Emniert-Buck M Steinberg W Kohn E Liotta L Serum proteomic patterns for detection of prostate cancer Journal of the National Cancer Institute 2002 94 1576 1578 12381711
Petricoin EF Liotta LA SELDI-TOF-based serum proteomic pattern diagnostics for early detection of cancer Current Opinion in Biotechnology 2004 15 24 30 15102462 10.1016/j.copbio.2004.01.005
Link to NCI-FDA Clinical Proteomics Databank
Reeves C Rowe J Genetic Algorithms : Principles and Perspectives : a Guide to GA Theory 2003 Boston: Kluwer Academic
Conrads T Fusaro V Ross S Johann D Rajapakse V Hitt B Steinberg S Kohn E Fishman D Whiteley G Barrett J Liotta L III EPVeenstra T High-resolution serum proteomic features for ovarian cancer detection Endocrine-Related Cancer 2004 11 163 178 15163296 10.1677/erc.0.0110163
Wu B Abbott T Fishman D McMurray W Mor G Stone K Ward D Williams K Zhao H Comparison of statistical methods for classifcation of ovarian cancer using mass spectrometry data Bioinformatics 2003 19 1636 1643 12967959 10.1093/bioinformatics/btg210
Hastie T Tibshirani R Friedman J The Elements of Statistical Learning: Data Mining, Inference, and Prediction 2001 New York: Springer-Verlag
Tibshirani R Hastie T Narasimhan B Chu G Diagnosis of multiple cancer types by shrunken centroids of gene expression PNAS 2002 99 6567 6572 12011421 10.1073/pnas.082099299
Supporting Documentation for Evaluating the Genetic Algorithm
Baggerly K Morris J Coombes K Reproducibility of SELDI-TOF protein patterns in serum comparing data sets from different experiments Bioinformatics 2004 20 777 785 14751995 10.1093/bioinformatics/btg484
Efron B Tibshirani R An Introduction to the Bootstrap 1993 New York: Chapman & Hall
Grizzle W Meleth S Clarification in the point/counterpoint discussion related to surface-enhanced laser desorption/ionization time-of-flight mass spectrometric identification of patients with adenocarcinomas of the prostate Clinical Chemistry 2004 50 1475 1476 15277364 10.1373/clinchem.2004.033456
Petricoin E Liotta L Proteomic pattern complexity reveals a rich and uncharted continent of biomarkers Clinical Chemistry 2004 50 1476 1477
Fung E Enderwick C ProteinChip Clinical proteomics: computational challenges and solutions Biotechniques Compututational Proteomics Supplement 2002 34 S34 S41
The R Project for Statistical Computing
Biowulf Cluster at NIH
| 15555060 | PMC539275 | CC BY | 2021-01-04 16:02:47 | no | BMC Bioinformatics. 2004 Nov 19; 5:180 | utf-8 | BMC Bioinformatics | 2,004 | 10.1186/1471-2105-5-180 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-5-1941558829710.1186/1471-2105-5-194Methodology ArticleOptimized LOWESS normalization parameter selection for DNA microarray data Berger John A [email protected] Sampsa [email protected]ärvinen Anna-Kaarina [email protected] Henrik [email protected] Sanjit K [email protected] Jaakko [email protected] Department of Electrical and Computer Engineering, University of California, Santa Barbara, CA 93106-9560, USA2 Institute of Signal Processing, Tampere University of Technology, P.O. Box 553, 33101 Tampere, Finland3 Biomedicum Biochip Center, University of Helsinki, P.O. Box 63, 00014 Helsinki, Finland4 Medical Biotechnology Group, VTT Technical Research Center of Finland and University of Turku, P.O. Box 106, 20521 Turku, Finland2004 9 12 2004 5 194 194 6 6 2004 9 12 2004 Copyright © 2004 Berger et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Microarray data normalization is an important step for obtaining data that are reliable and usable for subsequent analysis. One of the most commonly utilized normalization techniques is the locally weighted scatterplot smoothing (LOWESS) algorithm. However, a much overlooked concern with the LOWESS normalization strategy deals with choosing the appropriate parameters. Parameters are usually chosen arbitrarily, which may reduce the efficiency of the normalization and result in non-optimally normalized data. Thus, there is a need to explore LOWESS parameter selection in greater detail.
Results and discussion
In this work, we discuss how to choose parameters for the LOWESS method. Moreover, we present an optimization approach for obtaining the fraction of data points utilized in the local regression and analyze results for local print-tip normalization. The optimization procedure determines the bandwidth parameter for the local regression by minimizing a cost function that represents the mean-squared difference between the LOWESS estimates and the normalization reference level. We demonstrate the utility of the systematic parameter selection using two publicly available data sets. The first data set consists of three self versus self hybridizations, which allow for a quantitative study of the optimization method. The second data set contains a collection of DNA microarray data from a breast cancer study utilizing four breast cancer cell lines. Our results show that different parameter choices for the bandwidth window yield dramatically different calibration results in both studies.
Conclusions
Results derived from the self versus self experiment indicate that the proposed optimization approach is a plausible solution for estimating the LOWESS parameters, while results from the breast cancer experiment show that the optimization procedure is readily applicable to real-life microarray data normalization. In summary, the systematic approach to obtain critical parameters in the LOWESS technique is likely to produce data that optimally meets assumptions made in the data preprocessing step and thereby makes studies utilizing the LOWESS method unambiguous and easier to repeat.
==== Body
Background
DNA microarray technology has become a standard tool in biomedical research for large-scale transcriptional monitoring [1]. A growing number of microarray experiments seek to compare samples labeled with two different dyes, such as Cyanine5 (Cy5) and Cyanine3 (Cy3). However, several studies report that the dyes bind on a microarray slide differently due to the variations in their chemical characteristics [2-6]. In addition, the image scanner settings also affect dye intensity measurements. Should these discrepancies not be corrected, the resulting data may not be useful for analysis purposes. Thus, there is a need for dye normalization for the microarray slide prior to actual data analysis to reduce systematic variability.
Microarray data preprocessing contains three phases: quality control, within-slide normalization, and between-slide normalization. Within-slide normalization aims to correct dye incorporation differences which affects all the genes similarly, or genes with the same intensity similarly [7]. One scatterplot-based normalization technique that is particularly suitable for balancing the intensities is called locally weighted scatterplot smoothing (LOWESS) and its original application was for smoothing scatterplots in a weighted, least-squares fashion [8]. This technique is typically chosen to calibrate microarray data because a popular, freely available implementation is available in the statistical software package R [9] and in many commercial microarray analysis software suites such as the Agilent Feature Extraction Software. Moreover, several other freely available microarray data handling packages have incorporated this normalization technique [10,11]. It is noted that many normalization studies simply call the function without rigorous consideration for the actual algorithmic parameters [12,13]. Our analysis reports that the choices of different parameter values drastically affect the quality of the normalization results. The original work on LOWESS clearly mentions the problem of obtaining parameter values and even offers some ideas for finding suitable data-dependent choices [8,14]. However, many microarray studies have omitted such rationale and made arbitrary selections for different experimental data sets [13,15,16] and some studies even failed to report their parameter assumptions in their methods [17-19]. Although this practice has not lead to significant consequences for most of the parameters in LOWESS, we show that the parameter that represents the fraction f of neighboring samples to be included in the weighted polynomial fit is particularly sensitive and its variation greatly affects the normalization results. This parameter should be carefully chosen through a systematic procedure where experimental assumptions are clearly specified. Benefits in the normalization process may be considered to be small in their own right, but these improvements are extremely meaningful in the context of searching for subtle biological differences in gene expression.
We outline an optimization-based procedure for obtaining a systematic value for f in print-tip LOWESS normalization. Results are compared to common, arbitrary selections of f. The proposed procedure first examines a case study where we have utilized three quality filtered, self versus self hybridization experiments. With self versus self experiments, we are able to clearly detect normalization differences. Such analysis also verifies that the optimized method produces properly calibrated ratios. Our proposed technique is also demonstrated on a typical set of quality filtered microarray data. We utilize a set of breast cancer data that has replicated measurements for four different tumor cell lines [20]. In addition to visual comparisons, we quantitatively assess the performance of the different normalization procedures using a goodness-of-fit test. Our results demonstrate that arbitrarily selecting the LOWESS bandwidth parameter produces statistically different results for certain print-tips compared to the proposed optimized parameter selection formulation. Moreover, for genes that have been verified using reverse transcription-polymerase chain reaction (RT-PCR) experiments, we show that calibrated results are substantially affected by the choice of f. Our self versus self data, including the original TIFF images, are available online [21] and the replicated breast cancer data is posted by the original authors of that study [22].
Results and discussion
Within-slide normalization
Within-slide normalization is used to correct the dye intensity errors introduced across one microarray slide. The result of this step provides the normalized, calibrated ratios. Let denote the background corrected selection for the intensity of the jth gene of the Cy3 (green) colored sample. Similarly, let denote the jth gene of the Cy5 (red) colored sample. One key issue for the dyes is that they are consistently imbalanced [12,13]. Different labelling effciency between the two fluorescent dyes exists and in some labelling schemes Cy5 is systematically less intense than Cy3. Normalization techniques are needed in order to render the gene expression levels measured by the two different dyes comparable [23,24]. Dye biases can stem from a wide variety of factors, including physical properties of the dyes, effciency of dye incorporation, and processing errors. Such errors may be introduced by slight variations in the amount of mRNA used to create the target hybridized to each microarray or in the quantity of dye used to fluorescently label each target.
For a single microarray experiment, there are n total gene expression ratios and we denote the observed vector of ratios for a single experiment as r ∈ ℝn × 1. The calibrated ratio of expression for each gene is obtained by dividing the test by the reference sample intensities with the proper normalization factor in the denominator,
for i = 1, 2..., n, where n is the total number of spots on a microarray. The normalization factor, denoted by Φ(·), is a function of data-dependent variables. If the dyes are linearly dependent, it can be assumed that the normalization function is a constant, namely Φ(·) = φ. Many studies have looked at linear dependencies [25], as well as a generalized form of the normalization factor Φ(·) that is a function of an often times unknown number of experiment-specific parameters.
Many studies perform within-slide normalization in a global manner by assuming the error effects are stationary across an entire slide. This is currently true for the cases of Affymetrix GeneChip or Agilent oligonucleotide microarrays. For cDNA microarrays, however, the sources of variation typically originate in a localized or spatial manner [13], mainly from the different print tips for each sub-array of the slide [26]. The process of determining the values for Φ(·) is highly dependent on the characteristics of the data for each print-tip [12]. For example, some print-tips have highly nonlinear effects, while other print-tips in the same experiment behave quite differently and may exhibit linear trends in dye bias. Furthermore, the systematic manner in which the experiment has been conducted also influences the results of different slides, but it is our intention that such effects will be satisfactorily captured in the behavior of the print-tip statistics. As a consequence, we omit global calibration considerations that neglect print-tip distinction and focus solely on scatterplot-based normalization in a termed localized manner.
LOWESS method
One of the most widely used nonlinear correction techniques is the LOWESS method, which was first applied to microarray data by Yang et al. [16]. The main idea behind LOWESS is to utilize a locally weighted polynomial regression of the intensity scatterplot in order to obtain the calibration factor. Compared to other techniques, like housekeeping-based normalization or dye-swap experiments, scatterplot-based normalization is more robust in many types of scenarios where assumptions of constantly expressed genes may break down [23]. Subsequent microarray studies have also chosen this method due to the robustness of fit in the presence of a few extreme outliers. Original studies have examined the (Ig, Ir)-scatterplot in log2-space for determining the value of Φ(·). It has been suggested in separate works by Dudoit et al. [15] and Yang et al. [16] that a log2-based scatterplot of the average fluorescence intensity A versus the transformed ratio M should be used instead of a simple, log2-based intensity scatterplot. This type of scatterplot is commonly known as a Bland-Altman plot in the statistics literature. The values for A and M are given as,
for i = 1, 2,..., n. Equations (2) and (3) are preferred over the original intensities because the (A, M)-scatterplot may reveal artifacts that are not clearly visible in the ordinary intensity scatterplot. Such a transformation represents a scaled, 45° rotation of the (Ig, Ir)-coordinate system [16].
The smoothing procedure has been designed to accommodate measured scatterplot data obeying the form Mj = g(Aj) + εj, where the jth transformed ratio Mj is a function of the corresponding overall intensity Aj and a zero mean random variable εj. The smoothed point at Aj using LOWESS with a degree d polynomial is (Aj, ), where is the fitted value of the regression. The LOWESS estimate, , is a weighted linear combination of the Mi
where the hi(Aj) depend on Ai, ∀i, but not on the Mi. The LOWESS algorithm contains four data-specific parameters, namely the polynomial order d, the number of LOWESS algorithmic iterations t, the weight function w(·), and the fraction of the data points used in the local regression f. Consequently, these parameters all affect the values of the weights hi(Aj) in Eq. (4). For a complete outline of the LOWESS algorithm, consult [8,14,27]. In practice, the polynomial order for DNA microarray data is usually selected as being either d = 0, 1, or 2, depending on the choice of (Ig, Ir)- or (A, M)-coordinate systems, the tri-cube weight function is quite standardized for all types of data [8], and the number of iterations is usually fixed at t = 3. The final parameter must be chosen where f ∈ (0, 1] and it is often times assigned an arbitrary value without any justification. However, since the choice of f ultimately determines the magnitude of calibration, it is essential to put heavy emphasis on choosing this parameter carefully. In the literature, many microarray studies neglect such concerns and arbitrarily select f for different experimental data sets [12,13,16]. Formal consideration of the parameter f is typically glossed over by simply stating that the larger the f value, the smoother the fit. Although this is a true statement, the consequences are deeper than the statement leads on. Different types of data may require smoother fits but DNA microarray data takes all shapes. Also, what defines a smoother fit is also highly subject to interpretation depending on the actual data.
The optimized approach
For a microarray experiment, there are a total of ℓ print-tips used on a single slide. In order to reliably determine the value of f for each print-tip group, we introduce an optimization approach based on the actual microarray data for each print-tip group. We slightly modify our notation to include print-tip indices as a subscript k for each transformed ratio. The goal is to select the appropriate values of fk that minimizes the mean squared difference between the LOWESS fit of the ith transformed ratio in the kth print-tip group, , and the corresponding normalization reference level, ψk,i(·). The value of each ψk,i(·) is a function of experiment-specific parameters such as temperature or other environment settings which may differ from sample to sample in a single experiment. Accordingly, the cost function to be minimized for the kth print-tip group across all transformed ratios is
with the constraint that fk ∈ (0, 1]. Here, the value nk is the total number of ratios for the kth print-tip group. Correspondingly, for a total of ℓ print-tip groups, we have . For certain experiments, like self versus self hybridizations, the true expression value is known a priori. If ψk,i(·) is unknown, reliable estimates that reflect experiment-specific assumptions may be used. Usually there are tens of thousands of genes in a microarray study and a plausible assumption is that the mean of the log2-transformed ratios after normalization is zero. Also, in a variety of experiments, platform-dependent control transcripts that are known to have certain expression at a constant level may be utilized in the optimized approach. Furthermore, in our breast cancer case study we show how to obtain statistically reliable estimates of ψk,i(·) from replicate slides. We also show how our approach may be used if replicates are not available for typical microarray studies. Ultimately, the optimized approach requires experimenters to explicitly state their assumptions behind the study, which is systematically better than arbitrarily choosing parameter values. In addition, determining an experiment-specific fk by trial and error may be time consuming and will oftentimes lead to non-optimal results. The chosen optimization algorithm for minimizing the corresponding cost function is based on a combination of golden-section search and successive parabolic interpolation as outlined by Forsythe et al. [28]. This approach finds the best fk for minimizing δk(fk) for each print-tip, k = 1,..., ℓ within a tolerance of ±0.01. Each print-tip, resultingly, may have a different, optimal bandwidth parameter.
Normalization step
After the estimates have been obtained, calibrating the intensities for all the Ak,i is given as
for i = 1,..., nk, and k = 1,..., ℓ. For the local LOWESS normalization within each print-tip group, the issue of how the total intensities are spread about the sample mean for the group becomes a factor to consider when normalizing the data [16]. After normalization, all the log2-ratios from the different print-tip groups are usually centered around zero. Some print-tips may have larger variances compared to others and an appropriate scale adjustment is needed to account for such differences. One proposed approach is to find the maximum likelihood estimate for the scale of the variance for each print-tip group [16]. This method assumes that all log2-ratios from the kth print-tip group follow a normal distribution with mean zero and variance σ2, where σ2 is the variance of the true log2-ratios and is the estimated scale factor for the kth print-tip group. However, this is only valid for certain types of data that reasonably follow a normal distribution and in our work we observe that this assumption may often times lead to undesirable results. Refer to [16] for further details.
Another approach proposed here that is able to deal with the variance scaling issue is to introduce a weighting factor in the calibration function that is of the form
for i = 1,..., nk, k = 1,..., ℓ, and where the weight is given as . The bias-corrected sample variance for the kth print-tip is denoted by and is given as
where denotes the sample mean for print-tip k. Furthermore, the minimum sample variance is given as
Compared to the maximum likelihood method outlined by [16], this method stresses higher weighting on print-tip groups that exhibit less variance and lower weighting for highly variant print-tips. If such a weight is not introduced, the normalization may improperly calibrate highly variant print-tip groups that have extreme sample means and many genes may erroneously be considered as differentially expressed as a consequence. Other treatments, such as the one suggested by Quackenbush [12] examine the geometric mean of the tip variances as a scale factor for the normalization estimate. However, such a treatment may not always scale the tips properly since some tips may still be overly compensated. Our proposed scaling factor λk takes values over (0, 1] while other scaling methods may have larger upper limits. By calibrating data using Eq. (9), we have obtained nearly identical sample means, but less total variance for the resulting data compared to previously published techniques. The computation of λk is straightforward and easy to calculate but our novel variance stabilization procedure does not take into account any heteroscedasticity in the data, namely observed increasing ratio variance with decreasing measurement intensity A. A rigorous comparison of print-tip scaling is beyond the scope of this contribution, but it is noted that the different scaling procedures affect the overall calibration scheme.
Case studies
To demonstrate the utility of our optimized LOWESS normalization procedure, we first utilized a set of three self versus self experiments [21], BT-474, MCF-7, and HBL-100, which were obtained using the protocols delineated in the methods section. In addition, we calibrated a set of four breast cancer cell lines [22], BT-474, MCF-7, MDA-MB-436, and MDA-MB-361, each measured in comparison to the reference cell line HBL-100, which were obtained using the protocols outlined by Järvinen et al. [20]. For each cancer cell line, three replicate slide hybridizations were available. In order to reduce the effects of spots whose intensities are not reliable due to experimental or printing errors, we used two separate quality filtering methods and normalized the intensities after discarding values that were detected unreliable. The assessment of ratio quality was performed using the method proposed by Chen et al. [29] and the evaluation of spot quality was performed using the method of Hautaniemi et al. [30]. Optimized parameter selection for fk was performed and print-tip LOWESS normalization results are compared to the results using arbitrary choices of the parameter fk. The implementation took a few minutes to run on a standard desktop PC running MATLAB.
Self versus self experiments
Self versus self experiments provide a trivial application to test our method since the amount of mRNA in both the test and the reference samples is the same. Thus, the points of an intensity scatterplot in the log2 - log2 space should be distributed along a straight line that intersects zero with a slope of unity. In the (A, M)-coordinate system, all values of M should lie on a straight line at M = 0 for all values of A; this means that the calibrated ratios should ideally be unity for all variables. Correspondingly, the cost measure is given when ψk,i(·) = 0, (∀k, i), in Eq. (5) for the (A, M)-coordinate systems. Separate trials were conducted using weighted, zeroth-order (d = 0), first-order (d = 1), and quadratic (d = 2) polynomial fits. For all trials, the number of print-tip LOWESS iterations was fixed at t = 3. The weight function used is given by Cleveland [8]. For each experiment, the local print-tip groups were separately normalized with their respective, optimized values of fk. As a comparison to arbitrary selections of fk, the print-tip normalization was also carried out using fk = 0.2, 0.4, 0.6, and 0.8 in separate trials. Figure 1 shows the (M(Arb), M(Opt))-scatterplot comparison between the calibration results with d = 1 using optimal fk and arbitrary fk for the BT-474 self versus self experiment. The points that deviate from the blue line are the genes that are most affected by the choice of fk. The M(Arb) data in this figure was calibrated using fk = 0.4, ∀k.
In all three self versus self experiments, the global sample means of M were nearly the same after calibration, regardless of the choice of fk. However, the calibrations that used optimized selections of fk for each print-tip resulted in data that contained less overall variance compared to the arbitrary selections. The ultimate goal of calibration is to adjust the dynamic range for the transformed ratios and reduce the variability within the data. By using optimized selection of fk, we outperform all arbitrary formulations to achieve these goals.
Typical microarray experiments
One immediate concern for typical experimental microarray data is that many genes may be over- or under-expressed and the true, transformed gene expression ratio ψk,i(·) surely will not be equal to zero for all genes. Accordingly, implementing the cost function in Eq. (5) becomes an immediate challenge since the normalization reference level of all the genes for a typical microarray experiment may be diffcult to determine with complete accuracy. We note that our cost function still may be used with the assumption that the sample mean for each print tip before log2-transformation is unity. In most microarray experiments, many genes may be assumed to have constant RNA concentrations while smaller numbers of genes may be over or under expressed, namely their sample mean over all the genes is zero, . Using this assumption in Eq. (5), our experiments show that by minimizing the cost function in this context, like in the self versus self case study, we are able to systematically choose fk and the only consequence is that the minimum of the cost will not be as low as in the self versus self scenario where all genes should be constantly expressed. The main benefit of utilizing LOWESS for microarray normalization is that it is robust to extreme outliers and the cost function implemented in this fashion further restricts the effects of such extreme points in the regression. Ultimately, this implementation results in reliably calibrated ratios compared to the arbitrary formulation where different choices of fk affect the resulting data.
Since a single microarray experiment represents an observation, multiple observations would be needed to compute a reliable estimate of the true transformed ratio values. The use of only a small number of replicate slides may be satisfactorily used to determine reliable estimates of true gene expression and one study showed that three replicates suffce for significantly reducing experimental variability [31]. With the growing number of publicly available microarray data, conducting replicate experiments is becoming a popular solution to assess experimental errors and reduce noise bias in the measurements [32]. The advantages of replicate slides also greatly help the analysis of between-slide variability and help address formal statistical considerations when drawing biological conclusions. Here, we show that the optimized normalization approach may be directly extended in an iterative manner to use the estimates of the true ratio values for further specifying fk. After an initial round of optimized LOWESS normalization for each replicate slide with ψk,i(·) = 0 in Eq. (5), the sample mean for each gene may then be calculated using the replicates. The normalization reference levels ψk,i(·) were reassigned these average gene expression values in Eq. (5). Each experiment was then separately calibrated a second and final time using the optimization approach and the final results were noticeably different compared to the normalized data using f = 0.2 that Järvinen et al. posted on their website [20]. A noteworthy consideration to address here is the overall effect of an iterative calibration process on the underlying structure of the data. Experimentally, once the optimized LOWESS regression is computed using the average value for each gene and normalization is performed, subsequent calibration attempts using the cost function-based method do not result in drastically different data. The subsequent regressions are nearly constant lines near M = 0 in the (A, M)-scatterplot if the cost function approach is used. Consequently, the calibrated data reach a stable domain with a small dynamic range. Empirically, we found that performing optimized normalization in an iterative manner will not propagate regression effects through to disrupt the underlying structure of the data.
Figure 2 shows the scatterplot comparison between the calibration results using optimal and arbitrary selections of fk for the first replicate BT-474 hybridization. Some genes in this plot report 4-fold differences and ultimately these differences affect data analysis. Consequently, the errant choice of this parameter fk may have deleterious effects on different biological studies. To illustrate the differences for one representative print-tip in this breast cancer study for the first replicate of the BT-474 cell line, Figure 3 plots the regressions obtained by both methods. All the data points for this hybridization are shown as a two-dimensional histogram [33], while the spots given by print-tip k = 16 are highlighted in black. In this plot, we show that the regression obtained by the optimized choice of f16 differs from the one obtained by arbitrarily selecting f16 = 0.2 and the calibration results are thus affected. Figure 4 reports arbitrary calibration results and Figure 5 shows optimized results. The data in Figure 5 has less overall variance when calibrated with the optimized choices of fk.
As further illustration of the calibration differences between the optimized and arbitrary calibration results, we employ a goodness-of-fit test [34]. We wish to make a direct test of the data, independent of any underlying parent distribution of the ratios, and we use the following statistic for the kth print-tip group
where M(Arb) and M(Opt) are the arbitrary and optimized calibration results, and the denominator within the summation is simply the variance of the difference between M(Arb) and M(Opt). The null hypothesis is defined to be H0: the normalized ratios using arbitrary f are comparable to ones using optimized f. We tested against p < 0.05 for the distribution and reported the alternative hypothesis for a few print-tip groups on almost all the slides. In this analysis, we compared optimized choices of f for each print-tip to the arbitrary choices f = 0.2, 0.4, 0.6, and 0.8. By looking across each replicate of the calibrated data for all four breast cancer cell lines, almost all slides in this study reported at least one print-tip to have statistically different calibration results based on the choice of fk. Often times a single slide would report two or three print-tip groups that had statistically different calibration results.
In addition to statistical analysis, genes that exhibit known over-expression in the BT-474 cell line data [35] were selected here for more detailed analysis. In particular, genes that were verified experimentally using reverse transcription-polymerase chain reaction (RT-PCR) were of the highest interest. Comparing our optimized calibration results utilizing the replicate data to the normalized data by Järvinen et al. [20], our results conform strongly with most of the over-expressed genes given in a list from a parallel study [35]. Two genes in particular stand out to demonstrate the benefits of utilizing our proposed method: homeo box B7, which was validated with RT-PCR [35], and v-erb-b2, which is known to be over-expressed in the BT-474 cell line [35]. The results posted by Järvinen et al. [20] for calibrating the homeo box B7 gene shows that it falls within the top 18% of overall gene expression, but by using the optimized approach we report it to be within the top 13%. For the v-erb-b2 gene, both calibration techniques report that this gene falls within the top 1% of the genes in terms of expression. As a result, for the homeo box B7 gene, the calibration factor fk is responsible for about 5% change in the reported gene expression. This is a dramatic result that may influence how the expression for this gene may be interpreted in comparison to the accepted biological knowledge of a certain experiment. As public data from microarray experiments continues to become available, the knowledge of certain genes will undoubtedly be uncovered for well-studied cell lines and this information will help further assess normalization and microarray quality control tasks.
Conclusions
The LOWESS method has recently been applied in other applications for the biological sciences. Comparative genomic hybridization (CGH) is a molecular cytogenetic method of screening a tumor for genetic changes. The alterations are classified as DNA gains and losses and they reveal a characteristic pattern that includes mutations at chromosomal and subchromosomal levels. Our proposed optimized scheme is directly applicable to the application of calibrating CGH microarray experiments, as well as for data analysis aspects. For example, the work of Clark et al. [36] utilized the LOWESS method for identifying the regions where gene copy numbers were aberrantly high or low in prostate cancer using CGH microarray technology. The parameter f was chosen arbitrarily and its value was not reported in the study. Consequently, reproduction and verification of these results may be diffcult. For instance, some of the important biological findings, such as start and end points of amplifications and deletions, may be adversely affected by different choices of f.
In addition to CGH analysis, LOWESS has found application in case-control studies where logistic regression has been used to model the relationship between binary responses and continuous predictor variables [37]. In these types of studies one may use LOWESS to remove systematic trends that contaminate the laboratory measurements of predictor variables. The analysis reported by Borkowf et al. [37] clearly shows that different choices of f result in noticeably different correction effects and the optimization method proposed here may be suitable for enhancing such a study. Adaptations to the cost function may be utilized to handle this type of data. In addition, analysis of other types of scatterplot data by utilizing the LOWESS method with an arbitrary choice for the bandwidth parameter is undoubtedly susceptible to varied interpretations or errant conclusions [38,39].
Another result of this optimized calibration study is that we uncovered a better understanding of choosing the parameter d in the weighted polynomial fit. A higher-order (d > 2), weighted polynomial is rarely needed based on the argument that such an assumption is, to a certain extent, over-fitting the data. From the findings of our study, we find that it is better to use a linear estimate based on minimizing the estimate errors across (A, M)-scatterplots. Consequently, different choices of d resulted in different optimized values for f. The reason is that for the higher-order polynomial, it is beneficial in general to retain a larger fraction of the values of A for the weight function in computing the polynomial coeffcients. It is very important to carefully select f since ultimately, the bandwidth is a function of the polynomial order.
Here, we also reaffirmed the idea that the quality filtering of ratios and spots is a necessary step that should precede all experimental microarray data handling procedures, whether it is scatterplot-based normalization or any other normalization method, since errant ratios would surely have a deleterious affect on the calibration. For instance, in the BT-474 data, the first replicate slide had poor ratio quality for a handful of genes. Calibration without considering or removing these errant spots resulted in less reliable results. This study addresses the issue of locating sources of experimental error for print-tips that have high sensitivity for the parameter f . For one, print-tips are physically different and they are considered to have different types of errors introduced based on these properties. In the formulation of normalization, it is imperative to address such subtle issues when choosing and implementing any algorithm.
The systematic choice of the parameters in the LOWESS algorithm has not been previously addressed in the microarray literature and the method proposed here may be utilized in different microarray platforms. Such a treatment is also important for a wide variety of applications that employ scatterplot-based regression. The findings of this study illustrate that by choosing different values of f for the LOWESS algorithm results in noticeably different normalization results. This proposed method requires the calibration step to clearly state the assumptions used for within-slide normalization. Our optimization algorithm is more systematic than simply choosing an arbitrary parameter value or through trial and error techniques since the optimized approach relies on the actual underlying structure of the data. We also stress that such an optimization algorithm may also be utilized for other studies in addition to DNA microarray normalization treatments. Proper changes need to be made to Eq. (5) to reflect the ideal model for the data captured in the function ψk,i(·), but in some studies, such a function may be satisfactorily determined or estimated from the data.
Methods
Data resources
For the self versus self hybridizations, custom cDNA microarray experiments proceed as follows. Altogether, three microarray hybridizations were performed using custom printed cDNA microarray slides from the same print batch. Labelling, hybridization and washing were done as described previously by Monni et al. [40] and Järvinen et al. [20]. Briefly, total RNA was extracted from cell lines BT-474, HBL-100, and MCF-7 and labelled with Cy3-dUTP and Cy5-dUTP (Amersham Biosciences, Piscataway, NJ). The custom printed cDNA microarrays comprised of 11,520 clones from Incyte Genomics IRAL cDNA library and 1,136 clones from Research Genetics library. Microarrays were printed on poly-l-lysine coated slides using an Omnigrid arrayer (GeneMachines) as described previously [20]. Microarrays were scanned with an Agilent laser confocal scanner (Agilent Technologies, Palo Alto, CA) and gridded using the DEARRAY software developed by Chen et al. [29]. For the four breast cancer cell lines, custom cDNA microarray experiments were provided in a separate contribution by Järvinen et al. [20] and detailed protocols are described in that work. The relevant genes in our study were verified using RT-PCR in a parallel study by Hyman et al. [35].
Data quality filtering
All microarray experiments contained in this work were conducted and spotted using groups of ℓ = 32 print-tips, with each tip being responsible for either 384 or 420 spots in their respective subarray. In order to reduce the effects of spots whose intensities are not reliable due to experimental or printing errors, we used two separate quality filtering methods and normalized the intensities after discarding values that were detected unreliable. The assessment of ratio quality was performed using the method proposed by Chen et al. [29] and ratios that had a quality value below the threshold 0.5 were discarded from our analysis. This quality cutoff value has, in the past, been shown to represent less reliable cDNA microarray measurements due to either low signal intensity, high local background level, uneven distribution of the target intensity, and/or small target size. The evaluation of spot quality was performed using the method of Hautaniemi et al. [30]. In this Bayesian networks-based method, we utilized the following features in determining spot quality. Bleeding, spot roundness, and spot intensity were assessed for the Cy5 channel and bleeding, spot size, spot roundness, background intensity, and fitting error were evaluated for the Cy3 channel. These features were chosen since this set was found to result in the best classification accuracies [30]. The trained Bayesian network was applied to each slide in this study and all the spots having a quality value of zero were excluded from the subsequent analysis.
Authors' contributions
JAB developed the mathematical formulation of the problem, implemented the optimized normalization algorithm in MATLAB, developed the statistical analysis, and wrote the manuscript. SH developed the LOWESS normalization software in MATLAB, coordinated spot quality filtering, and assisted in drafting the manuscript. AKJ conducted the self versus self microarray experiments and performed ratio quality filtering for data analysis. HE assisted in data preparation and in drafting the manuscript. SKM participated in the design and coordination of the study and assisted in drafting the manuscript. JA reviewed the statistical analysis and participated in the design and coordination of the study. All authors read and approved the final manuscript.
Acknowledgements
The authors thank the anonymous reviewers for their comments and contributions. This work was supported in part by a University of California MICRO grant with matching support from Philips Research Laboratories and in part by Microsoft Corporation and the Academy of Finland.
Figures and Tables
Figure 1 (M(Arb), M(Opt))-Scatterplot analysis of BT-474 self versus self data This plot compares the calibrated ratios obtained by LOWESS (d = 1) with the arbitrary choice of fk = 0.4 for all print-tips compared to ratios obtained with optimized fk for each print-tip group. The line of unity slope that passes through the origin shows where all the points should lay if both calibration methods produced identical ratios. For this self versus self experiment, the group of points that lay under this line shows that the arbitrary fk may improperly under-normalize these points.
Figure 2 (M(Arb), M(Opt))-Scatterplot analysis of BT-474_01 data This plot compares the calibrated ratios obtained by LOWESS (d = 1) with arbitrary (fk = 0.2) and optimized bandwidth windows for the first replicate hybridization of the BT-474 breast cancer cell line. Again, the line of unity slope shows where all the points should lay if both calibration methods produced identically calibrated ratios. Many points deviate from the similarity line in this example and such results are commonly observed for the microarray data used in this study. Consequently, it is clear that the choice of fk greatly affects how the data is calibrated. Points that are furthest away from the similarity line are highly influenced by the choice of fk in LOWESS calibration.
Figure 3 Print-tip LOWESS comparisons for BT-474_01 data This (A, M)-scatterplot shows a two-dimensional histogram [33] or all the spots for the first replicate BT-474 breast cancer hybridization, where the bright red color indicates a high concentration of spots. Print-tip k = 16 is highlighted by black dots. The LOWESS estimates obtained by using f16 = 0.2 are shown by the dark blue line and the estimates using optimal f16 is shown here in light blue. This result is typical for the print-tips in this study based on minimizing the cost function given in Eq. (5).
Figure 4 Arbitrary calibration results for BT-474_01 data All spots are shown using a two-dimensional scatterplot with the spots from print-tip k = 16 are highlighted here in black. LOWESS calibration has been performed using the choice of fk = 0.2 for all print-tips.
Figure 5 Optimized calibration results for BT-474_01 data This scatterplot shows LOWESS calibration after optimized choices of fk have been obtained for all print-tips. Compared to the results in Figure 4, the normalized data here has less overall variance. In addition, genes that have been verified experimentally conform in better agreement with the well-known biology.
==== Refs
Schena M Shalon D Davis RW Brown PO Quantitative monitoring of gene expression patterns with a complementary DNA microarray Science 1995 270 467 470 7569999
Goryachev AB MacGregor PF Edwards AM Unfolding of microarray data Journal of Computational Biology 2001 8 443 461 11571077 10.1089/106652701752236232
Ideker T Thorsson V Siegel AF Hood LE Testing for differentially-expressed genes by maximum-likelihood analysis of microarray data Journal of Computational Biology 2000 7 805 817 11382363 10.1089/10665270050514945
Kerr MK Martin M Churchill GA Analysis of variance for gene expression microarray data Journal of Computational Biology 2000 7 819 837 11382364 10.1089/10665270050514954
Tseng GC Oh MK Rohlin L Liao JC Wong WH Issues in cDNA microarray analysis: quality filtering, channel normalization, models of variations and assessment of gene effects Nucleic Acids Research 2001 29 2549 2557 11410663 10.1093/nar/29.12.2549
Wang X Ghosh S Guo SW Quantitative quality control in microarray image processing and data acquisition Nucleic Acids Research 2001 19 e75 10.1093/nar/29.15.e75
Dobbin K Shih JH Simon R Statisical design of reverse dye microarrays Bioinformatics 2003 19 803 810 12724289 10.1093/bioinformatics/btg076
Cleveland WS Robust locally weighted regression and smoothing scatterplots Journal of the American Statistical Association 1979 74 829 836
Ihaka R Gentleman R R: a language for data analysis and graphics Journal of Computational and Graphical Statistics 1996 5 299 314
Engelen K Coessens B Marchal K Moor BD MARAN: normalizing micro-array data Bioinformatics 2003 19 893 894 12724302 10.1093/bioinformatics/btg085
Venet D MatArray: a Matlab toolbox for microarray data Bioinformatics 2003 19 659 660 12651728 10.1093/bioinformatics/btg046
Quackenbush J Microarray data normalization and transformation Nature Genetics 2002 Suppl 32 496 501 10.1038/ng1032
Yang YH Dudoit S Luu P Lin DM Peng V Ngai J Speed TP Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation Nucleic Acids Research 2002 30 e15 11842121 10.1093/nar/30.4.e15
Cleveland WS Devlin SJ Locally weighted regression: an approach to regression analysis by local fitting Journal of the American Statistical Association 1988 83 596 610
Dudoit S Yang YH Callow MJ Speed TP Statistical methods for identifying genes with differential expression in replicated cDNA microarray experiments Statistical Sinica 2002 12 111 139
Yang YH Dudoit S Luu P Speed TP Bittner M, Chen Y, Dorsel A, Dougherty ER Normalization for cDNA microarray data In Microarrays: optical technologies and informatics 2001 4266 San Jose, CA, USA: SPIE 141 152
Bolstad BM Irizarry RA Åstrand M Speed TP A comparison of normalization methods for high density oligonucleotide array data based on variance and bias Bioinformatics 2003 19 185 193 12538238 10.1093/bioinformatics/19.2.185
Edwards D Non-linear normalization and background correction in one-channel cDNA microarray studies Bioinformatics 2003 19 825 833 12724292 10.1093/bioinformatics/btg083
Wilson DL Buckley MJ Helliwell CA Wilson IW New normalization methods for cDNA microarray data Bioinformatics 2003 19 1325 1332 12874043 10.1093/bioinformatics/btg146
Järvinen AK Hautaniemi S Edgren H Auvinen P Saarela J Kallioniemi OP Monni O Are data from different gene expression microarray platforms comparable? Genomics 2004 84 1164 1168 10.1016/j.ygeno.2004.01.004
Supplementary Webpage (Self vs. Self data)
Supplementary Webpage (Breast Cancer data)
Dobbin K Shih JH Simon R Questions and answers on design of dual-label microarrays for identifying differentially expressed genes J Nat Cancer Inst 2003 95 1362 1369 13130111
Duggan DJ Bittner M Chen Y Meltzer P Trent JM Expression profiling using cDNA microarrays Nature Genetics 1999 21 10 14 9915494 10.1038/4434
Finkelstein D Ewing R Gollub J Sterky F Cherry JM Somerville S Microarray data quality analysis: lessons from the AFGC project Plant Molecular Biology 2002 48 119 131 11860205 10.1023/A:1013765922672
Holloway AJ van Laar RK Tothill RW Bowtell DDL Options available – from start to finish – for obtaining data from DNA microarrays II Nature Genetics 2002 32 481 489 12454642 10.1038/ng1030
Fan J Gijbels I Local Polynomial Modelling and its Applications 1996 London: Chapman and Hall
Forsythe GE Malcolm MA Moler CB Computer Methods for Mathematical Computations 1977 Englewood Cliffs, NJ: Prentice-Hall, Inc
Chen Y Kamat V Dougherty ER Bittner ML Meltzer PS Trent JM Ratio statistics of gene expression levels and applications to microarray data analysis Bioinformatics 2002 18 1207 1215 12217912 10.1093/bioinformatics/18.9.1207
Hautaniemi S Edgren H Vesanen P Wolf M Järvinen AK Yli-Harja O Astola J Kallioniemi O Monni O A Novel Strategy for Microarray Quality Control Using Bayesian Networks Bioinformatics 2003 19 2031 2038 14594707 10.1093/bioinformatics/btg275
Lee M Kuo F Whitmore G Sklar J Importance of replication in microarray gene expression studies: statistical methods and evidence from repetitive cDNA hybridizations Proc Natl Acad Sci USA 2000 97 9834 9839 10963655 10.1073/pnas.97.18.9834
Yang YH Speed TP Design issues for cDNA microarray experiments Nature Reviews Genetics 2002 3 579 588 12154381
Eilers PHC Goeman JJ Enhancing scatterplots with smoothed densities Bioinformatics 2004 20 623 628 15033868 10.1093/bioinformatics/btg454
Bevington PR Robinson DK Data Reduction and Error Analysis for the Physical Sciences 1992 2 Boston, MA: WCB/McGraw-Hill
Hyman E Kauraniemi P Hautaniemi S Wolf M Mousses S Rozenblum E Ringnér M Sauter G Monni O Elkahloun A Kallioniemi OP Kallioniemi A Impact of DNA amplification on gene expression patterns in breast cancer Cancer Research 2002 62 6240 6245 12414653
Clark J Edwards S Feber A Flohr P John M Giddings I Crossland S Stratton MR Wooster R Campbell C Cooper CS Genome-wide screening for complete genetic loss in prostate cancer by comparative hybridization onto cDNA microarrays Oncogene 2003 22 1247 1252 12606952 10.1038/sj.onc.1206247
Borkowf CB Albert PS Abnet CC Using lowess to remove systematic trends over time in predictor variables prior to logistic regression with quantile categories Statistics in Medicine 2003 22 1477 1493 12704611 10.1002/sim.1507
Mazerolle M Detrimental effects of peat mining on amphibian abundance and species richness in bogs Biological Conservation 2003 113 215 223 10.1016/S0006-3207(02)00360-9
Hen I Sakov A Kafkafi N Golani I Benjamini Y The dynamics of spatial behavior: how can robust smoothing techniques help? Journal of Neuroscience Methods 2004 133 161 172 14757357 10.1016/j.jneumeth.2003.10.013
Monni O Bärlund M Mousses S Kononen J Sauter G Heiskanen M Paavola P Avela K Chen Y Bittner M Kallioniemi A Comprehensive copy number and gene expression profiling of the 17q23 amplicon in human breast cancer Proc Natl Acad Sci 2001 98 5711 5716 11331760 10.1073/pnas.091582298
| 15588297 | PMC539276 | CC BY | 2021-01-04 16:02:47 | no | BMC Bioinformatics. 2004 Dec 9; 5:194 | utf-8 | BMC Bioinformatics | 2,004 | 10.1186/1471-2105-5-194 | oa_comm |
==== Front
BMC Cell BiolBMC Cell Biology1471-2121BioMed Central London 1471-2121-5-461559601010.1186/1471-2121-5-46Research ArticleAltered protein dynamics of disease-associated lamin A mutants Gilchrist Susan [email protected] Nick [email protected] Paul [email protected]Östlund Cecilia [email protected] Howard J [email protected] Wendy A [email protected] MRC Human Genetics Unit, Crewe Road, Edinburgh EH4 2XU, UK2 Departments of Medicine and of Anatomy and Cell Biology, College of Physicians, Columbia University, New York, NY 10032, USA2004 13 12 2004 5 46 46 3 9 2004 13 12 2004 Copyright © 2004 Gilchrist et al; licensee BioMed Central Ltd.2004Gilchrist et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Recent interest in the function of the nuclear lamina has been provoked by the discovery of lamin A/C mutations in the laminopathy diseases. However, it is not understood why mutations in lamin A give such a range of tissue-specific phenotypes. Part of the problem in rationalising genotype-phenotype correlations in the laminopathies is our lack of understanding of the function of normal and mutant lamin A. To investigate this we have used photobleaching in human cells to analyse the dynamics of wild-type and mutant lamin A protein at the nuclear periphery.
Results
We have found that a large proportion of wild-type lamin A at the nuclear periphery is immobile, but that there is some slow movement of lamin A within the nuclear lamina. The mobility of an R482W mutant lamin A was indistinguishable from wild-type, but increased mobility of L85R and L530P mutant proteins within the nuclear lamina was found. However, the N195K mutant shows the most enhanced protein mobility, both within the nucleoplasm and within the lamina.
Conclusion
The slow kinetics of lamin A movement is compatible with its incorporation into a stable polymer that only exchanges subunits very slowly. All of the myopathy-associated lamin A mutants that we have studied show increased protein movement compared with wild-type. In contrast, the dynamic behaviour of the lipodystrophy-associated lamin A mutant was indistinguishable from wild-type. This supports the hypothesis that the underlying defect in lamin A function is quite distinct in the laminopathies that affect striated muscle, compared to the diseases that affect adipose tissue. Our data are consistent with an alteration in the stability of the lamin A molecules within the higher-order polymer at the nuclear lamina in myopathies.
==== Body
Background
The nuclear lamina is a filamentous network of lamin proteins that underlies the inner nuclear membrane (INM). It is thought to make connections between both integral membrane proteins of the INM, and chromatin. It may therefore play a fundamental role in the functional organisation of the nucleus. Lamins are type V intermediate filament (IF) proteins, consisting of a central coiled-coil region, and globular N-terminal and C-terminal domains. The N-terminal domain has a nuclear localisation signal (NLS) and most lamins, except for lamin C, are farnesylated at their carboxy termini via a CaaX motif [1] (Figure 1A). The mammalian genome contains two lamin B genes (LamB1 and 2) and lamins A/C. The latter is alternatively spliced to produce lamins A and C, as well as other minor species. Lamin B is expressed in all cell types and is essential for cell viability. A-type lamins are expressed in more differentiated cells [2] and are non-essential for cell viability [3].
Figure 1 Structure of lamin A protein.A) Diagram of lamin A amino acid sequence showing the domains of the protein, and the position of the four laminopathy-associated missense mutations in DCM, FPLD and AD-EDMD. B) Structure of the C-terminal globular domain of Lamin A showing the relative positions of the FPLD associated R482W missense mutation and the AD-EDMD associated L530P mutation. (Adapted with permission from [25]).
Lamins readily form parallel coiled-coil dimers, which then associate into larger polymers. However, whereas cytoplasmic IF proteins assemble in vitro into 10 nm filaments that resemble those formed in vivo, lamins assemble in vitro into paracrystalline arrays rather than filaments [4]. This suggests that, in vivo, assembly of correct lamin higher-order structures requires the interaction with other molecules/proteins. Lamin A certainly has the ability to interact with other proteins, and also to influence their localisation. In the absence of lamin A, emerin relocates from the INM to the endoplasmic reticulum [3,5,6]. The interaction domain with emerin is in the C-terminal domain of lamin A [7,8]. The coiled-coil region can interact with chromatin [9,10] (Figure 1A). There is also an interaction between lamins A/C and the INM proteins LAP2β [11] and muscle-specific nesprin1 [12].
In addition to its localisation at the nuclear lamina, lamin A is also found within the nucleoplasm where it might interact with other nuclear proteins. Interaction and/or co-localisation between lamin A and; Rb, mRNA splicing factors, LAP2, sites of early DNA replication, and specific transcription factors have been reported [13-18].
Recent interest in the function of the nuclear lamina has been provoked by the discovery of lamin A/C mutations in several human diseases, termed the laminopathies [reviewed in [19]]. What is striking about these diseases is that so many apparently disparate phenotypes arise from mutations in one widely expressed gene. The overt phenotypes of the laminopathies can be grouped according to the major cell types that are affected. Striated (skeletal and cardiac) muscle is affected in autosomal dominant Emery-Dreifuss muscular dystrophy (AD-EDMD), limb girdle muscular dystrophy type 1 (LGMD-1B), and dilated cardiomyopathy (DCM). Adipose and bone tissues are affected in familial partial lipodystrophy (FPLD) and mandibuloacral dysplasia (MAD). Charcot-Marie-Tooth neuropathy type 2B1 (CMT2B1) is a demyelination disease of peripheral neurons. Lastly, Hutchinson-Gilford Progeria Syndrome (HGPS) [20,21] and atypical Werner's Syndrome [22] affect multiple tissue types, including many of those involved in the other laminopathies (muscle, fat, bone), and also results in some premature ageing phenotypes. There are currently three main hypotheses for laminopathy disease mechanisms – nuclear weakness, altered nuclear-cytoskeletal interactions, or changes in gene expression [19,23].
To understand the disease pathology of the laminopathies it will be necessary to better characterise the properties of mutant lamin As. The mutations in AD-EDMD are distributed throughout the coiled-coil domain and the first half of the C-terminal globular domain of lamin A. LGMD and DCM appear to be caused mainly by mutations in the coiled-coil domain [19], although an R571S mutation at the end of the globular domain, that affects only lamin C, has been found in a mild case of DCM. [24].
In contrast, FPLD and MAD mutations cluster tightly within part of the C-terminal globular domain. An explanation for this came from structural analysis of this domain. The residues mutated in FPLD and MAD are on the surface (solvent exposed), whereas residues mutated in other laminopathies are located internally within the hydrophobic core of the domain structure [25,26] (Figure. 1B). The latter mutations may therefore have more profound affects on the structure of the mutant protein, whereas FPLD and MAD mutations may leave the overall structure of the lamin A molecule largely unperturbed but might, for example, interfere with protein-protein interactions.
To better understand the affects of laminopathy-associated mutations on lamin A function we have used fluorescence recovery after photobleaching (FRAP) and fluorescence loss in photobleaching (FLIP) to investigate the protein dynamics of GFP-tagged wild-type and disease-associated mutant lamin As in living cells.
Results and discussion
Expression of mutant lamin A in human cells
To investigate the biological affect of different mutations on lamin A nuclear localisation and dynamics we expressed epitope tagged forms of the protein, carrying disease-associated missense mutations, in human HT1080 cells. The mutations chosen were; L85R (DCM), N195K (DCM), R482W (FPLD), and L530P (AD-EDMD). Although L85R and N195K are both located within the coiled-coil domain of lamin A, and associated with DCM (Figure 1A), they have been shown to have different behaviours when transiently expressed [27,28]. R482W and L530P are associated with different disease phenotypes (lipodystrophy and myopathy, respectively), and although they are both within the globular domain, R482W is a surface residue, whilst L530P is internal (Figure 1B).
Since these mutations are responsible for autosomal dominant forms of disease they should still exert their molecular phenotype in the cell in the presence of wild-type (wt) protein. Both FLAG-tagged and GFP-tagged prelamin As were transiently transfected into human fibrosarcoma cells. Each protein was processed into mature lamin A [29] and incorporated into the nuclear lamina, as evident by the bright nuclear ring of staining visualised either by immunofluorescence with anti-FLAG antibody or from the GFP signal (Figure 2). The mutant forms of lamin A generally had a more uneven distribution at the nuclear periphery, compared to wt, as has been reported previously [30]. We saw high levels of N195K lamin A in the nucleoplasm in addition to the nuclear periphery, but we did not see much evidence for its aggregation into intra-nuclear foci, as has been reported in mouse myoblasts and embryonic fibroblasts [27,28],. This might reflect differences in cell-type or relative expression levels of the mutant protein. Apparently internal sites of epitope-tagged lamin As are seen, but analysis of 3D image stacks (Figure 2B) shows that these are invaginations of the nuclear periphery and not intra-nuclear foci. Such invaginations has previously been reported in many types of cultured cells [31-33].
Figure 2 Sub-cellular localisation of epitope-tagged lamin As.A) Detection of FLAG-tagged wt and mutant lamin As transfected into human HT1080 fibrosarcoma cells. The FLAG tag was detected by immunofluorescence with M2 anti-FLAG (red in merge), in DAPI stained nuclei (blue in merge). Bar = 10 μm. B) Detection of GFP-tagged wt and mutant lamin As transfected into human HT1080 fibrosarcoma cells. GFP signal in images collected at 2 μm intervals from the top to the bottom of the nucleus is shown in black and white. The merged colour images (far right) show mid-plane images of the GFP signal (green) in DAPI stained nuclei (blue). Bar = 10 μm.
Analysis of lamin A dynamics by FRAP
The lamin A mutations that we have studied are within different domains of the protein (Figure 1A), or within different parts of the same structural domain (Figure 1B). Therefore they likely have different interactions, either with other molecules of lamin A, or with other proteins of the nuclear periphery or nucleoplasm. Such interactions affect the kinetic properties of a protein, and photobleaching and time-lapse imaging can probe this [34]. We therefore analysed the mobility of GFP tagged wt and mutant lamin As by FRAP in transiently transfected human cells.
In each case a region at the nuclear lamina was bleached. The fluorescence within a 1.8 × 1.8 μm region of interest (ROI) of this bleach region was then followed every 5 minutes over a period of up to 65 minutes. To calculate the loss of fluorescence attributed to the imaging process alone, the sum of pixel intensities was also calculated for a control (unbleached) cell in each case. This was used to normalise the fluorescence intensity for each ROI [35]. The mean relative fluorescence intensity for each time point was then calculated for 9 cells of each of the GFP-lamin A proteins (WT, L85R, N195K, R482W and L530P).
For wild-type lamin A, fluorescence at the nuclear lamina is visibly bleached (t = 0 in Figure 3A), and only about 20% of the signal recovers over the time course of the experiment (Figure 3C). This indicates that a large proportion (~80%) of lamin A at the nuclear periphery is immobile, at least within the time-frame of these experiments. This is similar to the reported immobility of 60% of lamin B receptor (LBR) in the INM [36]. The recovery curve shows that wt lamin A moves back into the bleach area only very slowly (Figure 3C). The extrapolated t1/2 is ~140 minutes, similar to that reported for lamin B1 (>180 mins) [37]. GFP-tagged lamin C expressed in CHO cells has also been reported to show very little recovery after 1 hour [33]. Most nuclear proteins e.g. transcription factors, and even chromatin-associated proteins such as HP1, are very dynamic with t1/2 values in the range of a few seconds [38]. Even the INM proteins emerin, Lap2β, and Man1 have recovery halftimes of about 1 minute [39]. The slow recovery of lamin A is compatible with its incorporation into a stable polymer that only exchanges subunits very slowly.
Figure 3 FRAP analysis of wild type and mutant lamin As. A and B) Single z-plane confocal images of GFP-tagged (A) wt and (B) N195K lamin A expressing cells. Images were captured before (t = -5) and immediately after (t = 0) photobleaching of an area of the nuclear periphery, and at 5 min intervals thereafter. The bleach region is boxed in red. C) Graphs of mean (± s.e.m) relative fluorescence in the bleach area during FRAP, averaged over 9 cells each. In each graph, data for wt (black) and a mutant (red) lamin A are compared.
The recovery kinetics for the R482W lamin A mutant are indistinguishable from wt and the extrapolated t1/2 = 145 mins (Figure 3C). However, the other lamin A mutants analysed show significant differences. The L85R and L530P mutant proteins appear to be more mobile than wild-type lamin A. They recover more rapidly: t1/2 L85R = 75 mins, L530P = 80 mins. Compared to wt, a higher proportion of the L85R fluorescence (35%) also recovers, suggesting that less of this mutant lamin A is in an immobile fraction.
The most dramatic difference in dynamics was seen for the N195K mutant. Compared to the other lamin As it does not bleach to the same extent, and this is attributable to rapid diffusion of the high levels of nucleoplasmic protein, since at t = 0 recovery of fluorescence can be seen in the nucleoplasmic part of the bleach region, but not in the nuclear periphery itself (Figure 3B). It is known that in early G1 cells the nucleoplasmic pools of lamin A recovery their fluorescence immediately following bleaching [37]. However, even within the nuclear periphery fluorescence recovers within the observation period (Figure 3B, t = 15) and the t1/2 = 30 mins (Figure 3C). Therefore the N195K lamin A mutant is considerably more mobile within the nuclear lamina than wt lamin A, or indeed the other lamin A mutants studied here.
Analysis of lamin A dynamics by FLIP
To further analyse the movement of lamin A within the nuclear lamina, and between the lamina and the nucleoplasm, FLIP experiments were performed on wt, and N195K and L530P mutant GFP-lamin A expressing cells. After successive rounds of photobleaching at a region of the nuclear periphery, the fluorescence at a region of the nuclear periphery distant from the bleach, and at a region within the nucleoplasm were measured (Figure 4) in 10 cells each. As in FRAP, the data was normalised for the loss of fluorescence caused by the successive rounds of imaging.
Figure 4 FLIP analysis of wild type and mutant lamin As. A) Single z-plane confocal images of a GFP-tagged wt lamin A expressing cell captured before (left) and immediately after (right) a round of photobleaching of an area of the nuclear periphery (red box). Fluorescence was also recorded for an unbleached area (blue box) of the nuclear periphery, and a region of the nucleoplasm (green box). Bar = 10 μm B) Graphs of mean (± s.e.m) relative fluorescence in the bleach area (red) during successive rounds of FLIP, and in unbleached regions of the nuclear periphery (blue). and the nucleoplasm (green). Data are averaged over 10 cells each for wt lamin A and for the L530P and N195K mutant lamin As.
For both wt and L530P lamin A there is little loss of fluorescence from either a distant region of the nuclear periphery, or the nucleoplasm after repeated rounds of photobleaching (Figure 4B). This reflects the slow FRAP recovery kinetics of these forms of lamin A (Figure 3). In contrast, the nucleoplasmic fraction of the N195K mutant lamin A shows a substantial decrease (24%) in fluorescence after successive rounds of bleaching at the nuclear periphery. This may reflect diffusion into the small region of nucleoplasm contained within the bleach region, but could also be due to exchange of protein between the nucleoplasm and the lamina. A 10% decrease in fluorescence is also seen at a non-bleached part of the nuclear periphery. This suggests that there is enhanced lateral movement of mutant lamin A within the nuclear lamina compared to wild-type protein.
Conclusions
For GFP-tagged wild-type lamin A we have determined that a large proportion of the protein at the nuclear periphery is immobile (Figure 3), and that any recovery of fluorescence that does occur there is very slow (t1/2 = ~140 mins). This is consistent with the slow recovery halftimes of lamin B1 [37], and the incorporation of lamin A into a stable IF polymer at the nuclear lamina.
Of the four laminopathy-associated mutant forms of lamin A studied by photobleaching all, except for R482, show altered dynamics relative to wt protein. The R482W mutation is associated with FPLD, and other lamin A mutations found in this disease are also either a loss of positive charge at R482, or K486, or the gain of a negative charge (G465D). The amino acid residues involved all map to a solvent-exposed surface in the structure of the Ig-like C-terminal domain [26] (Figure 1B). By NMR and circular dichroism the structure and thermostability of the R482W mutant is similar to that of wt lamin A [26]. Our analysis suggests that the dynamics of the R482W mutant protein within the cell are also similar to wt. It has been suggested that FPLD-associated mutations of lamin A do not destabilise the Ig-like domain of lamin A, but may alter the interaction of the protein with other cellular components. The Ig-like lamin A domain interacts with LAP2α [16], emerin [28], DNA [10] and SREBP1 [18]. Emerin can still interact with R482W lamin A [16], though altered emerin-lamin A interactions have been reported for the R482L mutation [40]. Mutations at R482 have a 5-fold lower affinity for DNA binding in in vitro assays [10], and a slightly lower affinity for SREBP1 [18]. We suggest that if lamin A-protein or -DNA interactions are perturbed by the R482W mutation they are not sufficient to affect the dynamics of lamin A movement within the nucleus.
The EDMD-associated L530P mutation is also within the Ig-like domain (Figure 1), but unlike R482W it is located inside of the structure and so is predicted to destabilise protein folding [25,26]. Compared with wild-type and R482W lamin A, we detected increased mobility of L530P lamin A within the nuclear lamina by FRAP (Figure 3). Expression of L530P has been reported to result in decreased emerin localisation at the INM [30]. Therefore, the stability of both emerin and lamin A at the nuclear periphery may be mutually dependent. In the absence of lamin A, emerin completely fails to localise at the INM [3,5,6]. Our analysis of protein dynamics suggests that an altered interaction between emerin and lamin A could alter the stability of the nuclear lamina, reflected in the increased mobility of lamin A.
Missense mutations in the coiled-coil domain of lamin A are associated with the myopathies, not FPLD (Figure 1). They likely impair the dimerization and formation of higher-order filaments of lamin A. The increased mobility of the L85R mutant lamin A, as assayed by FRAP (Figure 3), would be consistent with this. The most dramatic change in lamin A dynamics was seen with the N195K form. FRAP indicates that it is considerably more mobile than wt lamin A (Figure 3). FLIP suggests that there might be exchange between the nucleoplasmic and lamina pools of this mutant protein, as well as enhanced mobility within the nuclear lamina (Figure 4). Like L530P, this mutation is also in the coiled-coil domain, but clearly has a more drastic affect on lamin polymerisation and intra-nuclear dynamics.
Given the genetically dominant nature of many of the laminopathies, it would be interesting to determine whether the presence of a mutant lamin A has an affect on the mobility of the (GFP-tagged) wild-type protein in the same cells.
FLPD is clinically distinct from AD-EDMD and DCM. Patients with FPLD do not have striated muscle pathology, conversely adipose tissue is normal in AD-EDMD and DCM. Whereas we find increased mobility of all the myopathy-associated lamin A mutants we studied, we cannot distinguish between the protein dynamics of wt and an FLPD mutant form of lamin A (R482W). We conclude that in AD-EDMD and DCM laminopathies the structure of the nuclear lamina is perturbed in such a way as to allow for more rapid exchange of lamin A molecules. In contrast, we suggest that, in this respect, the structure of the lamin polymer is normal in FPLD.
Methods
Generation of GFP-tagged lamin A constructs
Green Fluorescent Protein (GFP)-tagged human lamin A cloned in pCDNA-3-EGFP (GFP-HLA) was obtained from L Karnitz (Mayo Clinic, Rochester). GFP-tagged mutant lamin As were then generated by transfer from N-terminal FLAG-tagged fusion constructs [27]. GFP-HLA was digested with AccI/EcoRI and fragments of 5.4 Kb (the vector backbone) and 894 bp (GFP coding sequence plus 0–175 bp of lamin A) were purified. FLAG-pre-lamin A coding sequences (carrying laminopathy mutations) were then excised from each of the pSVK vectors using EcoRI/SalI, and cloned into XhoI/EcoRI digested pUC21. A 1.8 kb AccI/SpeI fragment of the coding sequence (removing the FLAG tag and the first 175 bp of lamin A) was purified. The AccI site is upstream of each mutant codon and the SpeI site is downstream of the stop codon. A three way ligation was then performed of this fragment together with the 5.4 Kb and 894 bp AccI/EcoRI fragments from GFP-HLA.
Cell transfection and immunofluorescence
Human HT1080 fibrosarcoma cells were transfected with plasmids using Lipofectamine™2000 according to the manufacturer's recommendations. Cells grown on glass slides were fixed 24 hours later for immunofluorescence or GFP analysis. Cells were fixed for 10 mins in 4% paraformaldehyde and permeabilised for 10 mins in 0.05% Triton-X100. FLAG-tagged proteins were detected with 1:200 dilution of M2 anti-Flag mouse monoclonal antibody (Sigma) and 1:100 anti-mouse Texas Red Fab'2 heavy and light chains (Jackson Labs). Slides were counterstained with DAPI and analysed using a Zeiss Axioplan microscope fitted with a Xillig CCD camera and a focus motor to collect images at 0.5 μm intervals in the z plane [41].
Live cell analysis
Cells were grown on DeltaT 0.17 mm culture dishes (Bioptechs Inc) and were mounted onto a heated stage (Bioptechs Inc) on a Zeiss LSM510 confocal microscope. An objective warmer (Bioptechs Inc) was also used to help to maintain a stable temperature of the medium in the culture dish.
Photobleaching
For FRAP, a 1.8 × 1.8 μm region at the nuclear periphery in the mid-focal plane was bleached with 100 iterations at 100% power of the argon laser running at 6.1 mA (50% power). The pinhole size for the confocal was set at 1 Airy unit. The time series software option was used to specify the appropriate time delay between rounds of 3D image stack capture. Each bleached cell was imaged with a ×100 objective, in a window that included other non-bleached cells to allow for relative fluorescence levels to be normalised. Immediately following the bleach, images in the same z-plane were captured at 1s intervals (t = 0 in Figure 3). Thereafter 3D z-plane stacks were captured of the cell at 5 min intervals for a further 65 mins, using 8% of laser power.
Because of the length of FRAP analysis, nuclear rotation, cell movement and focus drift presented a problem in registering the bleach ROI between time points. To account for this, the best z-plane image for the bleach ROI was selected from each time point 3D stack. Each of these was then processed by an interactive rotation script (v3.6 IPLAB, Scanalytics) to correct for nuclear rotation and cell movement. This enabled all images to be superimposed with the pre-bleach image.
For FLIP, an ROI at the nuclear periphery was bleached with 10 laser iterations at 100% of 50% total laser output (~6.1 mA). Following the bleach, 5 images were taken at 2 sec intervals using 8% of laser output. The bleach procedure was repeated for 16 rounds.
In both FRAP and FLIP, the loss of fluorescence attributed to the imaging process alone was assessed from the sum of pixel intensities in a control (unbleached) cell, in each analysis. The relative fluorescence intensity over time was calculated for each defined ROI using a normalisation equation [35].
Authors' contributions
SG constructed the GFP-tagged lamin A constructs, did the cell transfections, the fluorescence microscopy and the photobleaching analysis. NG gave assistance and advice in the photobleaching studies. PP advised and assisted in confocal microscopy and wrote the scripts for image registration over the time course of FRAP. CO and HJW constructed the FLAG-tagged lamin A mutants and provided advice. WAB conceived of the study and drafted the manuscript. All authors read and approved the final manuscript
Acknowledgements
We thank Larry Karnitz (Mayo Clinic, Rochester) for the gift of pcDNA-EGFP-HLA. SG was funded by a PhD studentship form the Medical Research Council (UK). WAB is a Centennial fellow of the James S. McDonnell foundation.
==== Refs
Hutchison CJ Lamins: building blocks or regulators of gene expression? Nat Rev Mol Cell Biol 2002 3 848 858 12415302 10.1038/nrm950
Rober RA Sauter H Weber K Osborn M Cells of the cellular immune and hemopoietic system of the mouse lack lamins A/C: distinction versus other somatic cells J Cell Sci 1990 95 ( Pt 4) 587 598 2200797
Sullivan T Escalante-Alcalde D Bhatt H Anver M Bhat N Nagashima K Stewart CL Burke B Loss of A-type lamin expression compromises nuclear envelope integrity leading to muscular dystrophy J Cell Biol 1999 147 913 920 10579712 10.1083/jcb.147.5.913
Stuurman N Heins S Aebi U Nuclear lamins: their structure, assembly, and interactions J Struct Biol 1998 122 42 66 9724605 10.1006/jsbi.1998.3987
Harborth J Elbashir SM Bechert K Tuschl T Weber K Identification of essential genes in cultured mammalian cells using small interfering RNAs J Cell Sci 2001 114 4557 4565 11792820
Vaughan A Alvarez-Reyes M Bridger JM Broers JL Ramaekers FC Wehnert M Morris GE Whitfield WGF Hutchison CJ Both emerin and lamin C depend on lamin A for localization at the nuclear envelope J Cell Sci 2001 114 2577 2590 11683386
Clements L Manilal S Love DR Morris GE Direct interaction between emerin and lamin A Biochem Biophys Res Commun 2000 267 709 714 10673356 10.1006/bbrc.1999.2023
Sakaki M Koike H Takahashi N Sasagawa N Tomioka S Arahata K Ishiura S Interaction between emerin and nuclear lamins J Biochem (Tokyo) 2001 129 321 327 11173535
Glass CA Glass JR Taniura H Hasel KW Blevitt JM Gerace L The alpha-helical rod domain of human lamins A and C contains a chromatin binding site EMBO J 1993 12 4413 4424 8223451
Stierle V Couprie J Ostlund C Krimm I Zinn-Justin S Hossenlopp P Worman HJ Courvalin JC Duband-Goulet I The carboxyl-terminal region common to lamins A and C contains a DNA binding domain Biochemistry 2003 42 4819 4828 12718522 10.1021/bi020704g
Lang C Krohne G Lamina-associated polypeptide 2beta (LAP2beta) is contained in a protein complex together with A- and B-type lamins Eur J Cell Biol 2003 82 143 153 12691263
Mislow JM Kim MS Davis DB McNally EM Myne-1, a spectrin repeat transmembrane protein of the myocyte inner nuclear membrane, interacts with lamin A/C J Cell Sci 2002 115 61 70 11801724
Ozaki T Saijo M Murakami K Enomoto H Taya Y Sakiyama S Complex formation between lamin A and the retinoblastoma gene product: identification of the domain on lamin A required for its interaction Oncogene 1994 9 2649 2653 8058329
Jagatheesan G Thanumalayan S Muralikrishna B Rangaraj N Karande AA Parnaik VK Colocalization of intranuclear lamin foci with RNA splicing factors J Cell Sci 1999 112 ( Pt 24) 4651 4661 10574713
Kumaran RI Muralikrishna B Parnaik VK Lamin A/C speckles mediate spatial organization of splicing factor compartments and RNA polymerase II transcription J Cell Biol 2002 159 783 793 12473687 10.1083/jcb.200204149
Dechat T Korbei B Vaughan OA Vlcek S Hutchison CJ Foisner R Lamina-associated polypeptide 2alpha binds intranuclear A-type lamins J Cell Sci 2000 113 Pt 19 3473 3484 10984438
Kennedy BK Barbie DA Classon M Dyson N Harlow E Nuclear organization of DNA replication in primary mammalian cells Genes Dev 2000 14 2855 2868 11090133 10.1101/gad.842600
Lloyd DJ Trembath RC Shackleton S A novel interaction between lamin A and SREBP1: implications for partial lipodystrophy and other laminopathies Hum Mol Genet 2002 11 769 777 11929849 10.1093/hmg/11.7.769
Mounkes L Kozlov S Burke B Stewart CL The laminopathies: nuclear structure meets disease Curr Opin Genet Dev 2003 13 223 230 12787783 10.1016/S0959-437X(03)00058-3
Eriksson M Brown WT Gordon LB Glynn MW Singer J Scott L Erdos MR Robbins CM Moses TY Berglund P Dutra A Pak E Durkin S Csoka AB Boehnke M Glover TW Collins FS Recurrent de novo point mutations in lamin A cause Hutchinson-Gilford progeria syndrome Nature 2003 423 293 298 12714972 10.1038/nature01629
Sandre-Giovannoli A Bernard R Cau P Navarro C Amiel J Boccaccio I Lyonnet S Stewart CL Munnich A Le Merrer M Levy N Lamin a truncation in Hutchinson-Gilford progeria Science 2003 300 2055 12702809 10.1126/science.1084125
Chen L Lee L Kudlow BA Dos Santos HG Sletvold O Shafeghati Y Botha EG Garg A Hanson NB Martin GM Mian IS Kennedy BK Oshima J LMNA mutations in atypical Werner's syndrome Lancet 2003 362 440 445 12927431 10.1016/S0140-6736(03)14069-X
Burke B Stewart CL Life at the edge: the nuclear envelope and human disease Nat Rev Mol Cell Biol 2002 3 575 585 12154369 10.1038/nrm879
Fatkin D MacRae C Sasaki T Wolff MR Porcu M Frenneaux M Atherton J Vidaillet HJJ Spudich S De Girolami U Seidman JG Seidman C Muntoni F Muehle G Johnson W McDonough B Missense mutations in the rod domain of the lamin A/C gene as causes of dilated cardiomyopathy and conduction-system disease N Engl J Med 1999 341 1715 1724 10580070 10.1056/NEJM199912023412302
Dhe-Paganon S Werner ED Chi YI Shoelson SE Structure of the globular tail of nuclear lamin J Biol Chem 2002 277 17381 17384 11901143 10.1074/jbc.C200038200
Krimm I Ostlund C Gilquin B Couprie J Hossenlopp P Mornon JP Bonne G Courvalin JC Worman HJ Zinn-Justin S The Ig-like structure of the C-terminal domain of lamin A/C, mutated in muscular dystrophies, cardiomyopathy, and partial lipodystrophy Structure (Camb ) 2002 10 811 823 12057196 10.1016/S0969-2126(02)00777-3
Ostlund C Bonne G Schwartz K Worman HJ Properties of lamin A mutants found in Emery-Dreifuss muscular dystrophy, cardiomyopathy and Dunnigan-type partial lipodystrophy J Cell Sci 2001 114 4435 4445 11792809
Holt I Ostlund C Stewart CL Man N Worman HJ Morris GE Effect of pathogenic mis-sense mutations in lamin A on its interaction with emerin in vivo J Cell Sci 2003 116 3027 3035 12783988 10.1242/jcs.00599
Sinensky M Fantle K Trujillo M McLain T Kupfer A Dalton M The processing pathway of prelamin A J Cell Sci 1994 107 ( Pt 1) 61 67 8175923
Raharjo WH Enarson P Sullivan T Stewart CL Burke B Nuclear envelope defects associated with LMNA mutations cause dilated cardiomyopathy and Emery-Dreifuss muscular dystrophy J Cell Sci 2001 114 4447 4457 11792810
Bridger JM Kill IR O'Farrell M Hutchison CJ Internal lamin structures within G1 nuclei of human dermal fibroblasts J Cell Sci 1993 104 ( Pt 2) 297 306 8505362
Fricker M Hollinshead M White N Vaux D Interphase nuclei of many mammalian cell types contain deep, dynamic, tubular membrane-bound invaginations of the nuclear envelope J Cell Biol 1997 136 531 544 9024685 10.1083/jcb.136.3.531
Broers JL Machiels BM van Eys GJ Kuijpers HJ Manders EM van Driel R Ramaekers FC Dynamics of the nuclear lamina as monitored by GFP-tagged A-type lamins J Cell Sci 1999 112 ( Pt 20) 3463 3475 10504295
Lippincott-Schwartz J Altan-Bonnet N Patterson GH Photobleaching and photoactivation: following protein dynamics in living cells Nat Cell Biol 2003 Suppl S7 14 14562845
Phair RD Misteli T High mobility of proteins in the mammalian cell nucleus Nature 2000 404 604 609 10766243 10.1038/35007077
Ellenberg J Siggia ED Moreira JE Smith CL Presley JF Worman HJ Lippincott-Schwartz J Nuclear membrane dynamics and reassembly in living cells: targeting of an inner nuclear membrane protein in interphase and mitosis J Cell Biol 1997 138 1193 1206 9298976 10.1083/jcb.138.6.1193
Moir RD Yoon M Khuon S Goldman RD Nuclear lamins A and B1: different pathways of assembly during nuclear envelope formation in living cells J Cell Biol 2000 151 1155 1168 11121432 10.1083/jcb.151.6.1155
Phair RD Scaffidi P Elbi C Vecerova J Dey A Ozato K Brown DT Hager G Bustin M Misteli T Global nature of dynamic protein-chromatin interactions in vivo: three-dimensional genome scanning and dynamic interaction networks of chromatin proteins Mol Cell Biol 2004 24 6393 6402 15226439 10.1128/MCB.24.14.6393-6402.2004
Shimi T Koujin T Segura-Totten M Wilson KL Haraguchi T Hiraoka Y Dynamic interaction between BAF and emerin revealed by FRAP, FLIP, and FRET analyses in living HeLa cells J Struct Biol 2004 147 31 41 15109603 10.1016/j.jsb.2003.11.013
Capanni C Cenni V Mattioli E Sabatelli P Ognibene A Columbaro M Parnaik VK Wehnert M Maraldi NM Squarzoni S Lattanzi G Failure of lamin A/C to functionally assemble in R482L mutated familial partial lipodystrophy fibroblasts: altered intermolecular interaction with emerin and implications for gene transcription Exp Cell Res 2003 291 122 134 14597414 10.1016/S0014-4827(03)00395-1
Gilchrist S Gilbert N Perry P Bickmore WA Nuclear organization of centromeric domains is not perturbed by inhibition of histone deacetylases Chromosome Res 2004 12 505 516 15254368 10.1023/B:CHRO.0000034892.64739.ff
| 15596010 | PMC539277 | CC BY | 2021-01-04 16:31:37 | no | BMC Cell Biol. 2004 Dec 13; 5:46 | utf-8 | BMC Cell Biol | 2,004 | 10.1186/1471-2121-5-46 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-5-501557920410.1186/1471-2202-5-50Research ArticleEffects of the group I metabotropic glutamate receptor agonist, DHPG, and injection stress on striatal cell signaling in food-restricted and ad libitum fed rats Pan Yan [email protected] Yemiliya [email protected] Kenneth D [email protected] Department of Psychiatry, Millhauser Laboratories, New York University School of Medicine, 550 First Ave, New York, NY 10016, USA2 Department of Pharmacology, New York University School of Medicine, 550 First Ave, New York, NY 10016, USA2004 3 12 2004 5 50 50 1 10 2004 3 12 2004 Copyright © 2004 Pan 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
Chronic food restriction augments the rewarding effect of centrally administered psychostimulant drugs and this effect may involve a previously documented upregulation of D-1 dopamine receptor-mediated MAP kinase signaling in nucleus accumbens (NAc) and caudate-putamen (CPu). Psychostimulants are known to induce striatal glutamate release, and group I metabotropic glutamate receptors (mGluR) have been implicated in the cellular and behavioral responses to amphetamine. The purpose of the present study was to evaluate whether chronic food restriction increases striatal MAP kinase signaling in response to the group I mGluR agonist, DHPG.
Results
Western immunoblotting was used to demonstrate that intracerebroventricular (i.c.v.) injection of DHPG (500 nmol) produces greater activation of ERK1/2 and CREB in CPu and NAc of food-restricted as compared to ad libitum fed rats. Fos-immunostaining induced by DHPG was also stronger in CPu and NAc core of food-restricted relative to ad libitum fed rats. However, i.c.v. injection of saline-vehicle produced greater activation of ERK1/2 and CREB in CPu and NAc of food-restricted relative to ad libitum fed rats, and this difference was not seen when subjects received no i.c.v. injection prior to sacrifice. In addition, although DHPG activated Akt, there was no difference in Akt activation between feeding groups. To probe whether the augmented ERK1/2 and CREB activation in vehicle-injected food-restricted rats are mediated by one or more GluR types, effects of an NMDA antagonist (MK-801, 100 nmol), AMPA antagonist (DNQX, 10 nmol), and group I mGluR antagonist (AIDA, 100 nmol) were compared to saline-vehicle. Antagonist injections did not diminish activation of ERK1/2 or CREB.
Conclusions
These results indicate that a group I mGluR agonist induces phosphorylation of Akt, ERK1/2 and CREB in both CPu and NAc. However, group I mGluR-mediated signaling may not be upregulated in food-restricted rats. Rather, a physiological response to "i.c.v. injection stress" is augmented by food restriction and appears to summate with effects of the group I mGluR agonist in activating ERK1/2 and CREB. While the augmented cellular response of food-restricted rats to i.c.v. injection treatment represents additional evidence of enhanced CNS responsiveness in these subjects, the functional significance and underlying mechanism(s) of this effect remain to be elucidated.
==== Body
Background
Chronic food restriction increases central sensitivity to rewarding and motor-activating effects of psychostimulants and direct dopamine receptor agonists [1]. Corresponding adaptive changes at the cellular level include increased psychostimulant-induced DA release in nucleus accumbens core [2], and upregulation of D-1 DA receptor-mediated MAP kinase signaling in dorsal and ventral striatum, with consequent increased activation of CREB and the immediate early gene, c-fos [3,4]. Changes in striatal glutamate receptor function have not been investigated but are of interest in light of findings that (i) psychostimulants induce striatal glutamate release [5-7], (ii) amphetamine-induced activation of striatal MAP kinase, CREB, and c-fos is attenuated by a group I metabotropic glutamate receptor antagonist [8], and (iii) the augmented effects of striatal D-1 receptor stimulation in food-restricted rats include hyperphosphorylation of the NMDA receptor NR1 subunit [9].
The group I mGluRs (mGluR1/5) are of particular interest. Group I mGluRs are densely expressed in striatal medium spiny neurons [10] and activate phospholipase C, resulting in hydrolysis of phosphoinositides and activation of Ca2+-dependent signaling cascades [11]. The group I mGluR agonist, DHPG, increases the phosphorylation of extracellular signal-regulated kinase (ERK) and the transcription factor CREB when infused into dorsal striatum [12], as do cocaine injected systemically and the D-1 agonist SKF-82958 injected into the lateral ventricle [4,13]. DHPG also elicits hyperlocomotion resembling that induced by DA receptor agonists, and the effect is not attenuated by the D-1 DA receptor antagonist, SCH23390 [10]. It is therefore possible that glutamate and group I mGluR function contribute to the changes in psychostimulant-induced behavioral responses and striatal cell signaling in food-restricted subjects. The purpose of the first experiment of this study was to compare dorsal and ventral striatal cell signaling in response to intracerebroventricular (i.c.v.) administration of the group I mGluR agonist, DHPG, in ad libitum fed and food-restricted rats. A dose of 500 nmol was used based on pilot work and published reports indicating that this dose is sufficient to exert behavioral effects [14] and stimulate PI hydrolysis [11] while being below threshold for producing convulsive behavior [15]. A second experiment was based on results of the first experiment and sought to elucidate the observed increase in striatal ERK1/2 and CREB phosphorylation in food-restricted rats injected i.c.v. with saline-vehicle.
Results
Experiment 1
In both the caudate-putamen and nucleus accumbens ERK1/2 phosphorylation was increased by food restriction (Cpu: F1,18 = 10.8, p < .005; NAc: F1,18 = 8.2, p < .01) and by DHPG injection (Cpu: F1,18 = 6.6, p < .02; NAc: F1,18 = 10.6, p < .005). However, there was no interaction between feeding condition and drug treatment in either brain region (Cpu: F1,18 = 0.6; NAc: F1,18 = 0.3). This analysis supports the impression (see Figures 1 &2, top panel) that food restriction did not actually increase the group I mGluR-mediated response to DHPG; rather, increased signaling in the food-restricted group, irrespective of i.c.v. injection treatment, appears to have summated with DHPG to produce a greater net effect in the food-restricted than ad libitum fed group. An identical pattern of results was obtained for CREB phosphorylation (Figures 1 &2 center panel; Cpu: Fdiet; 1,18 = 6.2, p < .025; Fdrug; 1,18 = 3.9, p = .06; Fdiet × drug; 1,18 = 1.2; NAc: Fdiet; 1,18 = 12.9, p < .0025; Fdrug; 1,18 = 15.3, p < .001; Fdiet × drug; 1,18 = 2.7, p > .10).
Interestingly, Akt phosphorylation was induced by DHPG but did not differ between food-restricted and ad libitum fed groups in either the Cpu (Fdrug; 1,18 = 4.5, p < .05; Fdiet; 1,18 = 0.4; Fdiet × drug; 1,18 = 0.4) or NAc (Fdrug; 1,18 = 12.7, p < .0025; Fdiet; 1,18 = 0.4; Fdiet × drug; 1,18 = 2.4, p > .10; Figures 1 &2 bottom panel).
In previous studies, striatal Fos-immunostaining was greater in food-restricted relative to ad libitum fed rats injected i.c.v. with d-amphetamine or the D-1 DA receptor agonist SKF-82958, but not saline-vehicle [3,16]. Therefore, in a small sample of ad libitum fed and food-restricted subjects, brains were processed for Fos-immunostaining and revealed stronger DHPG-induced staining in CPu (t(7) = 3.6, p < .01) and NAc core (t(7) = 2.8 p < .025) but not NAc shell (t(7) = 1.1) of food-restricted, relative to ad libitum fed, subjects (see Figure 3).
During the period between DHPG injection and sacrifice, rats, which remained in their home cages, did not display any obvious behavioral responses to the drug.
Experiment 2
In rats without lateral ventricular cannulas that received no injection treatment prior to sacrifice, ERK1/2 phosphorylation did not differ between feeding groups (Figure 4; Cpu: t(8) = 1.2; NAc: t(8) = 1.0), nor did CREB phosphorylation (Cpu: t(8) = 1.2; NAc t(8) = 0.1).
In food-restricted rats injected with GluR antagonists prior to sacrifice, neither ERK1/2 nor CREB phosphorylation differed from that observed in food-restricted rats injected with saline-vehicle (Figure 5).
Discussion
Adding to the results of Choe and Wang who demonstrated DHPG-induced activation of ERK1/2 and CREB in dorsal striatum [12], the present study has demonstrated DHPG-induced activation of ERK/12 and CREB in both CPu and NAc following lateral ventricular infusion. While i.c.v. injection allows for the possibility that these effects were not directly mediated by striatal group I mGluRs, the proximity of striatum to the lateral ventrical and the high density of mGluRs in these structures support the likelihood of direct effects. Direct effects are also supported by the observed activation of Akt which mediates group I mGluR signal transduction [17].
As previously seen in rats challenged with the D-1 dopamine receptor agonist, SKF-82958, DHPG-induced activation of ERK1/2, CREB and c-Fos were greater in food-restricted than ad libitum fed rats. However, the pattern of results in this study – i.e. augmented ERK1/2 and CREB activation in food-restricted rats injected with vehicle, and no difference in Akt activation between feeding groups – suggests that group I mGluR signaling was not enhanced. Rather, a cellular-activating effect of the injection procedure was enhanced by food restriction and summated with the effect of DHPG. This response of vehicle-injected subjects was clearly a response to some aspect of the i.c.v. injection procedure because similarly food-restricted rats that were not challenged in any way prior to sacrifice displayed levels of pERK1/2 and pCREB that were essentially identical to those of ad libitum fed rats. Furthermore, the augmented activation of ERK1/2 and CREB in vehicle-injected food-restricted rats was not mediated by Akt because Akt, though activated by DHPG, did not differ between food-restricted and ad libitum fed rats injected with vehicle or DHPG.
Attribution of the augmented DHPG-induced ERK1/2 and CREB phosphorylation to a "nonspecific" response to i.c.v. injection distinguishes the group I mGluR from the D-1 DA receptor. Although a modestly enhanced ERK response to i.c.v. vehicle infusion was seen in the NAc of food-restricted rats in the prior study, the dramatically increased activation of ERK1/2 and CREB by SKF-82958 in CPu and NAc was dissociable from any response to the injection procedure [4]. Even mild stressors are known to stimulate DA [18] and Glu [19,20] release in striatum, and it is possible that food-restriction augments this physiological response or the cell signaling induced by it. Because the ionotropic (AMPA and NMDA) as well as group I mGluRs are abundant in striatum and all three receptor types mediate MAP kinase signaling [12,21,22], corresponding antagonists were injected i.c.v. and compared to vehicle. It was reasoned that if the physiological response to the injection procedure involves one of these GluRs, ERK1/2 and CREB phosphorylation would be decreased, relative to vehicle, in the group(s) receiving the corresponding antagonist. The MK-801 treatment was also a potential probe for mediation by D-1 DA receptors because D-1 DA agonist-induced activation of ERK1/2 and CREB are dependent on the NMDA receptor in food-restricted subjects [9]. Results of this test did not provide support for mediation by one of the GluR types. Because only one dose of each antagonist was used, these results must be considered preliminary. However, considering the proximity of striatum to the lateral ventricle, and the fact that the doses used have exerted measurable cellular, physiological or behavioral effects in other studies [23-26] there is some doubt about involvement of GluRs in the effect of "injection stress".
An interesting possibility to consider is mediation of the response by brain-derived neurotrophic factor (BDNF). Food-restricted rats have elevated striatal BDNF levels which appear to be involved in an enhanced neuroprotective response to diverse insults [27]. BDNF activates striatal ERK1/2 which, unlike glutamate-induced activation of Akt, ERK1/2 and CREB is not blocked by a PI 3-kinase inhibitor [21]. It is therefore possible that an enhanced BDNF-mediated activation of ERK1/2 and CREB in food-restricted rats represents a neuroprotective response to intraventricular infusion.
Conclusions
The present results indicate that a group I mGluR agonist activates Akt, ERK1/2 and CREB in both the CPu and NAc. Further, activation of ERK1/2, CREB, and c-Fos are stronger in food-restricted than in ad libitum fed rats. The augmented response is not attributed to increased Group I mGluR function but, instead, to an augmented response of food-restricted rats to the i.c.v. injection procedure. Some evidence casting doubt on attribution of the latter response to GluRs was obtained. It will be of interest to evaluate whether striatal MAPK signaling in food-restricted rats is generally augmented in response to stressors or whether this response is peculiar to i.c.v. infusion. In addition, it will be of interest to evaluate whether this physiological response is mediated by BDNF and its TrkB receptor.
Methods
Subjects and surgery
All experimental procedures were approved by the New York University School of Medicine Institutional Animal Care and Use Committee and were performed in accordance with the "Principles of Laboratory Animal Care" (NIH publication number 85-23, revised 1996).
Subjects were male Sprague-Dawley rats (375–425 g) housed individually in plastic cages with free access to food and water except when food restriction conditions applied. Animals were maintained on a 12-h light/dark cycle, with lights on at 07:00 h. Rats were anesthetized with ketamine (100 mg/kg; i.p.) and xylazine (10 mg/kg; i.p.) and stereotaxically implanted with a 26-gauge guide cannula (Plastics One, Roanoke, VA USA) in the right lateral ventricle. The cannula was permanently affixed to the skull by flowing dental acrylic around it and four surrounding mounting screws. Patency of the guide cannula was maintained with an occlusion stylet. Several days after surgery, cannula placements were confirmed by demonstration of a vigorous and short latency (i.e. <60 s) drinking response to 50 ng of angiotensin II.
Food restriction and habituation
Seven days following surgery, half the subjects were put on a food restriction regimen whereby a single 10 g meal of Purina (Gray Summit, MO USA) rat chow was delivered at approximately 17:00 h each day. Rats continued to have ad libitum access to water. Once body weight had declined by 20–25% (approximately 15 days) daily food allotments were titrated for an additional week to maintain stable body weight. During the 3 weeks required for food-restricted rats to attain and stabilize at target body weights, all rats were habituated, on five occasions, to the handling and injection procedures to be employed on the terminal day of the experiment.
Drug treatment
In Experiment 1, six food-restricted and six ad libitum fed rats received i.c.v injections of sterile 0.9% saline (5 μl). Six food-restricted and six ad libitum fed rats received i.c.v injections of the group I metabotropic glutamate receptor agonist, DHPG (500 nmol in 5 μl; Tocris Cookson, Ellisville, MO, USA). An additional five food-restricted and four ad libitum fed rats received i.c.v. injections of DHPG and brains were processed for Fos-immunostaining. In Experiment 2, six unoperated food-restricted rats and six unoperated ad libitum fed rats were prepared and habituated as above and received no unusual handling or drug injection on the terminal day of the experiment. In the second part of Experiment 2, twenty four food-restricted rats with lateral ventricular cannulas were prepared and habituated as in Experiment 1. On the terminal day of the experiment, six were injected i.c.v. with saline vehicle (5.0 μl), six were injected with the AMPA glutamate receptor antagonist, DNQX (10 nmol; Sigma-Aldrich, St. Louis, MO), six were injected with the NMDA glutamate receptor antagonist, MK-801 (100 nmol; Sigma-Aldrich), and six were injected with the group I mGluR antagonist, AIDA (100 nmol; Tocris Cookson). Antagonist doses were chosen on the basis of prior findings [e.g. [23,26,28]] and/or being just below threshold for producing noticeable motoric abnormalities. For i.c.v injection, solutions were loaded into a 30 cm length of PE-50 tubing attached at one end to a 250-μl Hamilton syringe filled with distilled water and at the other end to a 33-gauge injector cannula which extended 1.0 mm beyond the implanted guide. The 5.0 μl injection volume was delivered over a period of 95 s. One minute following injection, internal cannulas were removed, stylets replaced, and animals were returned to home cages.
Lysate preparation
Prior studies have indicated that MAP kinase activation is transient and the optimal time-point to study phosphorylation of ERK 1/2 after physiological or pharmacological treatment is 15–20 min [29]. Therefore, all rats, except the twelve unoperated/uninjected rats, were killed 20 min after injection by brief exposure to CO2 followed by decapitation. The twelve unoperated/uninjected rats were simply removed from home cages and exposed to CO2 followed by decapitation. Brains were rapidly removed and immediately frozen in powdered dry ice. Five hundred-micrometer sections were cut using an IEC Minotome cryostat, and CPu and NAc were micropunched, under an Olympus dissecting microscope, from a series of 8 consecutive frozen sections. The tissue was then homogenized in 10 volumes of 50 mM Tris-HCl, pH 7.5 containing 50 mM NaCl, 5 mM EDTA, 1 mM EGTA, 1 mM Na3VO4, 40 mM β-glycerophosphate, 50 mM NaF and 5 mM Na4P2O7, 1% Tx-100, 0.5 μM okadaic acid, 0.5% sodium deoxycholate and 0.1% SDS, followed by centrifugation and protein determination using BCA reagent kit as described by the manufacturer (Pierce) Supernatants were mixed with 5 × SDS-PAGE sample buffer, boiled for 5 min, cooled on ice and kept at -80°C until future use.
Western blotting
Protein (10–30 μg per lane) was separated by electrophoresis on precast 10% polyacrylamide gels (Cambrex, East Rutherford, NJ, USA). Precision Plus protein standard molecular weight markers (Bio-Rad, Hercules, CA, USA) were also loaded to assure complete electrophoretic transfer and to estimate the size of bands of interest. The gels were transferred to nitrocellulose membrane (Osmonics) for 2 h, with a constant voltage of 100 volts. Membranes were blocked for 1 hr at room temperature with blocking buffer, 5% non fat dry milk in 50 mM Tris-HCl, pH 7.5 containing 150 mM NaCl and 0.1% Tween 20 (TBS-T), then probed overnight at 4°C using primary monoclonal antibodies for phospho-(Thr202/Tyr204)-p44/42 ERK1/2 (mouse monoclonal, 1:2000; Cell Signaling, Beverly, MA, USA), or polyclonal antibodies for phospho-Akt (Ser 473) (rabbit polyclonal, 1:1000 dilutuion, Cell Signaling), and phospho (Ser 133) CREB (rabbit polyclonal, 1:2000; Upstate Biotechnology, Lake Placid, NY, USA). Total levels of ERK1/2, Akt and CREB were detected on the same blots using anti-rabbit p42/44 ERK1/2 antibody 1:2000, (Cell Signaling), anti-rabbit total Akt (1:2000 dilution, Cell Signaling), or anti-rabbit CREB antibody (1:2000 dilution, Calbiochem). After detection of phosphorylated ERK1/2, phospho-Akt and CREB blots were stripped with 25 mM Glycine, pH 2.0 containing 1% SDS for 30 min at room temperature, washed six times in TBS-T buffer, blocked in blocking buffer for 1 h and then incubated overnight at 4°C in total ERK1/2, total Akt or CREB antibody. After probing with primary antibodies and washing with TBS-T buffer (3 × 5 min), membranes were incubated with 1:2000 dilution horseradish peroxidase conjugated anti-mouse or 1: 2000 dilution anti-rabbit IgG (Cell Signaling). Proteins were visualized using a chemiluminescense ECL kit (Pierce). Densitometric analysis of the bands was performed using NIH image software. Phospho-p42/44 MAPK, phospho Akt and phospho CREB values were normalized to total p42/44 MAPK, Akt and CREB values respectively.
Immunohistochemistry
Ninety min after DHPG injection rats were anesthetized with sodium pentobarbital (50.0 mg/kg,i.p.) and transcardially perfused with isotonic phosphate buffered saline (PBS) followed by 4% paraformaldehyde in PBS. Brains were then removed and maintained in 20% sucrose at 4°C for 48 h. Forty μm sections were cut on an IEC Minotome cryostat and collected in a cryoprotective solution. Fos immunostaining was carried out using a rabbit polyclonal c-Fos antiserum (Oncogene Science-Calbiochem, La Jolla, CA) and the avidin-biotin peroxidase complex (ABC; Vector laboratories).
Sections were washed in 1% sodium borohydride followed by PBS and incubated for 2 hrs in 4% normal goat serum plus 1% BSA in PBS containing 0.2% Triton X-100 (Sigma-Aldrich) to block nonspecific binding. This was followed by incubation, overnight, with rabbit polyclonal c-fos antiserum (1:5000 dilution). Following several PBS washes, sections were incubated with a secondary antiserum (Vector, Burlingame, CA) for 60 min and subsequently reacted with avidin-biotin complex (ABC) (Vector). The peroxidase reaction was visualized with a chromogen solution containing 100 mM nickel sulfate, 125 mM sodium acetate, 10 mM imidazole, 0.03% diaminobenzidine (DAB), and 0.01% hydrogen peroxide at pH 6.5. Sections were then mounted on chrome-alum coated slides, dehydrated, and coverslipped.
Objective counting of c-Fos positive cells in CPu at coronal levels +1.7 and -0.3 mm, and NAc core and shell at coronal level +1.5 mm in relation to bregma [30] was accomplished with a light microscope (Olympus, CK2) equipped with a Sony XC-77 video camera module coupled to an MCID image analysis system (Imaging Research Inc., St. Catherines, Canada). For each region, bilateral grain counts from three to five consecutive sections were measured to arrive at an average bilateral value per rat. Because results obtained from anterior and posterior levels of CPu were essentially the same, results from the two levels were combined and averaged for each rat.
Data analysis
For each Western blot, film exposure time was set as needed to visualize distinct bands in the control samples of each experiment. Immunoblots were analyzed using NIH imaging software. For each blot, relative phospho-protein levels were calculated from the ratio of optical density of the phosphorylated protein/total protein to correct for small differences in protein loading. In addition, tubulin levels were analyzed in several representative gels and no differences were observed between treatment groups. Results were expressed by comparison to the normalized control, which in Experiment 1 was defined as the ad libitum fed group injected with vehicle. In the first part of Experiment 2, unoperated/uninjected ad libitum fed rats served as control. In the second part of Experiment 2, food-restricted rats injected with i.c.v. saline vehicle served as control. Differences between treatment conditions in Experiment 1 were analyzed by two-way analysis of variance (ANOVA; feeding condition × injection treatment). Differences between treatment conditions in the first and second parts of Experiment 2 were analyzed by student's t-test and one-way ANOVA, respectively.
Authors' contributions
YP conducted the majority of immunoblotting experiments plus the immunohistochemistry experiment, contributed to experimental design and assisted in manuscript preparation. YB provided technical supervision of immunoblotting experiments and assisted in manuscript preparation. KC contributed to design of the study, assisted in all experiments, and wrote the final draft of the manuscript.
Acknowledgements
This work was supported by T32 DA07254 (Y.P) and DA03956 and DA00292 (K.C.) from NIDA/NIH.
Figures and Tables
Figure 1 Effects of DHPG on activation of ERK1/2, CREB and Akt in Caudate-Putamen Ad libitum fed and food-restricted rats received i.c.v. injections of DHPG (500 nmol) or saline vehicle prior to sacrifice. Lysates were immunoblotted with anti-phospho ERK1/2 or anti-ERK1/2 (top), anti-phospho-CREB or anti-CREB (center), or anti-phospho-Akt or anti-Akt (bottom) antibodies. Following densitometry, intensities of bands corresponding to phosphorylated proteins were divided by intensities of the corresponding total protein bands to correct for small differences in protein loading. Results (mean ± S.E.M.) are expressed in comparison to the normalized control, which was defined as the ad libitum fed group injected with vehicle. Graphed results are displayed with representative immunoblots.
Figure 2 Effects of DHPG on activation of ERK1/2, CREB and Akt in Nucleus Accumbens Ad libitum fed and food-restricted rats received i.c.v. injections of DHPG (500 nmol) or saline vehicle prior to sacrifice. Lysates were immunoblotted with anti-phospho ERK1/2 or anti-ERK1/2 (top), anti-phospho-CREB or anti-CREB (center), or anti-phospho-Akt or anti-Akt (bottom) antibodies. Following densitometry, intensities of bands corresponding to phosphorylated proteins were divided by intensities of the corresponding total protein bands to correct for small differences in protein loading. Results (mean ± S.E.M.) are expressed in comparison to the normalized control, which was defined as the ad libitum fed group injected with vehicle. Graphed results are displayed with representative immunoblots.
Figure 3 Effects of DHPG on Fos-like immunostaining in Caudate-Putamen and Nucleus Accumbens core and shell Mean (±S.E.M.) bilateral number of c-Fos immunoreactive (IR) cells within caudate-putamen (CPu) at coronal levels +1.7 and -0.3 mm (combined) and nucleus accumbens (Nac) core and shell at coronal level +1.5 mm in relation to bregma. Ad libitum fed and food-restricted rats were injected i.c.v. with DHPG (500 nmol) 90 min prior to perfusion. *p < .025, **p < .01 compared to ad libitum fed rats. Representative digital photomicrographs depicting Fos-like immunoreactivity in NAc of ad libitum fed (left) and food-restricted (right) rats are included. LV = lateral ventricle; AC = anterior commissure.
Figure 4 Basal levels of phospho-ERK1/2 and -CREB in Caudate-Putamen and Nucleus Accumbens Ad libitum fed and food-restricted rats that were not implanted with i.c.v. cannulas were sacrificed without receiving experimental treatment. Lysates were immunoblotted with anti-phospho ERK1/2 or anti-ERK1/2, and anti-phospho-CREB or anti-CREB antibodies. Following densitometry, intensities of bands corresponding to phosphorylated proteins were divided by intensities of the corresponding total protein bands to correct for small differences in protein loading. Results (mean ± S.E.M.) for caudate-putamen (top) and nucleus accumbens (bottom) are expressed in comparison to the normalized control, which was defined as the ad libitum fed group.
Figure 5 Effects of Glutamate Receptor Antagonists on Activation of ERK1/2 and CREB in Caudate-Putamen and Nucleus Accumbens of Food-Restricted Rats Food-restricted rats received i.c.v. injections of either saline vehicle, the AMPA antagonist DNQX (10 nmol), the group I mGluR antagonist AIDA (100 nmol), or NMDA antagonist MK-801 (100 nmol) prior to sacrifice. Lysates were immunoblotted with anti-phospho ERK1/2 or anti-ERK1/2 (top) and anti-phospho-CREB or anti-CREB (bottom) Following densitometry, intensities of bands corresponding to phosphorylated proteins were divided by intensities of the corresponding total protein bands to correct for small differences in protein loading. Results (mean ± S.E.M.) for caudate-putamen (left in each figure) and nucleus accumbens (right in each figure) are expressed in comparison to the normalized control, which was defined as the group injected with saline vehicle.
==== Refs
Carr KD Augmentation of drug reward by chronic food restriction: Behavioral evidence and underlying mechanisms Physiol Behav 2002 76 353 364 12117572 10.1016/S0031-9384(02)00759-X
Cadoni C Solinas M Valentini V Di Chiara G Selective psychostimulant sensitization by food restriction: differential changes in accumbens shell and core dopamine Eur J Neurosci 2003 18 2326 2334 14622194 10.1046/j.1460-9568.2003.02941.x
Carr KD Tsimberg Y Berman Y Yamamoto N Evidence of increased dopamine receptor signaling in food-restricted rats Neuroscience 2003 119 1157 1167 12831870 10.1016/S0306-4522(03)00227-6
Haberny SL Berman Y Meller E Carr KD Chronic food restriction increases D-1 dopamine agonist-induced phosphorylation of extracellular signal regulated kinase 1/2 and cyclic AMP response element binding protein in caudate-putamen and nucleus accumbens Neuroscience 2004 125 289 298 15051167 10.1016/j.neuroscience.2004.01.037
Del Arco A Gonzalez-Mora JL Armas VR Mora F Amphetamine increases the extracellular concentration of glutamate in striatum of the awake rat: involvement of high affinity transporter mechanisms Neuropharmacology 1999 38 943 954 10428413 10.1016/S0028-3908(99)00043-X
Morari M Marti M Sbrenna S Fuxe K Bianchi C Beani L Reciprocal dopamine-glutamate modulation of release in the basal ganglia Neurochem Int 1998 33 383 397 9874089 10.1016/S0197-0186(98)00052-7
Reid MS Hsu K Berger SP Cocaine and amphetamine preferentially stimulate glutamate release in the rat striatum: an in vivo microdialysis study Synapse 1997 27 95 105 9266771
Choe ES Chung KT Mao L Wang JQ Amphetamine increases phosphorylation of extracellular signal-regulated kinase and transcription factors in the rat striatum via group I metabotropic glutamate receptors Neuropsychopharmacology 2002 27 565 575 12377393
Haberny SL Carr KD Food restriction upregulates D-1 dopamine receptor-mediated signaling in an NMDA receptor-dependent manner in nucleus accumbens [abstract] Soc Neurosci Abstr 2004 574.9
Wang JQ Mao L Sustained behavioral stimulation following selective activation of group I metabotropic glutamate receptors in rat striatum Pharmacol Biochem Behav 2000 65 439 447 10683484 10.1016/S0091-3057(99)00245-2
Johnson MP Chamberlain M Kelly GM Phosphoinositide hydrolysis in vivo with group I metabotropic glutamate receptor agonists Brain Res 1999 821 539 545 10064844 10.1016/S0006-8993(99)01065-3
Choe ES Wang JQ Group I metabotropic glutamate receptor activation increases phosphorylation of cAMP response element-binding protein, Elk-1, and extracellular signal-regulated kinases in rat dorsal striatum Molec Brain Res 2001 94 75 84 11597767 10.1016/S0169-328X(01)00217-0
Valjent E Corvol J-C Pages C Besson M-J Maldonado R Caboche J Involvement of the extracellular signal-regulated kinase cascade for cocaine-rewarding properties J Neurosci 2000 20 8701 8709 11102476
Zalewska-Winska A Wisniewski K Behavioral activity of (S)-3,5-DHPG, a selective agonist of group I metabotropic glutamate receptors Pharmacol Res 2000 42 239 245 10945930 10.1006/phrs.2000.0683
Camon L Vives P Vera de Nuria Martinez E Seizures and neuronal damage induced in the rat by activation of group I metabotropic glutamate receptors with their selective agonist 3,5-dihydroxyphenylglycine J Neurosci Res 1998 51 339 348 9486769
Carr K Kutchukhidze N Chronic food restriction increases Fos-like immunoreactivity (FLI) induced in rat forebrain by intraventricular amphetamine Brain Res 2000 861 88 96 10751568 10.1016/S0006-8993(00)02018-7
Hou L Klann E Activation of the phosphoinositide 3-kinase-Akt-mammalian target of rapamycin signaling pathway is required for metabotropic glutamate receptor-dependent long-term depression J Neurosci 2004 24 6352 6361 15254091 10.1523/JNEUROSCI.0995-04.2004
Inglis FM Moghaddam B Dopaminergic innervation of the amygdala is highly responsive to stress J Neurochem 1999 72 1088 1094 10037480 10.1046/j.1471-4159.1999.0721088.x
Bland ST Gonzale Ra Schallert T Movement-related glutamate levels in rat hippocampus, striatum and sensorimotor cortex Neurosci Lett 1999 277 119 122 10624824 10.1016/S0304-3940(99)00855-1
Ho YJ Chang YC Liu TM Tai MY Wong CS Tsai YF Striatal glutamate release during novelty exposure-induced hyperactivity in olfactory bulbectomized rats Neurosci Lett 2000 287 117 120 10854726 10.1016/S0304-3940(00)01152-6
Perkinton MS Ip JK Wood GL Crossthwaite AJ Williams RJ Phosphatidylinositol 3-kinase is a central mediator of NMDA receptor signaling to MAP kinase (ERK1/2), AKT/PKB and CREB in striatal neurons J Neurochem 2002 80 239 254 11902114 10.1046/j.0022-3042.2001.00699.x
Perkinton MS Sihra TS Williams RJ Ca2+ permeable AMPA receptors induce phosphorylation of cAMP response element-binding protein through a phosphatidylinositol 3-kinase-dependent stimulation of the mitogen-activated protein kinase signaling cascade in neurons J Neurosci 1999 19 5861 5874 10407026
Choe ES McGinty JF N-Methyl-D-aspartate receptors and p38 mitogen-activated protein kinase are required for cAMP-dependent cyclase response element binding protein and Elk-1 phosphorylation in the striatum Neurosci 2000 101 607 617 10.1016/S0306-4522(00)00379-1
Nadlewska A Car H Wisniewska RJ Wisniewski K Influence of AIDA on certain behaviors in rats subjected to experimental hypoxia Pol J Pharmacol 2003 55 171 180 12926544
Xu Z Herbert J Effects of intracerebroventricular dizocilpine (MK801) on dehydration-induced dipsogenic responses, plasma vasopressin and c-fos expression in the rat forebrain Brain Res 1998 784 91 99 9518565 10.1016/S0006-8993(97)01186-4
Xu Z Johnson AK Non-NMDA receptor antagonist-induced drinking in rat Brain Res 1998 808 124 127 9795183 10.1016/S0006-8993(98)00761-6
Duan W Lee J Guo Z Mattson MP Dietary restriction stimulates BDNF production in the brain and thereby protects neurons against excitotoxic injury J Mol Neurosci 2001 16 1 12 11345515 10.1385/JMN:16:1:1
Ma J Leung LS Metabotropic glutamate receptors in the hippocampus and nucleus accumbens are involved in generating seizure-induced hippocampal gamma waves and behavioral hyperactivity Behav Brain Res 2002 133 45 56 12048173 10.1016/S0166-4328(01)00445-4
Gerfen CR Miyachi S Paletzki R Brown P D1 dopamine receptor supersensitivity in the dopamine-depleted striatum results from a switch in the regulation of ERK1/2/MAP kinase J Neurosci 2002 22 5042 5054 12077200
Paxinos G Watson C The Rat Brain in Stereotaxic Coordinates 1998 Sydney: Academic Press
| 15579204 | PMC539278 | CC BY | 2021-01-04 16:03:46 | no | BMC Neurosci. 2004 Dec 3; 5:50 | utf-8 | BMC Neurosci | 2,004 | 10.1186/1471-2202-5-50 | oa_comm |
==== Front
BMC Med Res MethodolBMC Medical Research Methodology1471-2288BioMed Central London 1471-2288-4-271558832710.1186/1471-2288-4-27Research ArticleProportional odds ratio model for comparison of diagnostic tests in meta-analysis Siadaty Mir Said [email protected] Jianfen [email protected] Division of Biostatistics and Epidemiology, University of Virginia School of Medicine, Box 800717, Charlottesville, Virginia, 22908, USA2004 10 12 2004 4 27 27 30 8 2004 10 12 2004 Copyright © 2004 Siadaty and Shu; licensee BioMed Central Ltd.2004Siadaty and Shu; 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
Consider a meta-analysis where a 'head-to-head' comparison of diagnostic tests for a disease of interest is intended. Assume there are two or more tests available for the disease, where each test has been studied in one or more papers. Some of the papers may have studied more than one test, hence the results are not independent. Also the collection of tests studied may change from one paper to the other, hence incomplete matched groups.
Methods
We propose a model, the proportional odds ratio (POR) model, which makes no assumptions about the shape of ORp, a baseline function capturing the way OR changes across papers. The POR model does not assume homogeneity of ORs, but merely specifies a relationship between the ORs of the two tests.
One may expand the domain of the POR model to cover dependent studies, multiple outcomes, multiple thresholds, multi-category or continuous tests, and individual-level data.
Results
In the paper we demonstrate how to formulate the model for a few real examples, and how to use widely available or popular statistical software (like SAS, R or S-Plus, and Stata) to fit the models, and estimate the discrimination accuracy of tests. Furthermore, we provide code for converting ORs into other measures of test performance like predictive values, post-test probabilities, and likelihood ratios, under mild conditions. Also we provide code to convert numerical results into graphical ones, like forest plots, heterogeneous ROC curves, and post test probability difference graphs.
Conclusions
The flexibility of POR model, coupled with ease with which it can be estimated in familiar software, suits the daily practice of meta-analysis and improves clinical decision-making.
==== Body
Background
A diagnostic test, in its simple form, tries to detect presence of a particular condition (disease) in a sample. Usually there are several studies where performance of the diagnostic test is measured by some statistic. One may want to combine such studies to get a good picture of performance of the test, a meta-analysis. Also, for a particular disease there may be several diagnostic tests invented, where each of the tests is subject of one or more studies. One may also want to combine all such studies to see how the competing tests are performing with respect to each other, and choose the best for clinical practice.
To pool several studies and estimate a summary statistic some assumptions are made. One such assumption is that differences seen between individual study results are due to chance (sampling variation). Equivalently, this means all study results are reflecting the same "true" effect [1]. However, meta-analysis of studies for some diagnostic tests show that this assumption, in some cases, is not empirically supported. In other words, there is more variation between the studies that could be explained by random chance alone, the so-called "conflicting reports". One solution is to relax the assumption that every study is pointing to the same value. In other words, one accepts explicitly that different studies may correctly give "different" values for performance of the same test.
For example, sensitivity and specificity are a pair of statistics that together measure the performance of a diagnostic test. One may want to compute an average sensitivity and an average specificity for the test across the studies, hence pooling the studies together. Instead, one may choose to extract odds ratio (OR) from each paper (as test performance measure), and then estimate the average OR across the studies. The advantage is that widely different sensitivities (and specificities) can point to the same OR. This means one is relaxing the assumption that all the studies are pointing to the same sensitivity and specificity, and accepts that different studies are reporting "truly different" sensitivity and specificity, and that the between-study variation of them is not due to random noise alone, but because of difference in choice of decision threshold (the cutoff value to dichotomize the results). Therefore the major advantage of OR, and its corresponding receiver-operating-characteristic (ROC) curve, is that it provides measures of diagnostic accuracy unconfounded by decision criteria [2]. An additional problem when pooling sensitivities and specificities separately is that it usually underestimates the test performance [[3], p.670].
The above process may be used once more to relax the assumption that every study is pointing to the same OR, thus relaxing the "OR-homogeneity" assumption. In other words, in some cases, the remaining variation between studies, after utilizing OR as the summary performance measure, is still too much to be attributed to random noise. This suggests OR may vary from study to study. Therefore one explicitly assumes different studies are measuring different ORs, and that they are not pointing to the same OR. This difference in test performance across studies may be due to differences in study design, patient population, case difficulty, type of equipment, abilities of raters, and dependence of OR on threshold chosen [4]. Nelson [5] explains generating ROC curves that allow for the possibility of "inconstant discrimination accuracy", a heterogeneous ROC curve (HetROC). This means the ROC curve represents different ORs at different points. This contrasts with the fact that the homogeneous-ROC is completely characterized by one single OR.
There are a few implementations of the heterogeneous ROC. One may classify them into two groups. The first group is exemplified by Tosteson and Begg [6]. They show how to use ordinal regression with two equations that correspond to location and scale. The latent scale binary logistic regression of Rutter and Gatsonis [4] belong to this group. The second group contains implementations of Kardaun and Kardaun [7], and Moses et al [8]. Moses et al explain a method to plot such heterogeneous ROC curve under some parametric assumptions, and they call it summary ROC (SROC).
When comparing two (or more) diagnostic tests, where each study reports results on more than one test, the performance statistics (in the study results) are correlated. Then standard errors computed by SROC are invalid. Toledano and Gatsonis [9] use the ordinal regression model, and account for the dependency of measurements by generalized estimating equations (GEE). However, to fit the model they suggest using a FORTRAN code.
We propose a regression model that accommodates more general heterogeneous ROC curves than SROC. The model accommodates complex missing patterns, and accounts for correlated results [10]. Furthermore, we show how to implement the model using widely available statistical software packages. The model relaxes OR-homogeneity assumption. In the model, when comparing two (or more) tests, each test has its own trend of ORs across studies, while the trends of two tests are (assumed to be) proportional to each other, the "proportional odds ratio" assumption. We alleviate dilemma of choosing weighting schemes such that do not bias the estimates [[11], p.123], by fitting the POR model to 2-by-2 tables. The model assumes a binomial distribution that is more realistic than a Gaussian used by some implementations of HetROC. Also, it is fairly easy to fit the model to (original) patient level data (if available).
Besides accounting better for between-study variation, we show how to use the POR model to "explain why" such variation exists. This potentially gives valuable insights and may have direct clinical applications. It may help define as to when, where, how, and on what patient population to use which test, to optimize performance.
We show how to use "deviation" contrast, in parameterization of categorical variables, to relax the restriction that a summary measure may be reported only if the respective interaction terms in the model are insignificant. This is similar to using grand mean in a "factor effects" ANOVA model (compared to "cell means" ANOVA model).
We show how to use nonparametric smoothers, instead of parametric functions of true positive rate (TPR) and/or false positive rate (FPR), to generate heterogeneous ROC for a single diagnostic test across several studies.
Our proposed POR model assumes the shape of the heterogeneous ROC curve is the same from one test to the other, but they differ in their locations in the ROC space. This assumption facilitates the comparison of the tests. However, one may want to relax the POR assumption, where each test is allowed to have a heterogeneous ROC curve with a different shape. One may implement such generalized comparison of the competing diagnostic tests by a mixed effects model. This may improve generalizability of meta-analysis results to all (unobserved) studies. Also, a mixed effects model may take care of remaining between-study variation better.
Methods
Average difference in performances
To compare two diagnostic tests i and j, we want to estimate the difference in their performance. However, in reality such difference may vary from one paper (study) to the other. Therefore Δi,j,p = PERFi,p - PERFj,p, where the difference Δ depends on paper index p, where PERFi,p is observed performance of test i in paper p. To simplify notation, assume that a single number measures performance of each test in each paper. We relax this assumption later, allowing for the distinction between the two types of mistakes (FNR and FPR, or equivalently TPR and FPR). We decompose the differences
(1) Δi,j,p = PERFi,p - PERFj,p= δi,j + δi,j,p,
where δi,j is the 'average' difference between the two tests, and δi,j,p is deviation of the observed difference within paper p from the average δi,j. The δi,j is an estimator for the difference between performance of the two tests. Note by using deviation parameterization (similar to an ANOVA model) [[12], pp.51 & 45] we explicitly accept and account for the fact that the observed difference varies from one paper to the other, while estimating the 'average' difference. This is similar to a random-effects approach where a random distribution is assumed for the Δi,j,p and then the mean parameter for the distribution is estimated. In other words, one does not need to assume 'homogeneous' difference of the two tests across all the papers, and then estimate the 'common' difference [13].
The observed test performance, PERF, may be measured in several different scales, such as paired measures sensitivity and specificity, positive and negative predictive values, likelihood ratios, post test odds, and post test probabilities for normal and abnormal test results; as well as single measures such as accuracy, risk or rate ratio or difference, Youden's index, area under ROC curve, and odds ratio (OR). When using OR as the performance measure, the marginal logistic regression model
(2) logit(Resultpt) = β0 + β1*Diseasept + β2*PaperIDpt + β3*Diseasept*PaperIDpt + β4*TestIDpt + β7*Diseasept*TestIDpt + β6*TestIDpt*PaperIDpt + β7*Diseasept*TestIDpt*PaperIDpt
implements the decomposition of the performance. Model (2) is fitted to the (repeated measures) grouped binary data, where the 2-by-2 tables of gold-standard versus test results are extracted from each published paper. In the model (2) Result is an integer-valued variable for positive test result (depending on software choice, for grouped binary data, usually Result is replaced by number of positive test results over the total sample size, for each group); Disease is an indicator for actual presence of disease, ascertained by the gold standard; PaperID is a categorical variable for papers included in the meta-analysis; and TestID is a categorical variable for tests included. Regression coefficients β2 to β7 can be vector valued, meaning having several components, so the corresponding categorical variables should be represented by suitable number of indicator variables in the model. Indexes p and t signify paper p and test t. They define the repeated measures structure of the data [10]. Note model (2) fits the general case where there are two or more tests available for the disease, where each test has been studied in one or more papers. Some of the papers may have studied more than one test; hence the results are not independent. Also the collection of tests studied may change from one paper to the other, hence incomplete matched groups.
From model (2) one can show that
LORpt = β1 + β3*PaperIDpt + β5* TestIDpt + β7*TestIDpt*PaperIDpt
and therefore the difference between performance of two tests i and j, measured by LOR, is
LORpi - LORpj = β5* TestIDpi - β5* TestIDpj + β7*TestIDpi*PaperIDpi - β7*TestIDpj*PaperIDpj
where we identify δi,j of the decomposition model (1) with the β5* TestIDpi - β5*TestIDpj, and identify δi,j,p with β7*TestIDpi*PaperIDpi - β7*TestIDpj*PaperIDpj.
If there is an obvious and generally accepted diagnostic test that can serve as a reference category (RefCat) to which other tests can be compared, then a "simple" parameterization for tests is sufficient, However, usually it is not the case. When there is no perceived referent test to which the other tests are to be compared, a "deviation from means" coding is preferred for the tests. Using the deviation parameterization for both TestID and PaperID in the model (2), one can show that β5*TestIDpt is the average deviation of the LOR of test t from the overall LOR (the β1), where the overall LOR is the average over all tests and all papers. Therefore β5*TestIDpt of model (2) will be equivalent to the δi,j of the decomposition model (1), and β7*TestIDpt*PaperIDpt equivalent to δi,j,p.
Proportional odds ratio model
Model (2) expands each study to its original sample size, and uses patients as primary analysis units. Compared to a random-effects model where papers are the primary analysis units, it has more degrees of freedom. However, in a real case, not every test is studied in every paper. Rather majority of tests are not studied in each paper. Therefore the data structure of tests-by-papers is incomplete with many unmeasured cells. The three-way interaction model (2) may become over-parameterized. One may want to drop the term β6*Diseasept*TestIDpt*PaperIDpt. Then for the reduced model
(3) logit(Resultpt) = β0 + β1*Diseasept + β2*PaperIDpt + β3*Diseasept*PaperIDpt + β4*TestIDpt + β5*Diseasept*TestIDpt
we have LORpt = β1 + β3*PaperIDpt + β5* TestIDpt, where the paper and test effects are completely separate. We call this reduced model the Proportional Odds Ratio (POR) model, where the ratio of odds ratios of two tests is assumed to be constant across papers, while odds ratio of each test is allowed to vary across the papers. Note the difference with the proportional odds model where ratio of odds is assumed to be constant [14]. In the POR model
(4) ORpt = ORp * , t = 1, 2, ..., k, p = 1, 2, ..., m
where t is an index for the k diagnostic tests, and p is an index representing the m papers included in the analysis. ORp is a function capturing the way OR changes across papers. Then to compare two diagnostic tests i and j
ORpi / ORpj =
where the ratio of the two ORs depends only on the difference between the effect estimates of the two tests, and is independent of the underlying ORp across the papers. Thus the model makes no assumptions about the shape of ORp (and in particular homogeneity of ORs) but merely specifies a relationship between the ORs of the two tests.
One may want to replace the PaperID variable with a smooth function of FPR or TPR, such as natural restricted cubic splines. There are two potential advantages. This may preserve some degrees of freedom, where one can spend by adding covariates to the model to measure their potential effects on the performance of the diagnostic tests. Thus one would be able to explain why performance of the same test varies across papers. Also, this allows plotting a ROC curve where the OR is not constant across the curve, a flexible ROC (HetROC) curve.
(5) logit(Resultpt) = β0 + β1*Diseasept + β2*S(FPRpt) + β3*Diseasept*S(FPRpt) + β4*TestIDpt + β5*Diseasept*TestIDpt + β6*Xpt + β5*Diseasept*Xpt
To test the POR assumption one may use model (2) where the three-way interaction of Disease and TestID with PaperID is included. However, in majority of real datasets this would mean an over-parameterized model. Graphics can be used for a qualitative checking of the POR assumption. For instance, the y-axis can be LOR, while the x-axis is paper number. To produce such plot, it may be better to have the papers ordered in some sense. One choice is to compute an unweighted average of (observed) ORs of all the tests the paper studied, and use it as the OR of that paper. Then sort the papers based on such ORs. The OR of a test may vary from one paper to the other (with no restriction), but the POR assumption is that the ratio of ORs of two tests remains the same from one paper to another. If one shows ORs of a test across papers by a smooth curve, then one expects that the two curves of the two tests are proportional to each other. In the log-OR scale, this means the vertical distance of the two curves remains the same across the x-axis. To compute the observed LOR for a test in a paper one may need to add some value (like 1/2) to the cell counts, if some cell counts are zero. However, this could introduce some bias to the estimates.
Among the approaches for modeling repeated-measures data, we use generalized estimating equations to estimate the marginal logistic regression [15]. Software is widely available for estimation of parameters of a marginal POR model. These include SAS (genmod procedure), R (function geese), and STATA (command xtgee), with R being freely available open source software [16].
One may use a non-linear mixed effects modeling approach on the cell-count data for estimation of parameters of the POR model. The Paper effect is declared as random, and interaction of the random effect with Disease is included in the model, as indicated in model (2). However, such mixed effects non-linear models are hard to converge, especially for datasets where there are many papers studying only one or a small number of the included tests (such as the dataset presented as example in this paper). If the convergence is good, it may be possible to fit a mixed model with the interaction of Disease, Test, and the Paper random effect. Such model relaxes the POR assumption, besides relaxing the assumption of OR-homogeneity. In other words, one can use the model to quantitatively test the POR assumption. One should understand that the interpretation of LOR estimate from a marginal model is of a population-average, while that of a mixed model is a conditional-average. Therefore there is a slight difference in their meaning.
Expanding the proportional odds ratio model
One may use the frameworks of the generalized linear models (GLM) and the generalized estimating equations (GEE) to extend the POR model and apply it to different scenarios. By using suitable GLM link function and random component [[17], p.72], one may fit the POR model to multi-category diagnostic tests, like baseline-category logits, cumulative logits, adjacent-categories and continuation-ratio logits [[17], chapter 8]. A loglinear 'Proportional Performance' (PP) regression may be fitted to the cell counts, treating them as Poisson. Also, one may fit the PP model to the LORs directly, assuming a Gaussian random component with an identity link function. Comparing GEE estimates by fitting the model to 2-by-2 tables versus GEE estimates of the model fitted directly on LOR versus a Mixed model fitted on LOR, usually statistical power decreases across the three. Also, there is issue of incorporation of sample sizes that differ across studies. Note some nuisance parameters, like coefficients of all main effects and the intercept, won't need to be estimated as they are no longer present in the model fitted directly on LORs.
One may avoid dichotomizing results of the diagnostic test by using the 'likelihood ratio' as the performance measure, and fitting a PP model to such continuous outcome. For a scenario where performance of a single test has been measured multiple times within the same study, for example with different diagnostic calibrations (multiple thresholds), the POR estimated by the GEE incorporates data dependencies. When there is a multi-layer and/or nested clustering of repeated measures, software to fit a mixed-effects POR model may be more available than an equivalent GEE POR.
When POR is implemented by a logistic regression on 2-by-2 tables, it uses a grouped binary data structure. It takes a minimal effort to fit the same logistic model to the "ungrouped" binary data, the so-called "individual level" data.
Methods of meta-analysis that allow for different outcomes (and different numbers of outcomes) to be measured per study, such as that of Gleser and Olkin [18], or DuMouchel [19], may be used to implement the POR model. This would prevent conducting parallel meta-analyses that is usually less efficient.
Results
Deep vein thrombosis
To demonstrate how to fit the POR model, we use a recent meta-analysis of diagnostic tests for deep vein thrombosis (DVT) by Heim et al. [20]. In this meta-analysis there are 23 papers and 21 tests, comprising 483 potential performance measurements, while only 66 are actually observed, thus 86% of cells are not measured. We fitted the reduced marginal logistic regression model (3). Table 1 shows the parameter estimates for Test effects. SAS code to estimate the parameters is provided [see additional file 1].Data files are provided in Additional file 2.
Table 1 Parameter estimates for test effects
Coefficient Test Deviation* 95% Confidence Limits p value**
β5† 1 Asserachrom 0.524 0.2293, 0.8186 0.0005
2 Auto Dimertest 0.222 -0.1466, 0.5912 0.2376
3 BC D-Dimer -0.993 -2.4195, 0.4333 0.1724
4 D-Dimer test 0.225 0.1, 0.3494 0.0004
5 Dimertest -2.092 -2.3392, -1.8439 <.0001
6 Dimertest EIA -0.929 -1.1756, -0.6825 <.0001
7 Dimertest GOLD EIA -0.193 -0.4784, 0.0935 0.1871
8 Dimertest II -0.731 -0.9774, -0.4843 <.0001
9 Enzygnost 0.399 0.1209, 0.6766 0.0049
10 Fibrinostika 0.857 0.6865, 1.0266 <.0001
11 IL Test 0.809 0.0914, 1.5256 0.0271
12 Instant I.A. 0.558 0.216, 0.9006 0.0014
13 Liatest -0.143 -0.3375, 0.0511 0.1486
14 LPIA 0.182 -0.0354, 0.3997 0.1007
15 Minutex -0.323 -0.8394, 0.193 0.2197
16 Nephelotex 0.654 0.4325, 0.8745 <.0001
17 NycoCard -0.797 -1.0434, -0.5506 <.0001
18 SimpliRED 0.393 0.1467, 0.6398 0.0018
19 Tinaquant 0.703 0.0113, 1.3948 0.0464
20 Turbiquant -0.328 -1.6596, 1.0032 0.629
21 VIDAS 1.004 0.365, 1.6424 0.0021
β1 Overall LOR 2.489 2.4175, 2.5606 < .0001***
* estimate of deviation from overall LOR
** p-value for null hypothesis of Deviation = 0
***p-value for null hypothesis of LOR = 0
† LOR(Resultpt) = β1 + β3*PaperIDpt + β5*TestIDpt
Since we have used deviation contrast for the variables, estimate of β1 is the "overall mean" for the log-OR. This is similar to an ANOVA analysis where the overall mean is estimated by the model. Therefore the average OR is equal to exp(2.489) = 12.049. Components of β5 estimate deviation of LOR of each test from the overall LOR. Software gives estimates of SEs, plus confidence intervals and p-values, so inference is straightforward.
A forest plot may be used to present the results of the modeling in a graphical way. This may connect better with clinically oriented audience. In Figure 1 we have sorted the 21 tests based on their LOR estimate.
Figure 1 Comparing performance of each diagnostic test to the overall LOR
The horizontal axis is log-OR, representing test performance. The dashed vertical line shows overall mean LOR. For each diagnostic test the solid square shows the LOR, while the horizontal line shows the corresponding 95% CI. If the horizontal line does not intersect the vertical line, the test is significantly different from the overall mean LOR.
Note that the CIs in the plot are computed by adding the overall LOR to the CI for the deviation effect of each particular test. This ignores the variability of the overall LOR estimate. One can estimate the LOR of a test and its CI more accurately by some extra computations, or by fitting a slightly modified model. A method is illustrated and implemented [see additional file 1]. However, the gain in accuracy was small in this particular example. The model also estimates paper effects. However, one may not be interested in those primarily.
One can translate LOR to other measures of test performance. There are numerous types of these measures. We provide code to convert the LOR estimated by the POR model to such measures. Note that majority of them, unlike LOR, are in pairs. This means in order to compare two tests, one needs to use two numbers to represent each single test. For example, sensitivity-specificity is a pair. If a test has a higher sensitivity than the other test, while having a lower specificity, it is not immediately clear which test is better. Also, note that some performance measures are independent of disease prevalence, while others depend on prevalence. This means the same test would perform differently for populations with different disease prevalence.
Note in order to compute some of the performance measures, one needs to assume a prevalence and sensitivity or specificity. We assumed a disease prevalence of 40%, and a specificity of 90%, for Table 2, as the tests are mainly used for ruling out the DVT.
Table 2 Other performance measures for the 21 diagnostic tests of DVT
Diagnostic Test DOR Prev. Spec. Sens. AUC PPV NPV LRAT LRNT PTO PTOAT PTONT PTPAT PTPNT
1 Asserachrom 20.3 0.4 0.9 0.693 0.888 0.822 0.815 6.933 0.341 0.667 4.622 0.227 0.822 0.185
2 Auto Dimertest 15.0 0.4 0.9 0.626 0.864 0.807 0.783 6.258 0.416 0.667 4.172 0.277 0.807 0.217
3 BC D-Dimer 4.5 0.4 0.9 0.332 0.732 0.688 0.669 3.315 0.743 0.667 2.210 0.495 0.688 0.331
4 D-Dimer test 15.1 0.4 0.9 0.626 0.865 0.807 0.783 6.263 0.415 0.667 4.175 0.277 0.807 0.217
5 Dimertest 1.5 0.4 0.9 0.142 0.566 0.486 0.611 1.419 0.953 0.667 0.946 0.636 0.486 0.389
6 Dimertest EIA 4.8 0.4 0.9 0.346 0.741 0.697 0.674 3.459 0.727 0.667 2.306 0.485 0.697 0.326
7 Dimertest GOLD EIA 9.9 0.4 0.9 0.525 0.826 0.778 0.740 5.248 0.528 0.667 3.499 0.352 0.778 0.260
8 Dimertest II 5.8 0.4 0.9 0.392 0.766 0.723 0.689 3.920 0.676 0.667 2.613 0.450 0.723 0.311
9 Enzygnost 18.0 0.4 0.9 0.666 0.879 0.816 0.802 6.661 0.371 0.667 4.440 0.247 0.816 0.198
10 Fibrinostika 28.4 0.4 0.9 0.759 0.910 0.835 0.849 7.592 0.268 0.667 5.061 0.178 0.835 0.151
11 IL Test 27.0 0.4 0.9 0.750 0.907 0.833 0.844 7.503 0.277 0.667 5.002 0.185 0.833 0.156
12 Instant I.A. 21.1 0.4 0.9 0.701 0.890 0.824 0.818 7.006 0.333 0.667 4.671 0.222 0.824 0.182
13 Liatest 10.4 0.4 0.9 0.537 0.831 0.782 0.745 5.371 0.514 0.667 3.581 0.343 0.782 0.255
14 LPIA 14.5 0.4 0.9 0.616 0.861 0.804 0.779 6.163 0.426 0.667 4.109 0.284 0.804 0.221
15 Minutex 8.7 0.4 0.9 0.492 0.813 0.766 0.727 4.921 0.564 0.667 3.281 0.376 0.766 0.273
16 Nephelotex 23.2 0.4 0.9 0.720 0.897 0.828 0.828 7.202 0.311 0.667 4.801 0.207 0.828 0.172
17 NycoCard 5.4 0.4 0.9 0.376 0.758 0.715 0.684 3.763 0.693 0.667 2.509 0.462 0.715 0.316
18 SimpliRED 17.9 0.4 0.9 0.665 0.878 0.816 0.801 6.648 0.372 0.667 4.432 0.248 0.816 0.199
19 Tinaquant 24.3 0.4 0.9 0.730 0.900 0.830 0.833 7.300 0.300 0.667 4.867 0.200 0.830 0.167
20 Turbiquant 8.7 0.4 0.9 0.491 0.812 0.766 0.726 4.909 0.566 0.667 3.273 0.377 0.766 0.274
21 VIDAS 32.9 0.4 0.9 0.785 0.918 0.840 0.863 7.851 0.239 0.667 5.234 0.159 0.840 0.137
DOR = Diagnostic Odds Ratio
Prev. = Prevalence
Spec. = Specificity
Sens. = Sensitivity
AUC = Area Under Curve (assuming homogeneous OR)
PPV = Positive Predictive Value
NPV = Negative Predictive Value
LRAT = Likelihood Ratio For Abnormal Test
LRNT = Likelihood Ratio For Normal Test
PTO = Pre Test Odds
PTOAT = Post Test Odds Of Abnormal Test
PTONT = Post Test Odds Of Normal Test
PTPAT = Post Test Probability Of Abnormal Test
PTPNT = Post Test Probability Of Normal Test
We suggest graphs to compare tests when using such "prevalence-dependent paired performance measures" [21]. In Figure 2 we have used a pair of measures, 'probability of disease given a normal test result' and 'probability of disease given an abnormal test result', the dashed red curve and the dot-and-dash blue curve respectively.
Figure 2 Post-test probability difference for diagnostic test VIDAS
The way one may read the graph is that, given a particular population with a known prevalence of disease like 40%, we perform the diagnostic test on a person picked randomly from the population. If the test turns normal, the probability the person has disease decreases from the average 40% to about 4% (draw a vertical line from point 0.4 on x-axis to the dashed red curve, then draw a horizontal line from the curve to the y-axis). If the test turns abnormal, the probability the person is diseased increases from 40% to about 57%. The dotted green diagonal line represents a test no better than flipping a coin, an uninformative test. The farther the two curves from the diagonal line, the more informative the test is. In other words, the test performs better.
One can summarize the two curves of a test in a single curve, by computing the vertical distance between the two. The solid black curve in the figure is such "difference" curve. It seems this particular test is performing the best in populations with disease prevalence of around 75%.
One can use the difference curve to compare several tests, and study effect of prevalence on the way the tests compare to each other. In Figure 3 two tests VIDAS and D-Dimer from the DVT example are compared. From the model estimates we know that both tests perform better than average. And that VIDAS performs better than D-Dimer.
Figure 3 Comparing post-test probability difference for VIDAS – D-Dimer
The black solid curve is comparing the two tests. For populations with low disease prevalence (around 17%), the D-Dimer is performing better than VIDAS. However, when the prevalence is higher (around 90%), VIDAS is preferred. Simultaneous confidence bands around the comparison curve would make formal inference possible.
Random effects
A nonlinear mixed effects POR model fitted to cell counts of the DVT dataset does not converge satisfactorily. We fitted the mixed model to a subset of the data where only two tests and seven papers are included, Table 3. For codes refer to the additional file 1.
Table 3 Data structure for two diagnostic tests
Test
Paper Instant I.A. NycoCard
3 Elias, A. 1996 (171) X X
8 Legnani, C. 1997 (81) X X
11 Leroyer, C. 1997 (448) X
12 Scarano, L. 1997 (126) X X
13 van der Graaf, F. 2000 (99) X X
21 Wijns, W. 1998 (74) X
22 Kharia, HS. 1998 (79) X
TOTAL 6 5
Five of the seven papers have studied both the tests. Result of SAS Proc NLMixed still is sensitive to initial values of parameters. The three-way interaction term of disease, test, and paper in the mixed model (where POR is not assumed) is insignificant, Table 4. A POR assumption for the two tests may be acceptable.
Table 4 Comparing parameter estimates from three models
POR-relaxed Mixed * POR Mixed POR Marginal
overall LOR 1.389 (0.993, 1.786) 0.868 (0.568, 1.169) 2.593 (2.522, 2.664)
Test (NycoCard) -0.903 (-1.811, 0.006) -0.93 (-1.104, -0.755) -0.561 (-0.829, -0.293)
Test*Paper 0.016 (-1.556, 1.588) --- ---
* logit(Result) = β0 + β1*Disease + β2*PaperID + β3*Disease*PaperID + β4*TestID + β5*Disease*TestID + β6*Disease*TestID*PaperID
The estimate of overall LOR from both the POR-mixed model and POR-marginal model are significantly different from zero. However, the mixed model estimate of LOR is much smaller than the marginal one. For non-linear models, the marginal model describes the population parameter, while the mixed model describes an individual's [[15], p.135]. The estimate of deviation of test (NycoCard) from the overall LOR is closer in the two models. Plus the marginal estimate is closer to 0 than the mixed estimate. One expects coefficient estimates of mixed model being closer to zero, compared to the fixed model, while the mixed model CI's being wider.
Meta-analysis of a single test: the baseline ORp function
Sometimes one may be interested in constructing the ROC curve for the diagnostic test. A homogeneous ROC curve assumes the performance of the test (as measured by LOR) is the same across the whole range of specificity. However, this assumption may be relaxed in a HetROC. We fitted a simplified version of model (5) for test SimpliRED,
logit(Resultpt) = β0 + β1*Diseasept + β2*S(FPRpt) + β3*Diseasept*S(FPRpt)
where index t is fixed, and then used estimates of the coefficients to plot the corresponding HetROC, Figure 4.
Figure 4 Heterogeneous ROC curve for diagnostic test SimpliRED
The eleven papers that studied test SimpliRED are shown by circles where the area is proportional to the sample size of the study. The black dashed curve is ROC curve assuming homogeneous-OR. The red solid curve relaxes the assumption, hence a heterogeneous ROC curve. The amount of smoothing of the curve can be controlled by the "degree-of-freedom" DF parameter. Here we have used a DF of 2. Codes to make such plots are presented in the additional file 1.
Model checking
Checking the POR assumption, model (2) may be used to reject significance of the three-way interaction term. However, the dataset gathered for the DVT meta-analysis is such that no single paper covers all the tests. Moreover, out of 21, there are 7 tests that have been studied in only one paper. For Figure 5 we chose tests that have been studied in at least 5 of the 23 papers. There are 5 such tests. Note that even for such "popular" tests, out of 10 pairwise comparisons, 3 are based on only one paper (so no way to test POR). Four comparisons are based on 4 papers, one based on 3 papers, and the remaining two comparisons are based on 2 papers.
Figure 5 Observed log-odds-ratios of each diagnostic test
We sorted the papers, the x-axis, based on average LOR within that paper. We fitted Lowess smooth lines to the observed LORs of each test separately. Figure 5 shows the smooth curves are relatively parallel. Note the range of LORs of a single test. The LORs vary considerably from one paper to the other. Indeed the homogeneity-of-ORs assumption is violated in four of the five tests.
Also, to verify how good the model fits the data, one may use an observed-versus-fitted plot. Plots or lists of standardized residuals may be helpful finding papers or tests that are not fitted well. This may provide a starting point for further investigation.
Discussion
A comparison of the relative accuracy of several diagnostic tests should ideally be based on applying all the tests to each of the patients or randomly assigning tests to patients in each primary study. Obtaining diagnostic accuracy information for different tests from different primary studies is a weak design [3]. Comparison of the accuracy of two or more tests within each primary study is more valid than comparison of the accuracy of two or more tests between primary studies [22]. Although a head-to-head comparison of diagnostic tests provides more valid results, there are real-world practical questions that meta-analysis provides an answer that is more timely and efficient than a single big study [23]. Meta-analysis can potentially provide better understanding by examining the variability in estimates, hence the validity versus generalizability (applicability). Also, there may be tests that have never been studied simultaneously in a single study, hence meta-analysis can "reconstruct" such a study of diagnostic tests.
Relaxing the assumption of OR homogeneity
In meta-analysis of two (or more) diagnostic tests, where attention is mainly on the difference between performances of two tests, having a homogeneous estimate of performance of each single test is of secondary importance, and it may be treated as nuisance. The POR model assumes differences between LORs of two tests are the same across all papers, but does not assume the OR of a test is the same in every paper. Hence no need for homogeneity of OR of a test across papers that reported it, but shifting the assumption one level higher to POR.
Common versus average effect size
The POR model uses "deviation from means" parameterization. Then one does not need to drop the interactions coefficient β3 in the model logit(Result) = β0 + β1*Disease + β2*PaperID + β3*Disease*PaperID, to interpret β1, the overall LOR. This means the POR model explicitly accepts that performance of the diagnostic test varies across the papers, but at the same time estimates its mean value. McClish explains if a test for OR homogeneity shows heterogeneity, there may be no 'common' measure to report, but still there is an 'average' measure one can report. [13]
Advantages of using 2-by-2 tables
We demonstrated how to fit the POR model to the cell counts, rather than to the OR values. This, we believe, has several advantages. 1. One does not need assuming normality of some summary measure. This results in binomial distributional assumption that is more realistic. 2. Also, different study sample sizes are incorporated into the POR model without faulty bias-introducing weighting schemes, as shown by Mosteller & Chalmers [25]. And extension of the POR model to individual level patient data is much easier. 3. The effective sample size for a meta-analysis by a random model is the number of papers included, which is usually quite small. There is a great danger for overfitting. And the number of explanatory variables one could include in the model is very restricted. Since we use the grouped binary data structure, the patients are the effective sample size, hence much bigger degrees of freedom.
The way the random-effects model is usually implemented is by extracting OR from each paper, and assuming LOR being normally distributed. Then the distinction between the two types of mistakes (FNR and FPR, or equivalently TPR and FPR) is lost, since one enters the LOR as datapoints into the model. The bivariate model by Houwelingen et al [26] tries to fix this, by entering two datapoints into the model for each test from each paper. A fourth advantage of fitting the POR model to the cell counts is that the two types of mistakes are included in the model. Consider the logistic regression logit(Result) = β0 + β1*Disease + β2*PaperID . Then we have log(true positive/false negative) = β0 + β1 + β2*PaperID. Substituting a value for the covariate (here PaperID) such as a modal or average value, and using the model estimates for the betas, one gets the log-odds. Then one exponentiates it to get the TP/FN, call it Q. Now it is easy to verify that sensitivity = Q/(1+Q). Likewise we have log(false positive/true negative) = β0 + β2*PaperID, that we call = log(W). Then specificity = 1/(1+W). Also, one can apply separate weights to the log(true positive/false negative) and log(false positive/true negative), to balance the true positive and false positive rates for decision making in a particular clinical practice.
When collecting papers from biomedical literature for meta-analysis of a few diagnostic tests, it is hard to come up with a complete square dataset, where every paper has included all the tests of interest. Usually the dataset contains missing values, and a case-wise deletion of papers with missing tests means a lot of data is thrown away. A method of analysis that can utilize incomplete matched groups may be helpful. The POR model allows complex missing patterns in data structure. Convergence of marginal POR model seems much better than non-linear mixed model, when fitted to cell counts of incomplete matched groups. This is an advantage for using GEE to estimate POR.
The fact that one can use popular free or commercial software to fit the proposed models, facilitates incorporation of the POR modeling in the practice of meta-analysis.
Unwanted heterogeneity versus valuable variability
The POR model utilizes the variation in the observed performance of a test across papers. Explaining when and how the performance of the test changes, and finding the influential factors, is an important step in advancing science. In other words, rather than calling it 'heterogeneity', treated as 'unwanted' and unfortunate, one calls it 'variability' and utilizes the observed variability to estimate and explain when and how to use the agent or the test in order to optimize their effects.
Victor [32] emphasizes that results of a meta-analysis can only be interpreted if existing heterogeneities can be adequately explained by methodological heterogeneities. The POR model estimates effect of potential predictors on between-study variation, hence trying to 'explain' why such variation exists.
The POR model incorporates risk of events in the control group via a predictor, such as observed prevalence, hence a 'control rate regression'. [26]
ROC curve
Although implementing the HetROC means that one accepts the diagnostic test performs differently in different FPRs along the ROC curve, in some implementations of HetROC, such as method of summary ROC, one compares tests by a single point of their respective ROCs. This is not optimal. (The Q test of the SROC method is a single point test, where that point on the ROC may not be the point for a specific cost-benefit case.) In such method although one produces a complete SROC, but one does not use it in comparing the diagnostic tests. In the POR model, one uses LOR as the measure for diagnostic discrimination accuracy, and builds statistical test based on the LOR-ratio, hence the test corresponds to whole ROCs (of general form).
The ROC graph was designed in the context of the theory of signal detectability [27,28]. ROC can be generated in two ways, by assuming probability distribution functions (PDFs) for the two populations of 'diseased' and 'healthy', or by algebraic formulas [29]. Nelson claims the (algebraic) ROC framework is more general than the signal detection theory (and its PDF-based ROC) [5]. The location-scale regression models implement ROC via PDFs, while the method of summary-ROC uses algebraic approach. The POR model uses a hybrid approach. While POR may be implemented by logistic regression, the smoothing covariate resembles the algebraic method. Unlike location-scale regression models that use two equations, POR uses one equation, hence it is easier to fit by usual statistical packages. One may use a five-parameter logistic to implement the HetROC. However, the model cannot be linearized, then according to McCullagh [14] it won't have good statistical properties. The POR model not only relaxes assumption of Var1/Var2 = 1, where Var1 and Var2 are variances of the two underlying distributions for the two populations, but even monotonicity of ROC. Hence the model can be used to represent both asymmetric ROCs and non-regular ROCs (singular detection).
In building HetROC curve, the POR model accommodates more general heterogeneous ROCs than SROC, because it uses nonparametric smoother instead of arbitrary parametric functions used in SROC method. When in the POR model the smoother covariate is replaced by log{TPR*FPR/ [(1-TPR)*(1-FPR)]}, a HetROC similar to SROC of Moses et al is produced.
When one uses a smooth function of FPR in the POR model, it is equivalent to using a function of outcome as predictor. This resembles a 'transition model'. Ogilvie and Creelman [30] claim that for estimating parameters of a best fitting curve going through observed points in the ROC space, least squares is not good since both axes are dependent variables and subject to error. They claim maximum likelihood is a preferred method of estimation. Crouchley and Davies [31] warn that, although GEE is fairly robust, it becomes inconsistent if any of the covariates are endogenous, like a previous or related outcome or baseline outcome. They claim a mixed model is better for studying microlevel dynamics. We have observed that the smooth HetROC curve may become decreasing at right end, due to some outlier points. Using less smoothing in the splines may be a solution.
When there is only one diagnostic test, and one is mainly interested in pooling several studies of the same test, the POR model estimates effect sizes that are more generalizable. By using the smoother (instead of PaperID), one fits a sub-saturated model that allows inclusion of other covariates, hence it is possible to estimate effect of study level factors on performance and explain the heterogeneity. Also it does not assume any a priori shape of the ROC, including monotonicity. Plus, it enables graphing of the HetROC. It does not need omission of interaction terms to estimate the overall performance, and it does not need assumption of OR homogeneity. If several performance measurements of the same test is done in a single study, like evaluating the same test with different diagnostic calibrations, the POR model provides more accurate estimates, by incorporating the dependence structure of the data.
Random effects
When there is heterogeneity between a few studies for the same diagnostic test, one solution to absorb the extra between-study variation is to use a random/mixed effects model. However, Greenland [33] cautions when working with random effect models: 1. if adding random effect changes the inference substantially, it may indicate large heterogeneity, needing to be explained; 2. specific distributional forms for random effects have no empiric, epidemiologic, or biologic justification. So check its assumptions; 3. the summary statistic from random-effect model has no population-specific interpretation. It represents the mean of a distribution that generates effects. Random models estimate unit specific coefficients while marginal models estimate population averages. The choice between unit-specific versus population-average estimates will depend on the specific research questions that are of interest. If one were primarily interested in how a change in a covariate affect a particular individual cluster's mean, one would use the unit-specific model. If one were interested in how change in covariate can be expected to affect the overall population mean, one would use the population-average model. The difference between "unit-specific" models and "population-average" models arises only in the case of a nonlinear link function. In essence random-effect model exchanges questionable homogeneity assumption for a fictitious random distribution of effects. Advantage of a random model is that SE and CI reflect unaccounted-for sources of variation, and its drawback is that simplicity of interpretation is lost. When residual heterogeneity is small, fixed and random should give same conclusions. Inference about the fixed effects (in a mixed model) would apply to an entire population of cases defined by random effect, while the same coefficient from a fixed model apply only to particular units in the data set. Crouchley and Davies [31] explain one of the drawbacks of their random model is that it rapidly becomes over-parameterized, and also may encounter multiple optima.
Follow-ups
We suggest these follow-ups: 1. the POR model has been implemented both by marginal and mixed models. It would be useful to implement a marginalized mixed POR model; 2. in clinical practice, usually a group of diagnostic tests is performed on an individual, for a particular disease. Some of these tests are requested simultaneously and some in sequence. It would be useful, and practically important, to extend the POR model such that it incorporates such sequence of testing and a priori results; 3. the utility of POR model may be extended to meta-analysis of therapeutics.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MSS conceived of the model, and participated in its design and implementation. JS participated in implementation of the model and performing of the example analysis. Both authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
In this file we present sample codes for a few of the models presented in the paper. The estimation mostly has been done in SAS, while the graphing (and some model-fitting) has been done in R.
Click here for file
Additional File 2
This zipped file contains 8 data files, in the .csv (comma separated value) and .xls (MS Excel) formats. They are to be used with the SAS and R codes we presented in the Appendix [additional file 1]. Five files are for the SAS codes presented in the Appendix. The file names are "data5.xls", "data5_t12&17.xls", "u125.xls", "data5_t18.xls", "data6.xls". Three files are for the R codes presented in the Appendix. The file names are "obsVSfit.csv", "dataNewExcerpt2.csv", and "data6_lor2.csv".
Click here for file
==== Refs
L'Abbe KA Detsky AS O'Rourke K Meta-analysis in clinical research Ann Intern Med 1987 107 224 33 3300460
Dorfman DD Berbaum KS Metz CE Receiver operating characteristic rating analysis Invest Radiol 1992 27 723 731 1399456
Irwig L Tosteson ANA Gatsonis C Lau J Colditz G Chalmers TC Mosteller F Guidelines for meta-analyses evaluating diagnostic tests Ann Intern Med 1994 120 667 676 8135452
Rutter CM Gatsonis CA Regression methods for meta-analysis of diagnostic test data Acad Radiol 1995 2 S48 S56 9419705
Nelson TO ROC curves and measures of discrimination accuracy: A reply to Swets Psychol Bull 1986 99 128 132 3737791
Tosteson AN Begg CB A general regression methodology for ROC curve estimation Med Decis Making 1988 8 204 215 3294553
Kardaun JW Kardaun OJWF Comparative diagnostic performance of three radiological procedures for the detection of lumbar disk herniation Meth Inform Med 1990 29 12 22 2308524
Moses LE Shapiro D Littenberg B Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations Stat Med 1993 12 1293 316 8210827
Toledano A Gatsonis CA Regression analysis of correlated receiver operating characteristic data Acad Radiol 1995 2 S30 S36 9419703
Siadaty MS Philbrick JT Heim SW Schectman JM Repeated-measures modeling improved comparison of diagnostic tests in meta-analysis of dependent studies Journal of Clinical Epidemiology 2004 57 698 710 15358397
Irwig L Macaskill P Glasziou P Fahey M Meta-analytic methods for diagnostic test accuracy J Clin Epidemiol 1995 48 119 130 7853038 10.1016/0895-4356(94)00099-C
Hosmer DW Lemeshow S Applied Logistic Regression 1989 New York: Wiley-Interscience
McClish DK Combining and comparing area estimates across studies or strata Med Decis Making 1992 12 274 279 1484476
McCullagh P Regression models for ordinal data J R Statist Soc B 1980 42 109 142
Diggle P Heagerty P Liang KY Zeger S Analysis of Longitudinal Data 2002 New York: Oxford University Press
Ihaka R Gentleman RR A language for data analysis and graphics Journal of Computational and Graphical Statistics 1996 5 299 314
Agresti A An Introduction to Categorical Data Analysis 1996 New York: Wiley-Interscience
Gleser LJ Olkin I Cooper H, Hedges LV Stochastically dependent effect sizes In The Handbook of Research Synthesis 1994 New York: Russell Sage Foundation 339 56
DuMouchel W Repeated measures meta-analyses Bulletin of the International Statistical Institute, Session 51, Tome LVII, Book 1 1998 285 288
Heim SW Schectman JM Siadaty MS Philbrick JT D-dimer testing for deep venous thrombosis: a metaanalysis Clin Chem 2004 50 1136 47 15142977 10.1373/clinchem.2004.031765
Hamilton GW Trobaugh GB Ritchie JL Gould KL DeRouen TA Williams DL Myocardial imaging with Thallium 201: an analysis of clinical usefulness based on Bayes' theorem Semin Nucl Med 1978 8 358 364 754289
Cochrane methods group on systematic review of screening and diagnostic tests: recommended methods
Spitzer WO The challenge of meta-analysis J Clin Epidemiol 1995 48 1 4 7853036 10.1016/0895-4356(94)00109-4
Neter J Kutner MH Wasserman W Nachtsheim CJ Applied Linear Statistical Models 1996 Boston: McGraw-Hill/Irwin
Mosteller F Chalmers T Some progress and problems in meta-analysis of. clinical trials Statistical Science 1992 7 227 236
van Houwelingen HC Arends LR Stijnen T Advanced methods in meta-analysis: multivariate approach and meta-regression Stat Med 2002 21 589 624 11836738 10.1002/sim.1040
Peterson WW Birdsall TG Fox WC The theory of signal detectability Transactions of the IRE professional group on information theory 1954 4 171 212
Tanner WP Swets JA A decision-making theory of visual detection Psychol Rev 1954 61 401 409 13215690
Swets JA Indices of discrimination or diagnostic accuracy: Their ROCs and implied models Psychol Bull 1986 99 100 117 3704032 10.1037//0033-2909.99.1.100
Ogilvie JC Creelman CD Maximum likelihood estimation of receiver operating characteristic curve parameters J Math Psychol 1968 5 377 391
Crouchley R Davies RB A comparison of population average and random-effect models for the analysis of longitudinal count data with base-line information J R Statist Soc A 1999 162 331 347 10.1111/1467-985X.00139
Victor N "The challenge of meta-analysis": Discussion. Indications and contra-indications for meta-analysis J Clin Epidemiol 1995 48 5 8 7853048 10.1016/0895-4356(94)00107-2
Greenland S Quantitative methods in the review of epidemiologic literature Epidemiol Rev 1987 9 1 30 3678409
| 15588327 | PMC539279 | CC BY | 2021-01-04 16:32:50 | no | BMC Med Res Methodol. 2004 Dec 10; 4:27 | utf-8 | BMC Med Res Methodol | 2,004 | 10.1186/1471-2288-4-27 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-4-561557596510.1186/1471-2334-4-56Research ArticleAge-specific prevalence, transmission and phylogeny of TT virus in the Czech Republic Saláková Martina [email protected]ěmeček Vratislav [email protected]önig Jaroslav [email protected] Ruth [email protected] Department of Experimental Virology, Institute of Hematology and Blood Transfusion, U Nemocnice 1, 128 20 Prague 2, Czech Republic2 National Reference Laboratory for Hepatitis, National Institute of Health, Prague, Czech Republic2004 3 12 2004 4 56 56 6 8 2004 3 12 2004 Copyright © 2004 Saláková 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
TT virus is prevalent worldwide, but its prevalence and genotype distribution in Central and East-Europe has not been determined. The high prevalence of TTV in multiply-transfused patients points to the importance of a parenteral mode of transmission, but since more than half of the general population is infected other possible routes of transmission must be considered.
Methods
In our study, we investigated the epidemiology, transmission and phylogeny of TTV in the Czech Republic. The following groups were selected: a control group of 196 blood donors, 20 patients with hemophilia, 49 intravenous drug users, 100 sex workers, 50 penitentiary prisoners, 208 healthy children aged 1 to 14 years, 54 cord blood samples, 52 patients with non-A-E hepatitis, 74 patients with hepatitis C, and 51 blood donors with increased ALT levels. Primers specific for the non-coding region were used. The genotype distribution was studied in 70 TTV-positive samples.
Results
The prevalence rate of TTV among the Czech population was 52.6%. We have shown that TTV is not transmitted prenatally. Children were infected after birth with two peaks: one at the age of two years and the other after the beginning of primary school. Adults have shown a further increase in the TTV prevalence with age. The highest TTV prevalence was found in the group of patients who had received multiple blood transfusions. The TTV prevalence rate in subjects at an increased risk of sexual transmission was not significantly higher than in the general population. Genotypes G2 and G1 were most prevalent among the Czech population, followed by G8 and G3. The subjects positive for markers of HBV and/or HCV infection tested significantly more often TTV DNA positive, which is suggestive of a common route of transmission of these three infections.
Conclusions
This study on TTV prevalence, mode of transmission and age-specific prevalence is the most extensive study performed in Central and Eastern Europe. It showed insights into the epidemiology of TTV infection, but failed to associate TTV infection with clinical manifestations.
==== Body
Background
TTV, a non-enveloped small circular single-stranded DNA virus, was recently placed in a novel virus family named Circinoviridae [1]. In spite of being a DNA virus, TTV has an extremely wide range of sequence divergence. At least 40 TTV genotypes from five major phylogenetic groups (G1–G5) have been identified [2,3]. The evolutionary distance between the classified genotypes, as measured by nucleotide substitutions per site, is greater than 30% in the N22 region of ORF1 [4].
TTV has a worldwide distribution. The prevalence of the most common TTV genotypes G1 and G2 is similar all over the world, while the reports on the distribution of other genotypes are scarce and not conclusive. The rate of TTV DNA detection is influenced by the selection of the primer annealing sites. PCR using primers which target the ORF1 region can detect only TTV genotypes 1–6 of group 1, but PCR primers designed for the NCR can detect nearly all genotypes or genetic groups known so far [2].
The prevalence of TT virus in the general population has ranged between 1.9 and 98%, with the highest rates detected in the African and South American countries [5,6]. The TTV prevalence and genotype distribution in Central and East Europe have not yet been determined. One Polish study has reported the prevalence of TTV in blood donors to be 10% by ORF1 primers and 78% by NCR primers [7]. In a small group of blood donors in the Czech Republic, Krekulova et al. detected a TTV prevalence rate of 13.5% by ORF1 primers [8].
The mechanisms of TTV transmission have not yet been elucidated. Even though numerous studies have suggested that the parenteral transmission via transfusion of contaminated blood and blood products is the most common route of TTV infection [9], the detection of TTV in many individuals with no history of blood transfusion indicates that other routes of transmission of TTV may exist. This assumption has been further supported by the detection of TTV in saliva [10], breast milk [11], semen [12] and vaginal fluid [13]. There is evidence that TTV is excreted into feces of infected individuals, suggestive of possible fecal-oral transmission [14]. Some studies have reported placental transmission of TTV [15-17], whereas others have not detected TTV in cord blood and amniotic fluid [18,19]. Since children of TTV-infected mothers apparently tend to get infected more often and earlier after birth than children of TTV negative mothers, the role of postnatal transmission of TTV is being considered [20,21]. Furthermore, variation in the TTV prevalence in children from 5.1% in Japan [22] to 54% in the Democratic Republic of Congo [23] is also suggestive of the possible involvement of other specific environmental factors in the acquisition of TTV infection.
TTV was originally isolated from a blood-transfused patients in which increased alanine aminotransferase (ALT) levels were detected [24]. Therefore, TTV was thought to be the possible etiological factor of non-A-G hepatitis. However, further research has ruled out the notion that a clinically evident liver disease is a consequence of TTV infection. Other studies, which investigated the possible link of TTV to other than hepatic diseases are scarce and so far fail to show any association of TTV infection with clinical manifestation. Nevertheless, the spectrum of diseases studied in association with TTV infection is very narrow and justifies keeping TTV in the category of "orphan" viruses [25].
To investigate the possible routes of TTV transmission, age-specific prevalence and genotype distribution in the population of the Czech Republic, we analyzed sera of 854 subjects divided into 5 groups based on the type of risk of TTV transmission. The age-specific distribution of TTV and correlation with the anti-HBV and/or anti-HCV status were determined. TTV genotypes were determined in 70 patients selected from the different groups.
Methods
Population studied
The Human Subjects Committee of the Institutional Review Boards approved all experimental protocols, and all subjects enrolled in the study signed an informed consent form. Additional samples were obtained from the collection of the national reference laboratory for viral hepatitis (NRL-VH). Five groups of subjects were analyzed:
Group 1 (normal population, control group) was selected from 778 healthy blood donors (mean age 29 years, age range 18–59 years). All donors had normal ALT levels, and were negative for anti-HCV, anti-HIV and HBsAg. Since the TTV DNA prevalence in the first 100 subjects tested was very high we decided to randomly select sera [26] to include 20 first-time donors and 10 regular donors from each of five age groups (18–20, 21–30, 31–40, 41–50 and 50–60 years). In total, 136 sera of first-time donors and 60 sera of regular blood donors were analyzed.
Group 2 (at high risk of parenterally transmitted infection) consisted of 20 hemophiliacs (peripheral blood mononuclear cells (PBMCs) were collected from 10 patients) (mean age 42 years, age range 21–74 years) and 49 IVDUs. Patients with hemophilia were screened for anti-HCV, anti-HIV, anti-HBc and HBsAg, IVDUs were tested for anti-HBc and anti-HCV.
Group 3 (at high risk of sexually transmitted infection) included 85 sex workers and 15 promiscuous men (mean age 25 years, age range 18–43 years). All of them were screened for anti-HCV, anti-HBc, HbsAg and HIV.
Fifty penitentiary prisoners (mean age 29, age range 16–58 years) were tested for a potential increased risk of parenteral and/or sexual transmission of the virus. They were screened for anti-HCV and HBsAg.
Group 4 (at risk of transuterine and mother to child virus transmission) consisted of 54 cord blood samples and 208 sera of children selected by age (we analyzed 28–30 subjects in each of the following 7 age groups: 1, 2, 3, 5, 8, 11 and 14-year-olds).
Group 5 (at risk of potential etiological involvement of TTV) included 52 patients with non-A-E hepatitis (mean age 40 years, age range 9–76 years), 51 blood donors with elevated ALT levels (mean age 39 years, age range 25–64 years) and 74 patients with hepatitis C (mean age 27 years, age range 2–56 years). All blood donors with elevated ALT levels and all patients with non-A-E hepatitis tested negative for anti-HCV and HBsAg. All blood donors were also negative for anti-HIV. All patients with hepatitis C were HBsAg negative.
DNA purification
DNA extraction from sera
DNA was extracted from 200 μL of serum using the QIAamp Blood kit (QIAGEN Ltd., Crawley, UK) and dissolved in 100 μL of elution buffer (QIAGEN Ltd., Crawley, UK). Extracted DNA was stored at -20°C.
DNA extraction from PBMCs
PBMCs were separated by centrifugation from the whole blood on a Ficoll-Paque gradient (SIGMA, St. Louis, MO) according to the manufacturer's protocol. PBMCs were digested with proteinase K (100 μg/ml) (SIGMA, St. Louis, MO) in 1 ml of lysis buffer (50 mM Tris-HCl, pH 8.0; 1 % Tween; 5 mM EDTA, pH 8.0). Thereafter, proteinase K was inactivated for 10 min at 95°C and samples were stored at -20°C.
Polymerase chain reaction
Five microlitres of total DNA were analyzed in a PCR thermocycler PTC 200 (MJ Research, Inc, Waltham, MA). TTV DNA detection was performed with two sets of primers. A 271 bp long fragment of the ORF1 was amplified in a semi-nested PCR with a modified primer set designed by Okamoto [4]. By nested PCR a 110 bp fragment was amplified from the NCR [2]. For the first PCR with ORF1 specific primers, 50 pmol of both modified primers NG059mod (3' CAGACAGAGGMGAAGGMAAYATG 5') and NG063mod (3'CTGGCATYTYWCC MTTTCCAAARTT 5') were used in a 50 μl reaction mixture containing 10 mM Tris-HCl, pH 8.8, 50 mM KCl, 0.8% Nonidet P40, 200 μM each dNTP and 2.5 U Taq polymerase (Fermentas, Hanover, MD). Each of the 35 cycles consisted of 30 s of denaturation at 94°C, 30 s of annealing at 58°C and 45 s of elongation at 72°C. The last cycle was followed by 7 min incubation at 72°C. One microliter of the product from the first PCR was transferred to 50 μl reaction mixture with primers NG061 (3' GGMAAYATGYTRTGGATAGACTGG 5') and NG063mod. The reaction mixture and cycling conditions for the second PCR were the same except that 25 cycles were run.
For the PCR with NCR specific primers the same conditions as for ORF1 specific PCR were used. In the first PCR we used 50 pmol of primer NG133 (3'GTAAGTGCACTTCCGAATGGCTGAG 5') and NG147 (3'GCCAGTCCCGAGCCCG AATTGCC 5'), for the second PCR 50 pmol of primers NG134 (3'AGTTTTCCA CGCCCGTCCGCAGC 5') and NG132 (3'AGCCCGAATTGCCCCTTG AC 5'). Each of the 35 cycles of the first PCR and 25 cycles of the second PCR consisted of 30 s of denaturation at 94°C, 30 s of annealing at 58°C and 45 s of elongation at 72°C, with a final extension for 7 min at 72°C. Ten microlitres of the PCR product were separated electrophoretically on a 3% agarose gel (NuSieve 3:1, FMC BioProduct, Rockland, ME).
TTV viral load in serum and PBMCs of patients with hemophilia
DNA extracted from sera and PBMCs of 4 patients with hemophilia was serially diluted (in 10-fold steps) in distilled water and TTV DNA was determined by PCR with NCR primers. The highest dilution (10N) testing positive was used as a relative titer for determining the viral titer per 1 ml of TTV DNA in serum and PBMCs.
Sequence analysis
A nucleotide DNA sequence of PCR ORF1 products (selected from all groups of study subjects), which revealed a clear band on the agarose gel, was determined. PCR products were excised from a 3% agarose gel and purified with MinElute Gel Extraction kit (QIAGEN Ltd., Crawley, UK) according to the manufacturer's protocol. Both strands of the 271 bp long products generated by PCR with ORF1-specific primers were sequenced directly with the NG061 and NG063mod primers using the ABI Big Dye Sequencing kit (Applied Biosystems, Foster City, CA). The sequencing was performed on an ABI PRISM 310 automated DNA sequencer (Applied Biosystems, Foster City, CA).
Phylogenetic analysis
DNA sequences (of the following accession numbers: AY429576 – AY429589, AY433961 – AY434008, AY456097 – AY456103, AY484597) were aligned using the CLUSTAL X program [27] with the corresponding 222 bp long ORF1 region of previously reported sequences (of the following accessions numbers: AB017768 – AB017770, AB017774 – AB017779, AB017886, AB018889, AB018961, AB021796, AB021798, AB021800, AB021803, AB021815, AF060546, AF060547, AF072749, AF077274, AF079541, AF123914, AF123948, AF124009, AF124027) obtained from GenBank at NCBI (NCBI, Bethesda, MA). Phylogenetic trees were constructed using the neighbor-joining and maximum likelihood method in PHYLIP package, version 3.5 [28].
Serology of hepatitis B and C
HBsAg (V2) AxSYM, CORE AxSYM, AUSAB AxSYM, HCV3.0 AxSYM (Abbott, Chicago, IL) tests were used for the detection of HBsAg, anti-HBc, anti-HBs and anti-HCV markers. HBsAg reactive samples were confirmed with the HBsAg Confirmatory AxSYM test.
Statistical analysis
The statistical analysis was performed using the Fisher exact test. Odds ratios (OR) with 95% confidence intervals (CI) and two-tailed P values were calculated in 2 – 2 tables using the EPI INFO statistical package (version 2002) and GraphPad InStat (version 3.05) (GraphPad Software, San Diego, CA). In all tests, the basic significance level was P = 0.05.
Results
In total, 854 samples were screened for the presence of TTV DNA using a NCR-PCR. The TTV genotype was determined by sequencing part of the ORF1 region from 70 TTV isolates.
Prevalence of TTV DNA in different groups of subjects
The prevalence rates of TTV DNA in different groups are shown in Table 1. The prevalence of TTV DNA in healthy blood donors representing the normal population was 52.6% (103/196). We found no difference in the prevalence of TTV DNA between first-time and regular blood donors (results not shown). The highest prevalence rates were recorded in group 2 (at higher risk of parenteral transmission of infection): 95% (19/20) (OR = 17.16, CI 2.25–130.73, P = 0.0002) for hemophiliacs and 91.8% (45/49) (OR = 10.16, CI 3.52–29.34, P < 0.0001) for IVDUs. In group 3, sex workers and promiscuous men had a prevalence of TTV DNA comparable with that of the normal population, i.e. 62% (62/100), but penitentiary prisoners had a significantly higher prevalence of TTV DNA, i.e. 74% (37/50) (OR = 2.38, CI 1.21–4.69, P = 0.011) than blood donors. We detected no TTV DNA in the cord blood samples, but children had a similar prevalence of TTV DNA as the control group, i.e. 67.8% (141/208). Patients in group 5 had a slightly higher prevalence of TTV DNA than healthy blood donors, the difference being significant only for patients with hepatitis C virus infection, i.e. 89.2% (66/74) (OR = 7.45, CI 3.40–16.3, P < 0.0001).
Age-specific prevalence of TTV DNA
As indicated in Figure 1, we observed a dramatic increase in TTV-prevalence during the first two years of life (at the age of 2 years it equaled 85.7%). The prevalence decreased to 43.3% for 8-year-olds and started to rise again to 73.3% in 14-year-olds. Figure 2 shows an age-dependent distribution pattern of TTV DNA prevalence in the three groups of subjects (blood donors, a group at increased risk of sexual infection, and a combined group of patients with non-A-E hepatitis and hepatitis C). TTV prevalence showed a tendency to increase with a significant linear trend for the group at a higher risk of sexual transmission (Chi square for trend= 8.002, p = 0.0047). Even although the control group showed an obvious increase in the TTV DNA prevalence with age, this trend was not significant.
TTV presence and viral load in serum and PBMCs
As indicated in Table 2 for patients with hemophilia, TTV DNA was more often detected in serum than in PBMCs. TTV DNA was found in sera of all patients (10/10, 100%) and in most (7/10, 70%) PBMCs. Seven patients were positive in serum and PBMCs samples, three subjects had TTV DNA detectable only in serum. The amplification of the internal control beta-globin gene was positive for all PBMCs samples. Viral loads in PBMCs and in corresponding serum samples were compared for four hemophiliacs. All individuals had TTV DNA titers 10 to 100 times higher in their sera than in PBMCs.
Heterogeneity of TTV genotypes
The results of our sequencing and phylogenetic analysis are summarized in Table 3 and Figure 3. The topography of the tree using either the neighbour-joining or maximum likelihood method was identical. We analysed 9 samples from group 1, 23 from group 2, 15 from group 3, 10 from group 4 and 13 from group 5. According to Okamoto's classification [2], 9 isolates were classified into genotype G1a (2 patients with hemophilia, 3 penitentiary prisoners, 1 child, 3 patients with hepatitis C), 18 isolates into genotype G1b (3 blood donors, 3 IVDUs, 1 prostitute, 3 penitentiary prisoners, 4 children, 2 blood donors with increased ALT levels and 2 patients with non-A-E hepatitis), 18 isolates into genotype G2b (4 blood donors, 5 patients with hemophilia, 1 IVDU, 3 sex workers, 2 children, 1 blood donors with an elevated ALT level, 1 patient with hepatitis C and 1 patient with non-A-E hepatitis), 21 isolates into genotype G2c (2 blood donors, 5 patients with hemophilia, 3 IVDUs, 4 sex workers, 1 penitentiary prisoner, 3 children, 2 blood donors with increased ALT levels, 1 patient with hepatitis C), 3 isolates into genotype G8 (3 patients with hemophilia) (Table 3) and 1 isolate from a IVDU (I5s) was most closely related to genotype G2. The similarity of this product with a G2a reference genotype sequence (AB017770) was 74.8 %; its similarity with the reference sequence of G2c (AB017768) was 73.9% on the nucleotide level. In the Czech population, the most prevalent genotype was G2c (30.0%), followed by G1b (25.7%), G2b (25.7%), G1a (12.9%) and G8 (4.3%). No association between any of the detected genotypes and a particular population group was revealed.
Comparison of TTV genotypes present in sera and PBMCs of the same patient
Tree of four patients positive in both serum and PBMCs by the ORF1 PCR system yielded a PCR product adequate for sequencing analysis. Genotypes and sequence homologies are shown in Table 2. Two paired sequences were of the same genotype, G2b and G2c, respectively, and the sequence similarity between the genotype detected in PBMCs and serum was 96.8% and 100%, respectively. In one patient we detected genotype G8 in PBMCs and genotype G2b in serum (similarity 32.6%).
Co-infection markers in association with TTV DNA presence
Four groups of subjects were compared for correlation of past HBV and/or HCV infection with TTV DNA prevalence. We compared subjects at a higher risk of sexual transmission, IVDUs, patients with hemophilia and penitentiary prisoners. There was no evidence of present or past HCV or present HBV infection in blood donors. The prevalence of anti-HBc in blood donors in the Czech Republic is very low (1–2%, unpublished data). These data imply that our control group was at a low risk of sexually and parenterally acquired infections. Also all children were anti-HBc and anti-HCV negative. Both the groups of hemophiliacs and IVDUs presented evidence of frequent past or current HBV (50% and 22%) and HCV (82% and 72%) infection. Thirteen percent of sex workers and promiscuous men showed past or current HBV exposure. The study subjects positive for HBV and/or HCV markers had a significantly higher prevalence of TTV DNA regardless of the primer set used (Table 4).
Discussion
The present study on TTV prevalence, mode of transmission and age-specific prevalence is the largest study performed in Central and East Europe. This study showed that TTV infection is quite common among the Czech population. The prevalence rates of TTV in blood donors (52.6%) were similar to those found in other developed countries (for review see [29]).
The most interesting result was the lack of TTV DNA in 54 cord blood samples, suggestive of the absence of transuterine transmission of TTV. Furthermore, we have shown that the prevalence of TTV DNA was age dependent. In children, we observed a dramatic increase of TTV prevalence within the first two years of age. The prevalence of TTV further gradually increased in children aged 8 to 14 years. Similarly, Ohto et al. reported that the TTV prevalence in children at the age of 2 years was comparable with that in mothers, while children younger than 3 months of age were infected only exceptionally [19]. In the Czech Republic, children start schooling at the age of six or seven years. Based on our data, postnatal infection from mother to child and an increased number of social contacts are likely to be the most important routes of TTV transmission in children.
In adults, we have shown the increase of TTV prevalence with age irrespective of study group. Similar results have been reported by others [21,30-33].
We observed an increased prevalence of TTV in various groups of hepatitis patients. Since many hepatitis viruses share the same modes of transmission, multiple viral infections may occur in one patient [34]. Our results showed a significantly higher prevalence of TTV infection in patients with hepatitis C than in healthy individuals, implying that HCV and TTV may share common modes of transmission.
In agreement with other studies we detected the highest prevalence of TTV in hemophiliacs and IVDUs, which supports the importance of the parenteral route of transmission of TTV [9,35-38]. Nevertheless, the prevalence of 52.6% in blood donors and in healthy children suggests that the higher prevalence in hemophiliacs and IVDUs can also be attributed to a higher TTV viral load. Different TTV concentrations of virus in hemophiliacs and blood donors have been previously reported by Touinssi [39]. Additionally, Simmonds has shown that in hemophiliacs the prevalence of TTV increased with the amount of clotting factor treatment received and was also dependent on whether the blood concentrates tested had been virally inactivated [9].
As for TTV sexual transmission, the results of our study suggest that if TTV is sexually transmitted, this mode of transmission is likely to be less important. Even though 13% subjects of the group at risk for sexual transmission of TTV showed past or current HBV exposure, indicative of high promiscuity, the prevalence of TTV was not significantly different from that of the control group (62% and 52.6%). To the best of our knowledge, so far only one study found a significant difference in the TTV prevalence in sex workers [40], while others did not.
In the group of penitentiary prisoners the significantly higher prevalence of TTV seems to be most probably a consequence of intravenous drug abuse rather than sexual promiscuity since the prevalence of HCV was four-times higher than in the group at risk for sexual transmission (20% versus 5%).
An increased prevalence of TTV in non-A-E hepatitis patients was observed in the present study in agreement with many previous studies. Additionally, our data showed an increased TTV prevalence in blood donors with increased ALT levels. Several studies have shown a correlation between TTV-titer and elevation of serum ALT levels [24,30] but the experimental infection of chimpanzees with TTV did not show any biochemical or histological evidence of hepatitis [1].
The heterogeneity of TTV is extreme. Because the NCR region is too conserved for evolutionary analyses, the ORF1 region of the TTV genome is most often used for genotyping. Out of the 70 sequenced isolates, G2 followed by G1 were the most frequent genotypes among the Czech population. The phylogeny analysis showed no evidence of association of particular TTV genotypes with any of the risk groups.
Conclusions
Our results demonstrated a high prevalence of TTV in the Czech population. Our data show the absence of transuterine transmission of TTV, but postnatal route of transmission from mother to child and infection via frequent social contacts seem to be very important modes of transmision in children. The sexual mode of transmission is most likely to be low effective. No convincing evidence was found to support the involvement of TTV in the pathogenesis of hepatitis.
Our data, as well as the results of other studies, show that optimization of the primer set for more standard TTV detection and genotyping is still needed. Improved serological approach to TTV detection could be of value. It is evident that more data are still needed for a better understanding of the natural history of TTV infection.
Competing interests
The author(s) declare that they have no competing interests.
Authors'contributions
RT and VN conceived of the study and participated in its design and coordination. MS carried out most of the experimental work and participated on the preparation of the initial draft of the manuscript. Part of the experimental work was done by JK. RT did all the statistical analysis and evaluation of the results and writing of the final version of the manuscript. All authors contributed to the preparation of the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by the Grant Agency of the Czech Republic (grant project 310/00/1060).
Figures and Tables
Figure 1 Age specific prevalence of TTV in children Age-dependent prevalence of TTV DNA in children
Figure 2 Age specific prevalence of TTV in adults Age-specific prevalence of TTV DNA in the control group, in the group at increased risk of sexually transmitted infection and in patients with non-A-E hepatitis or hepatitis C as detected with NCR-specific primers
Figure 3 Phylogeny analysis of the Czech TTV isolates Phylogenetic tree constructed by neibor-joining method from a 222 bp fragment of the ORF1 region obtained by sequencing 70 isolates from the Czech population (accession numbers: AY429576 – AY429589, AY433961 – AY434008, AY456097 – AY456103, AY484597) and from previously reported sequences (accessions numbers: AB017768 – AB017770, AB017774 – AB017779, AB017886, AB018889, AB018961, AB021796, AB021798, AB021800, AB021803, AB021815, AF060546, AF060547, AF072749, AF077274, AF079541, AF123914, AF123948, AF124009, AF124027). The groups in which the isolates from our study were situated are circled and the I5s isolate which is located on a separate branch is in bold. Bootstrap values above 75% are shown.
Table 1 Prevalence of TTV DNA by study groups
Number Study group Number of subjects Number of TTV DNA positives (%) Difference (p value)
1 blood donors 196 103 (52.6%)
2 patients with hemophilia 20 19 (95.0%) <0.001
2 IDVUs 49 45 (91.8%) <0.0001
3 sex workers 100 62 (62.0%) NS
2, 3 penitentiary prisoners 50 37 (74.0%) <0.05
4 children 208 141 (67.8%) <0.005
4 cord blood samples 54 0 (0%) -
5 patients with non A-E hepatitis 52 39 (75.0%) <0.005
5 patients with HCV 74 66 (89.2%) <0.0001
5 blood donors with increased ALT levels 51 31 (60.8%) NS
Study groups: 1 = controls, 2 = subjects at high risk of parenterally transmitted infection, 3 = subjects at high risk of sexually transmitted infection, 4 = subjects at risk of transuterine or mother to child transmitted infection, 5 = subjects at risk of potential etiological involvement of TTV
Table 2 Comparison of TTV viral loads in sera and PBMCs of patients with hemophilia
Sample Type of material Genotype Similarity of genotypes (%) Viral load (relative titer 10N/ml)
1 serum G2c 100 107
PBMCs G2c 106
2 serum G2b 96.8 107
PBMCs G2b 106
3 serum G2b 32.6 107
PBMCs G8 106
4 serum - - 107
PBMCs G2c 105
Table 3 TTV GENOTYPES BY STUDY GROUPS
TTV genotype
Group Number of positive subjects Number of sequenced samples G1a G1b G2b G2c G8 other
blood donors 14 9 - 3 4 2 - -
patients with hemophilia 14 15a 2 - 5b 5c 3d -
IDVUs 14 8 - 3 1 3 - 1
sex workers 10 8 - 1 3 4 - -
penitentiary prisoners 10 7 3 3 - 1 - -
children 17 10 1 4 2 3 - -
cord blood samples 0 0 - - - - - -
patients with non A-E hepatitis 7 3 - 2 1 - - -
patients with HCV 14 5 3 - 1 1 - -
blood donors with increased ALT levels 8 5 0 2 1 2 - -
Total (%) 108 70 9 (12.9) 18 (25.7) 18 (25.7) 21 (30.0) 3 (4.3) 1 (1.4)
a) 11 PCR products amplified from sera and 4 PCR products amplified from DNA extracted from PBMCs
b) 1 PBMC sample positive for G2b
c) 2 PBMC samples positive for G2c
d) 1 PBMC sample positive for G8
Table 4 TTV presence in relation to the past infection with HBV and/or HCV
HBV and HCV serology TTV positive subjects
PCR OR CI P
anti-HBc positive 39/42 (92.9%) 5.452 1.61–18.44 <0.01
negative 124/176 (70.5%)
Anti-HCV positive 54/59 (91.5%) 4.954 1.87–13.14 0.0004
negative 109/159 (68.6%)
==== Refs
Mushahwar IK Erker JC Muerhoff AS Leary TP Simons JN Birkenmeyer LG Chalmers ML Pilot-Matias TJ Dexai SM Molecular and biophysical characterization of TT virus: evidence for a new virus family infecting humans Proc Natl Acad Sci U S A 1999 96 3177 3182 10077657 10.1073/pnas.96.6.3177
Okamoto H Takahashi M Nishizawa T Ukita M Fukuda M Tsuda F Miyakawa Y Mayumi M Marked genomic heterogeneity and frequent mixed infection of TT virus demonstrated by PCR with primers from coding and noncoding regions Virology 1999 259 428 436 10388666 10.1006/viro.1999.9770
Peng YH Nishizawa T Takahashi M Ishikawa T Yoshikawa A Okamoto H Analysis of the entire genomes of thirteen TT virus variants classifiable into the fourth and fifth genetic groups, isolated from viremic infants Arch Virol 2002 147 21 41 11855633 10.1007/s705-002-8301-7
Okamoto H Nishizawa T Kato N Ukita M Ikeda H Iizuka H Miyakawa Y Mayumi M Molecular cloning and characterization of a novel DNA virus (TTV) associated with posttransfusion hepatitis of unknown etiology Hepatology Research 1998 10 1 16 10.1016/S1386-6346(97)00123-X
Niel C de Oliveira JM Ross RS Gomes SA Roggendorf M Viazov S High prevalence of TT virus infection in Brazilian blood donors J Med Virol 1999 57 259 263 10022797
Prescott LE Simmonds P Global distribution of transfusion-transmitted virus N Engl J Med 1998 339 776 777 9742036 10.1056/NEJM199809103391118
Grabarczyk P Brojer E Polymorphism of the TT virus and its frequency in Polish blood donors Vox Sang 2002 82 177 181 12047511 10.1046/j.1423-0410.2002.00178.x
Krekulova L Rehak V Killoran P Madrigal N Riley LW Genotypic distribution of TT virus (TTV) in a Czech population: evidence for sexual transmission of the virus J Clin Virol 2001 23 31 41 11595582 10.1016/S1386-6532(01)00185-8
Simmonds P Davidson F Lycett C Prescott LE MacDonald DM Ellender J Yap PL Ludlam CA Haydon GH Gillon J Jarvis LM Detection of a novel DNA virus (TTV) in blood donors and blood products Lancet 1998 352 191 195 9683208 10.1016/S0140-6736(98)03056-6
Ross RS Viazov S Runde V Schaefer UW Roggendorf M Detection of TT virus DNA in specimens other than blood J Clin Virol 1999 13 181 184 10443794 10.1016/S1386-6532(99)00015-3
Schroter M Polywka S Zollner B Schafer P Laufs R Feucht HH Detection of TT virus DNA and GB virus type C/Hepatitis G virus RNA in serum and breast milk: determination of mother-to-child transmission J Clin Microbiol 2000 38 745 747 10655378
Inami T Konomi N Arakawa Y Abe K High Prevalence of TT Virus DNA in Human Saliva and Semen J Clin Microbiol 2000 38 2407 2408 10835017
Fornai C Maggi F Vatteroni ML Pistello M Bendinelli M High prevalence of tt virus (ttv) and ttv-like minivirus in cervical swabs J Clin Microbiol 2001 39 2022 2024 11326040 10.1128/JCM.39.5.2022-2024.2001
Okamoto H Akahane Y Ukita M Fukuda M Tsuda F Miyakawa Y Mayumi M Fecal excretion of a nonenveloped DNA virus (TTV) associated with posttransfusion non-A-G hepatitis J Med Virol 1998 56 128 132 9746068
Gerner P Oettinger R Gerner W Falbrede J Wirth S Mother-to-infant transmission of TT virus: prevalence, extent and mechanism of vertical transmission Pediatr Infect Dis J 2000 19 1074 1077 11099089
Goto K Sugiyama K Ando T Mizutani F Terabe K Tanaka K Nishiyama M Wada Y Detection rates of TT virus DNA in serum of umbilical cord blood, breast milk and saliva Tohoku J Exp Med 2000 191 203 207 11038012
Morrica A Maggi F Vatteroni ML Fornai C Pistello M Ciccorossi P Grassi E Gennazzani A Bendinelli M TT virus: evidence for transplacental transmission J Infect Dis 2000 181 803 804 10669384 10.1086/315296
Iso K Suzuki Y Takayama M Mother-to-infant transmission of TT virus in Japan Int J Gynaecol Obstet 2001 75 11 19 11597614 10.1016/S0020-7292(01)00450-7
Ohto H Ujiie N Takeuchi C Sato A Hayashi A Ishiko H Nishizawa T Okamoto H TT virus infection during childhood Transfusion 2002 42 892 898 12375662 10.1046/j.1537-2995.2002.00150.x
Hsieh SY Wu YH Ho YP Tsao KC Yeh CT Liaw YF High prevalence of TT virus infection in healthy children and adults and in patients with liver disease in Taiwan J Clin Microbiol 1999 37 1829 1831 10325332
Simmonds P Prescott LE Logue C Davidson F Thomas AE Ludlam CA TT virus--part of the normal human flora? J Infect Dis 1999 180 1748 1750 10516086 10.1086/315103
Goto K Sugiyama K Terabe K Mizutani F Wada Y Detection rates of TT virus among children who visited a general hospital in Japan J Med Virol 1999 57 405 407 10089054 10.1002/(SICI)1096-9071(199904)57:4<405::AID-JMV13>3.0.CO;2-F
Davidson F MacDonald D Mokili JL Prescott LE Graham S Simmonds P Early acquisition of TT virus (TTV) in an area endemic for TTV infection J Infect Dis 1999 179 1070 1076 10191206 10.1086/314730
Nishizawa T Okamoto H Konishi K Yoshizawa H Miyakawa Y Mayumi M A novel DNA virus (TTV) associated with elevated transaminase levels in posttransfusion hepatitis of unknown etiology Biochem Biophys Res Commun 1997 241 92 97 9405239 10.1006/bbrc.1997.7765
Bendinelli M Pistello M Maggi F Fornai C Freer G Vatteroni ML Molecular properties, biology, and clinical implications of TT virus, a recently identified widespread infectious agent of humans Clin Microbiol Rev 2001 14 98 113 11148004 10.1128/CMR.14.1.98-113.2001
Website title [http://www.randomizer.org]
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools Nucleic Acids Res 1997 25 4876 4882 9396791 10.1093/nar/25.24.4876
Felsenstein PHYLIP:Phylogeny Inference Package.Version 3.5 1993 Seatle,WA:Department of Genetics, University of Washington
Allain JP Emerging viral infections relevant to transfusion medicine Blood Rev 2000 14 173 181 11124105 10.1054/blre.2000.0140
Kanda T Yokosuka O Ikeuchi T Seta T Kawai S Imazeki F Saisho H The role of TT virus infection in acute viral hepatitis Hepatology 1999 29 1905 1908 10347137 10.1002/hep.510290613
MacDonald DM Scott GR Clutterbuck D Simmonds P Infrequent detection of TT virus infection in intravenous drug users, prostitutes, and homosexual men J Infect Dis 1999 179 686 689 9952377 10.1086/314642
Saback FL Gomes SA de PV da Silva RR Lewis-Ximenez LL Niel C Age-specific prevalence and transmission of TT virus J Med Virol 1999 59 318 322 10502263 10.1002/(SICI)1096-9071(199911)59:3<318::AID-JMV10>3.0.CO;2-Q
Yazici M Comert MR Mas R Guney C Cinar E Kocar IH Transfusion-transmitted virus prevalence in subjects at high risk of sexually transmitted infection in Turkey Clin Microbiol Infect 2002 8 363 367 12084105 10.1046/j.1469-0691.2002.00423.x
Pontisso P Ruvoletto MG Fattovich G Chemello L Gallorini A Ruol A Alberti A Clinical and virological profiles in patients with multiple hepatitis virus infections Gastroenterology 1993 105 1529 1533 8224658
Cao K Mizokami M Orito E Ding X Ge XM Huang GY Ueda R TT virus infection among IVDUs in south western China Scand J Infect Dis 1999 31 21 25 10381213 10.1080/00365549950161835
Katsoulidou A Paraskevis D Anastassopoulou CG Chryssou SE Sypsa V Boletis J Malliori M Karafoulidou A Tassopoulos NC Hatzakis A Prevalence and genotypic distribution of TT virus in Athens, Greece J Med Virol 2001 65 423 429 11536255 10.1002/jmv.2052
Poovorawan Y Theamboonlers A Jantaradsamee P Kaew-in N Hirsch P Tangkitvanich P Hepatitis TT virus infection in high-risk groups Infection 1998 26 355 358 9861559
Zehender G Manzin A De Maddalena C Colasante C Solforosi L Corsi F Bianchi-Bosisio A Girotto M Schirru I Russo U Galli M Clementi M Molecular epidemiology of TT virus in Italy and phylogenesis of viral isolates from subjects at different risk for parenteral exposure J Med Virol 2001 63 76 84 11130892
Touinssi M Gallian P Biagini P Attoui H Vialettes B Berland Y Tamalet C Dhiver C Ravaux I de Micco P de Lamballerie X TT virus infection: prevalence of elevated viraemia and arguments for the immune control of viral load J Clin Virol 2001 21 135 141 11378494 10.1016/S1386-6532(01)00157-3
Huang YH Wu JC Lin CC Sheng WY Lee PC Wang YJ Chang FY Lee SD Prevalence and risk factor analysis of TTV infection in prostitutes J Med Virol 2000 60 393 395 10686021 10.1002/(SICI)1096-9071(200004)60:4<393::AID-JMV5>3.0.CO;2-6
| 15575965 | PMC539280 | CC BY | 2021-01-04 16:03:31 | no | BMC Infect Dis. 2004 Dec 3; 4:56 | utf-8 | BMC Infect Dis | 2,004 | 10.1186/1471-2334-4-56 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-4-581559601210.1186/1471-2334-4-58Study ProtocolCan somatostatin control acute bleeding from oesophageal varices in Schistosoma mansoni patients?[ISRCTN63456799] Chatterjee Shyama [email protected] Marck Eric [email protected] Laboratory of Pathology, Faculty of Medicine, University of Antwerp, Universiteitsplein-1, 2610 Antwerp, Belgium2004 13 12 2004 4 58 58 15 11 2004 13 12 2004 Copyright © 2004 Chatterjee and Van Marck; 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
Management of patients with bleeding oesophageal varices comprises of mainly diagnostic endoscopy, sclerotherapy and band ligation. One of the major problems to do any of the above is the active bleeding which makes any intervention difficult. The neuropeptide hormone somatostatin administered exogenously has caused a reduction in portal hypertension and variceal bleeding in patients suffering from liver cirrhosis. We believe that the symptomatic use of somatostatin for variceal bleeding in Schistosoma mansoni infected subjects can reduce bleeding, thereby alleviating the pathology caused by schistosomiasis.
Methods/design
We herein present a study protocol for establishing this neuropeptide as a potential therapeutic agent in schistosomiasis. Adolescent subjects, age range varying from 12–17 years will be selected, based on several inclusion criteria, most important being infection with Schistosoma mansoni with bleeding from oesophageal varices in the last 24 hours. One group of schistosomiasis patients will be treated with somatostatin and praziquantel, the other with propanolol and praziquantel. Survival graphs will be set up to correlate somatostatin administration with survival time. A two part questionnaire will be set up to control treatment outcomes. The pre-treatment part of the clinical questionnaire will identify inclusion criteria questions, the post-treatment part of the questionnaire will identify treatment outcomes.
Discussion
We expect that the administration of somatostatin as a bolus followed by a 24 hour long infusion, will stop bleeding immediately, delay rebleeding as compared to the control study group and delay mortality in the somatostatin treated subjects.
==== Body
Background
Complications due to severe schistosomiasis include fibrosis, hepatomegaly, splenomegaly, haematemesis, varices, portal hypertension, ascites formation and death [1]. Complications resulting from hepatic fibrosis (such as portal hypertension and variceal bleeding) are the principal cause of death in S. mansoni infected patients. In such patients portal hypertension leads to the formation of portal-systemic collaterals, specifically gastro-oesophageal collaterals (varices). Bleeding of these oesophageal varices can occur depending on the severity of fibrosis, and can be fatal. Praziquantel, the most commonly used anti-schistosomal drug, is effective against the worm stages of the parasite, but has no activity against the pathology (fibrosis) caused by the egg stages or the variceal bleeding that can be fatal in its outcome. These observations have stressed the need to combine praziquantel treatment with symptomatic treatment like with somatostatin.
Somatostatin is emerging as the ideal vasoactive drug for the control of variceal bleeding, and is as effective as sclerotherapy [2-4]. In a recent clinical trial, cirrhotic patients with acute bleeding of oesophageal varices were treated with infusion followed by bolus injections of somatostatin just before sclerotherapy. Results showed fewer treatment failures, fewer deaths or use of rescue therapy, reduced blood transfusion and less frequent, active bleeding. This drug also prevents recurrence and is free from any major side effects even when administered over prolonged periods of time.
We have studied the potential role of somatostatin in modulating Schistosoma caused morbidity. In endemic regions, at any given time, only a fraction of infected patients develop severe hepatic fibrosis. There may be a direct correlation between the development of severe fibrosis and the inability to generate required somatostatin levels. Our ongoing research at the Laboratory of Pathology give evidence to this fact, since somatostatin levels in Senegalese patients with severe morbidity (haematemesis, portal hypertension, variceal bleeding, ascites, fibrosis) are significantly lower than that in patients with no severe morbidity [5-10]. This effect indicates that somatostatin could be administered exogenously for therapeutic purposes in chronic schistosomiasis patients.
Complications linked to hepato-intestinal schistosomiasis are increasing in the area of Richard-Toll, Northern Senegal. These cases are being identified at the local health centre. However, considering the high prevalence of schistosomiasis in this region, it is likely that the number of severe cases is much higher. Clear guidelines for the management of such severe complications and criteria for referral and surgery are required in this region. The establishment of an algorithm on how to treat these patients and create the appropriate infrastructure is urgently needed. The priority is how to take care of these patients and more so what is the best way of doing this under local conditions. One of the first steps is to critically evaluate the Niamey-Belo-Horizonte ultrasound methodology in patients with severe periportal fibrosis. Efficient management of bleeding varices in the afflicted patients is imperative.
Given the background that somatostatin is an ideal vasoactive drug in the field of liver pathology, it is our opinion that somatostatin will be more efficacious and safe as compared to currently used beta blocker drugs like propanolol, in the control of acute oesophageal variceal bleeding due to Schistosoma mansoni infection. Moreover using this neuropeptide may increase time to failure of drug treatment, decrease incidences of early re-bleeding (day 4, 8) and incidences of death during the follow up period. Decreased frequencies of late rebleeding (days 30, 60, 90) may occur, all indicating the safety of using somatostatin. Praziquantel cover would be given to all study patients.
Study design
1. Selection of patients
Age and Morbidity criteria – Adolescent subjects, age range varying from 12–17 years will be selected. The inclusion criterion will be schistosomiasis patients with bleeding from oesophageal varices in the last 24 hours. A random selection will be made to form two groups, a study group and a control group. Control of active infection will be done by means of CAA-strips on urine or blood. Subjects will be asked to fill in an informed consent form and the pre-treatment part of a questionnaire.
The inclusion criteria will be established fibrosis due to schistosomiasis of clinical history, physical examination and laboratory findings (and an examination compatible with the presence of portal hypertension due to fibrosis). Clinically active upper gastrointestinal bleeding (haematemesis of fresh or semi fresh blood and/or melena and/or haematochezia) with or without haemodynamic instability (systolic blood pressure < 80 mm Hg and heart rate > 120 bpm) will be selected. Subjects must be male or non-pregnant, non-lactating female subjects. Females of childbearing potential will have to utilize contraception for the duration of the study. Written or verbal documented informed consent will be needed from all subjects.
Exclusion criteria will include participation by subjects in another investigational study within the last 14 days. Subjects may not undergo treatment with endotherapy, i.e. band ligation, sclerotherapy or other (balloon tamponade). Treatment with somatostatin, vasopressin or their analogues will also be a exclusion criteria. Subjects with end stage liver disease with hepatorenal syndrome, diffuse hepatocellular carcinoma, patent porto-systemic shunts, known diagnosis of non-fibrotic portal hypertension, severe cardiovascular diseases, i.e. acute myocardial infarction and heart failure will be excluded. Concurrent use of metoclopramide is also not advised.
Conduct of trial – Active bleeding episode (haematemesis, haematochezia, melena) from a potential variceal source should be confimed by a medical team (the ER physician, the ICU physician, the investigator). Patients may be outpatients or already hospitalised patients. Patients will be randomised to either arm in a sequential manner. Randomisation, and the start of study drug infusion if in adjunctive therapy arm, should be accomplished as soon as possible following identification of a patient qualifying for the study and following the conduct of pre-randomisation study procedures.
Sample Size Criteria – Assuming a 80% chance of finding a significant difference <0.05 between the two study cohorts, the following statistics were established:
(A) If 99% of the untreated subjects and –
10% of somatostatin treated subjects bled -5 volunteers per group were sufficient;
20% treated subjects bled – 7 volunteers per group were required;
30% treated subjects bled-9 volunteers per group were needed.
(B) If 90% of untreated subjects bled, and –
10% of somatostatin treated group bled – 7 volunteers in each cohort were sufficient;
20% of somatostatin treated group bled – 9 volunteers in each group were required;
30% of somatostatin treated group bled – 10 volunteers in each group were needed.
(C) If 70% of untreated subjects bled, and:
10% of somatostatin treated group bled – 12 volunteers per group were sufficient;
20% of somatostatin treated group bled – 18 volunteers per group were required;
30% of somatostatin treated group bled – 28 volunteers per group were needed.
In cirrhotic patients with bleeding oesophageal varices somatostatin administration controls bleeding in more than 80% of the treated patients. Based on this report, we propose to start this pilot study with 10 subjects/group.
2. Treatment
Two groups of 10 schistosomiasis patients each will be identified. Group (1) will be treated with Somatostatin (3.5 μg/kg/hour; single bolus and i.v. infusion for 24 hours) + Praziquantel (40 mg/kg). For somatostatin treatment the i.v. infusion will be started first; 3 mg somatostatin will be dissolved in the 1 ml of saline provided. This solution will be added to the saline transfusion unit and administered to the patient for the next 12 hours. Once finished the second packet of 3 mg somatostatin will be used similarly for a second saline transfusion unit for the remaining 12 hours. The bolus dose of 250 μg will be dissolved in the 1 ml of saline provided and administered over 90 seconds soon after the start of the i.v. infusion.
Group (2) will be treated with Propanolol + Praziquantel.
3. Data analysis
Survival graphs will be set up to correlate somatostatin administration with survival time.
The primary efficacy variable is the number of patients meeting the failure of therapy definition during the infusion period. Failure criteria are defined as death during infusion, persistence of active bleeding (The haemodynamic instability criteria points to the inability to achieve and maintain a systolic blood pressure of 80 mm Hg OR presence of a 20 mmHg drop in systolic blood pressure from the highest post resuscitation value AND achieving a heart rate of 120 bpm OR a 20 bpm increase from highest post resuscitation value OR Inability to achieve and maintain a Hct of – 27% of Hb of – 9 g/dl despite blood transfusion of 2 units or more.
The clinical criteria of active bleeding include hematemesis (fresh or semi fresh blood), hematochezia, melena.
4. Control points
A two part questionnaire will be set up to control treatment outcomes. The pre-treatment part of the clinical questionnaire will identify inclusion criteria questions:
(1) When was the last haematemesis incidence?
(2) When did the present incident start?
The post-treatment part of the questionnaire will identify treatment outcomes. The following questions will be answered:
(1) What was the reaction to somatostatin infusion?
(2) When did bleeding stop after somatostatin infusion?
(3) GI disturbances: Is there abdominal pain, nausea, diarrhea, after somatostatin infusion?
(4) Control of early rebleeding: Is there early rebleeding in the 8 days following somatostatin administration?
(5) Control of late rebleeding: When is the next rebleeding incident? Subjects will be followed up on a monthly basis, by home visits.
(6) Control of mortality: Is there any difference in mortality time between the two groups.
(7) Control of fibrosis: Does ultrasonography detect anti-fibrotic effect of somatostatin after treatment?
Discussion
It is expected that the administration of somatostatin as a bolus followed by a 24 hour long infusion, will stop bleeding immediately, delay rebleeding as compared to the control study group and delay mortality in the somatostatin treated subjects. Our protocol that is based on a pilot study will help to establish the importance of somatostatin in schistosomiasis.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Both authors participated in the design of the study.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
WHO Expert Committee The control of schistosomiasis WHO Tech Rep Ser 1983 830 1 86
Fevery J Raptis SA 'Optimal management of upper gastrointestinal bleeding' in Research and Clinical forums 1997 19
Avgerinos A Nevens F Raptis S Fevery J Early administration of somatostatin and efficacy of sclerotherapy in acute oesophageal variceal bleeds: The European Acute Bleeding Oesophageal Variceal Episodes (ABOVE) randomised trial Lancet 1997 350 1495 1499 9388396 10.1016/S0140-6736(97)05099-X
Bloom SR Polak J Somatostatin Br Med J 1987 295 288 90 2888511
Chatterjee S Mbaye A Alfidja AT Weyler J Scott JT Van Damme P Van De Vijver K Deelder A Van Marck EAE Circulating levels of the neuropeptide hormone somatostatin may determine hepatic fibrosis in Schistosoma mansoni infections Acta Trop 2004 90 191 203 15177146 10.1016/j.actatropica.2003.12.002
Chatterjee S Mbaye A De Man JG Van Marck EAE Does somatostatin have therapeutic potential against schistosomiasis? Somatoselective 2003 2 10
Chatterjee S Mbaye A Van Marck EAE Lower levels of the circulating neuropeptide somatostatin in S. mansoni infected patients may have pathological significance Trop Med Int Health 2003 8 33 36 12535248 10.1046/j.1365-3156.2003.00989.x
Chatterjee S Mbaye A De Man JG Van Marck EAE Does the neuropeptide somatostatin have therapeutic potential in schistosomiasis? Trends Parasitol 2002 18 295 298 12379948 10.1016/S1471-4922(02)02294-8
Chatterjee S Van Marck E The role of somatostatin in schistosomiasis: a basis for immunomodulation in host-parasite interactions? Trop Med Int Health 2001 6 578 581 11555424 10.1046/j.1365-3156.2001.00758.x
Chatterjee S De Man JG Van Marck EAE Somatostatin and intestinal schistosomiasis: therapeutic and neuropathological implications in host-parasite interactions Trop Med Int Health 2001 6 1008 1015 11737838 10.1046/j.1365-3156.2001.00744.x
| 15596012 | PMC539281 | CC BY | 2021-01-04 16:03:31 | no | BMC Infect Dis. 2004 Dec 13; 4:58 | utf-8 | BMC Infect Dis | 2,004 | 10.1186/1471-2334-4-58 | oa_comm |
==== Front
BMC UrolBMC Urology1471-2490BioMed Central London 1471-2490-4-141558831010.1186/1471-2490-4-14Research ArticleExpression of pS2 in prostate cancer correlates with grade and Chromogranin A expression but not with stage Ather M Hammad [email protected] Farhat [email protected] Nuzhat [email protected] M [email protected] Shahid [email protected] Dept. of Surgery, Aga Khan University, Karachi, Pakistan2 Dept. of Pathology, Aga Khan University, Karachi, Pakistan2004 10 12 2004 4 14 14 22 9 2003 10 12 2004 Copyright © 2004 Ather 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 biological potential of prostate cancer is extremely variable. Particular interest is focused on markers not expressed in normal prostatic tissues. pS2 protein expression has been demonstrated in a range of malignant tissues in an oestrogen-independent pathway. Recently, it has been demonstrated that pS2, in prostate cancer, is closely associated with neuro-endocrine differentiation. In the present study, we have analyzed, the potential of Neuro-endocrine and pS2 (TFF1) expression in human prostate cancer determined by immunohistochemistry, in primary adenocarcinoma of the prostate and attempted to correlate this with the clinico-pathologic features of the patient and neuroendocrine expression.
Methods
Ninety-five malignant prostatic specimens from primary adenocarcinoma, obtained from either transurethral resection of prostate or radical retropubic prostatectomy, from 84 patients between January 1991 and December 1998 were evaluated by immuno-histochemical staining using selected neuroendocrine tumor markers i.e. chromogranin A (CgA) and estrogen inducible pS2 protein. The relationship between the expressions of pS2 was studied with CgA expression, clinical stage (TNM) and tumour grade (Gleason system). Fischer exact test was used for statistical analysis.
Results
The mean age of the patients was 70 + /- 9.2 years. The pS2 expression was seen in 10% of primary prostate cancers. Worsening histological grade was associated with greater expression of pS2 (p < 0.001). The expression of CgA was noted in 31% of malignant prostatic tissue. In pS2, positive cases 2/3rd of patients were also CgA +ve. However, there was no significant correlation between pS2 expression and the stage of disease.
Conclusion
pS2 expression in prostate cancer significantly correlates with histological grade and the neuroendocrine differentiation, as demonstrated by Chromogranin A expression but not with the clinical stage of the disease. However, the overall expression was low consequently; no definitive conclusions can be drawn. We feel further work is required in a larger series, both in primary and metastatic cancer.
==== Body
Background
The biological potential of prostate cancer is extremely variable [1]. It is perhaps the only cancer, which could be managed by, deferred treatment in its early course for selected cancers [2]. To define the biological potential of prostate cancer, prognostic markers are employed. There are numerous markers for assessing the biological aggressiveness of the prostate cancer [3]. However, large studies have shown that they lack sensitivity and specificity due primarily to their expression in normal prostatic epithelium as well. This justifies a recent surge in interest in markers specific to malignant prostatic tissue [4]. Recent studies have shown the potential of neuro-endocrine differentiation in adenocarcinoma of the prostate and its role in ascertaining the biological aggressiveness of the tumor [3]. Wang et al [5] has recently noted that the expression of the pS2 protein is implicated in the pathogenesis and progression of some neuro-endocrine tumors.
Maisakowski et al. first described the pS2 gene in the MCF-7 human breast cancer cell line [6]. The pS2 is a cysteine rich secretory protein, containing 84 amino acids and a molecular weight of 6.45 k-Da. The pS2 gene is highly expressed in estrogen-receptor positive breast cancer, and high levels of pS2 protein correlate with responsiveness to primary endocrine therapy and better patient survival in breast cancer. However, in prostate cancer it is linked with NE differentiation and poorer outcome [7].
In the present study, we have investigated the expression of pS2 in malignant primary prostatic tissue in specimens obtained from transurethral, open prostatectomy, and correlated this with neuro-endocrine differentiation and clinical stage and grade. This is a preliminary report on pS2 expression in prostate cancer, a larger study will better define the correlation between stage, grade of cancer with pS2 and CgA expression.
Methods
Demographic profile
Ninety-five malignant consecutive primary prostatic specimens were obtained from 84 patients by either trans-urethral resection of prostate (n = 69 patients) for urinary obstruction or from radical retro-pubic prostatectomy (n = 15 patients) between January 1991 and December 1998. These tissue specimens were taken from the archived records of the department of pathology. The age ranged from 52–93 years (mean 70 + 9.2 years).
Immuno-histochemical staining for pS2 and Chromogranin A Sections were stained for H & E as well as for pS2 (Novocastra, UK Cat. # NCL-pS2) and Chromogranin A (DAKO, Glostrup, Denmark Cat # A0430) by immunohistochemistry using indirect immunoperoxidase technique.
Briefly, 3 μm thick tissue sections were cut and mounted on poly-L-lysine (sigma) coated slides. Sections were deparaffinized in xylene and re-hydrated through graded alcohol series followed by water. Antigen retrieval was done in case of pS2 with 10 mM citrate buffer, 6.0 in a microwave oven 3 × 5 seconds at 450 W, then gradually cooled down to room temperature.
Sections were washed with water followed by Phosphate buffer saline (PBS) rinse.
Endogenous peroxidase in the sections was blocked for 30 minutes with 0.3% H2 02 in methanol. Sections were washed with PBS. All sections were treated with Normal Swine serum (NSS) prediluted 1:10 in PBS for 5 minutes. The sections were then incubated with the primary antibody to pS2 diluted with NSS (1:100) and Chromogranin A (1:20) for 90 minutes at room temperature. Slides were washed with PBS and incubated with peroxidase-conjugated swine anti rabbit secondary antibody (DAKO) at a dilution of 1:150 for 45 minutes at room temperature. 3, 3'-diaminobenzidine (DAB) was used as a final Chromogen. Harris Haemtoxylin was used as a counter nuclear stain. Positive and negative controls were used with all batches of IHC staining. A prostatic adenocarcinoma specimen section expressing pS2 was used as a positive control. Same case exhibiting the primary antibody was used a negative control with each staining procedure. The extent of pS2 reactivity was semi quantitatively assessed by estimating the percentage of positive acini present in the whole mounted sessions. Expression was graded ++ if more than 50% of the tissue showed expression, + if between 5 and 49% showed expression and focal if <5% showed expression.
Histological grading The Gleason system was used for grading of the cancer specimens; a senior histopathologist (SP) blinded of previous Gleason grading and clinical course did this. Based upon the Gleason score patients were divided into three groups i.e. well differentiated (Gleason sum 2–4), moderately differentiated (Gleason sum 5–7) and poorly differentiated (Gleason sum 8–10).
To study correlation and determine the p value Student t test was applied.
Results
The cancerous lesion composed of 35% (n = 29) stage T1, 32% (n = 27) stage T2, 25% (n = 21) stage T3 and 6% (n = 5) stage T4 disease according to the TNM classification. Based upon the stage of the disease patients were divided into three groups i.e. organ confined (T1-2), locally invasive (T3-4 and N1) and metastatic (M1) cancer.
In 95-cancer specimen from transurethral resection (n = 69) for urinary obstruction and radical retropubic prostatectomy for organ-confined cancers (n = 15), pS2 reactivity was detected in the adjoining normal or hyperplastic acini in only 4.2%. The pS2 expression in cancer was found in 10% (figure 1). The immuno-histochemical reactivity of pS2 in malignant epithelial cells was confined to the cytoplasm of with a tendency to a perinuclear accentuation.
Expression of pS2 was correlated with the stage of disease in Figure 1. Staining for NE marker (CgA) was seen in 31% (figure 1); correlation between the pS2 and CgA expression is summarized in table 1, it showed that 2/3rd of pS2 also showed CgA expression.
Worsening histological grade was associated with greater expression of pS2. In Gleason sum groups 2–4 and 5–6, expression of pS2 was noted in 6 and 8% respectively whereas in Gleason sum group 8–10 the expression was observed in 30% (p < 0.001). The expression of pS2 [figure 4(a) and 4(b)] in various prognostically and therapeutically distinct groups based upon the grade of cancer is described in figure 2.
Discussion
In the present study, we investigated the expression of pS2 protein in the adenocarcinoma of prostate and in the surrounding normal prostate tissue. We used a standard immunohistochemical method to assess pS2 expression in tissue sections of adenocarcinoma prostate instead of instead of biochemical or immuno-radiometric assay. The immuno-histochemical method for detection of pS2 expression has drawbacks in comparison to biochemical and immunoradiometric assay on tissue extracts. Both of the later methods allow precise quantification of levels of expression for a better correlation with other parameters studied.
However, as we are interested in the clinical utility of pS2 expression in our prostate cancer population, we used immunohistochemistry, which allows appreciation of intra-tumoral heterogeneity of expression and of both cancerous and non-cancerous cells. pS2 protein expression has been demonstrated in a range of malignant and benign pathologies. It is highly expressed in receptor positive human breast cancer [5] but expression in other cancers like ovarian [7], cervical [8], gastrointestinal [9], thyroid [10] and bladder [11] is variable.
A significant implication of pS2 in prostate cancer is the close association of this marker with Neuroendocrine (NE) differentiation. There is increasing evidence that focal NE differentiation frequently occurs in prostatic adenocarcinoma and it may have significant prognostic implications [12-14]. NE differentiation is also described in hormone refractory prostate cancer; Krijnen et al [14] noted that androgen receptors are not present in prostatic adenocarcinoma staining positive for CgA. While Higashiyama noted 17% expression of pS2 in all pulmonary cancers, Wang et al [5] noted 45% expression in small cell cancers of the lung (a neuroendocrine carcinoma). Recent evidence has suggested that expression of pS2 is closely associated with neuroendocrine differentiation in prostate cancer [15]. Colombel et al from in an RT-PCR study found a high expression of pS2 in prostate cancer; however, they found no correlation between with tumour stage or Gleason grade. Our present work [15] indicates that NE differentiation not only correlates with other prognostic markers like grade of the cancer but also has independent prognostic value. Bonkhoff et al [15] noted that pS2 expression was consistently confined to NE differentiation in untreated tumors and in carcinomas that relapsed after hormonal therapy. Our results have similarly shown that 6 out of 9 cancers that have expressed pS2 were also positive for CgA.
Conclusions
Our results demonstrate that although the expression of pS2 protein was noted in only 1/10th of prostate cancers, it significantly correlates with the histological grade and NE differentiation; both have independent and interdependent prognostic value. There is dearth of data exploring the correlation of pS2 expression and aggressiveness of prostate cancer cell behavior. Limited literature available at present show significant association of pS2 expression with prognosis in prostate cancer, however more work is required to explore the utility of this marker in defining the biological potential of prostate cancer.
Competing interest
The author(s) declare that they have no competing interests.
Authors' contributions
MHA, conceived of the idea, wrote the manuscript and conducted clinical part of the study. FA, helped in designing the study and reviewed the draft of the manuscript
NF, helped in conducting study, helped in data collection and analysis. MI, conducted the pathological part of the study. SP, conducted and supervised the pathological aspects of the study and reviewed the manuscript and wrote methods and results related to the pathology. All authors' have read and approve of the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Expression of pS2 % in organ confined, locally advanced (locoregional) and metastatic adenocarcinoma of the prostate.
Figure 2 Correlation of Gleason sum and pS2 expression.
Figure 3 Photomicrograph of prostatic adenocarcinoma stained with polyclonal antibody against CgA. Note prominent cytoplasmic staining within tumour cells. Magnification: 4×
Figure 4 Photomicrograph of prostatic adenocarcinoma stailed with a polyclonal against pS2. Note prominent perinuclear brown staining in tumour cells. Magnification: 40×
Figure 5 Photomicrograph of prostatic adenocarcinoma stailed with a polyclonal against pS2. Note prominent perinuclear brown staining in tumour cells. Higher magnification (100×).
Table 1 Correlation of pS2 expression with CgA expression and Gleason sum (GS)
n (%) CgA +ve CgA -ve Mean GS
pS2 +ve. 9(10) 67% 33% 7.5
pS2 -ve. 86(90) 28% 72% 6.2
+ve. Expression positive
-ve. Expression negative
CgA Chromogranin A
GS Gleeson sum
==== Refs
Harding MA Theodorescu D Prognostic markers in localized prostate cancer: from microscope to molecules Cancer-Metastases-Rev 1999 17 429 37 10.1023/A:1006137721823
Ather MH Abbas F Prognostic significance of neuroendocrine differentiation in prostate cancer Eur Urol 2000 38 535 42 11096233 10.1159/000020352
Brasso K Frijis S Juel K Jorgensen T Iversen P Morbidity in patients with clinically localized prostate cancer managed with non-curative intent. A population based case control study Prostate Cancer Prostatic Dis 1999 2 253 256 12497171 10.1038/sj.pcan.4500378
Erbersdobler A Fritz H Schnoger S Graefen M Hammereer P Huland H Henke RP Tumor grade, proliferatio, apoptosis,microvessel density, p53, and bcl 2 in prostate cancers: differences between tumours located in the transition zone and in the peripheral zone Eur Urol 2002 41 40 6 11999464 10.1016/S0302-2838(01)00021-5
Yang Y Chisholm GD Habib FK The distribution PSA, Cathepsin D and pS2 in BPH and cancer of the prostate The Prostate 1992 21 201 08 1279646
Wang DG Johnston CF Liu WH Sloan JM Buchanan KD Expression of abreast-cancer-associated protein (pS2) in human neuro-endocrine tumors Int J Cancer 74 270 4 1997 Jun 20 9221803 10.1002/(SICI)1097-0215(19970620)74:3<270::AID-IJC6>3.0.CO;2-W
Masiakowski P Breathnach R Bloch J Gannon F Krust A Chambon P Cloning of cDNA sequences of hormone-regulated genes from the MCF-7 human breast cancer cell line Nucleic Acids Res 10 7895 903 1982 Dec 20 6897676
Luqmani YA Ryall G Shousha S Coombes RC An immuno-histochemical survey of pS2 expression in human epithelial cancers Int J Cancer 50 302 4 1992 Jan 21 1730524
Looi LM Azura WW Cheah PL Ng MH pS2 expression in infiltrating ductal carcinoma of the breast correlates with oestrogen receptor positivity but not with histological grade and lymph node status Pathology 2001 33 283 6 11523925 10.1080/00313020120062848
Wysocki SJ Hahnel E Masters A Smith V McCartney AJ Hahnel R Detection of pS2 messenger RNA in gynecological cancers Cancer Res 50 1800 2 1990 Mar 15 2306733
Lefebvre O Chenard MP Masson R Linares J Dierich A LeMeur M Wendling C Tomasetto C Chambon P Rio MC Gastric mucosa abnormalities and tumorigenesis in mice lacking the pS2 trefoil protein Science 274 259 62 1996 Oct 11 8824193 10.1126/science.274.5285.259
Metaye T Millet C Kraimps JL Aubouin B Barbier J Begon F Estrogen receptors and cathepsin D in human thyroid tissue Cancer 72 1991 96 1993 Sep 15 8364879
Sagol O Yorukoglu K Tuna B Ozer E Sis B Guray M Mungan U Kirkali Z Expression of pS2 protein and its relation with the Ki-67 proliferative indices and tumor recurrence in superficial bladder carcinomas Eur Urol 2001 40 163 8 11528193 10.1159/000049767
Weinstein MH Partin AW Veltri RW Epstein JI Neuroendocrine differentiation in prostate cancer: enhanced prediction of progression after radical prostatectomy Hum Pathol 1996 27 683 7 8698312 10.1016/S0046-8177(96)90398-6
Krijnen JL Janssen PJ Ruizeveld de Winter JA van Krimpen H Schroder FH van der Kwast TH Do neuroendocrine cells in human prostate cancer express androgen receptor? Histochemistry 1993 100 393 8 8307781 10.1007/BF00268938
Bonkhoff H Stein U Welter C Remberger K Differential expression of the pS2 protein in the human prostate and prostate cancer: association with premalignant changes and neuroendocrine differentiation Hum Pathol 1995 26 824 8 7635445 10.1016/0046-8177(95)90002-0
Colombel M Dante R Bouvier R Ribieras S Pangaud C Marechal JM Lasne Y Differential RNA expression of the pS2 gene in the human benign and malignant prostatic tissue J Urol 1999 162 927 30 10458410 10.1097/00005392-199909010-00092
Ather MH Abbas F Faruqui N Israr M Pervez S Prognostic impact of neuroendocrine differentiation in prostate cancer Fourth international symposium on Urology, Nephrology and transplantation and Fourth international symposium on urology and transplantation, transplantation society of SAARC countries 2001
| 15588310 | PMC539282 | CC BY | 2021-01-04 16:03:53 | no | BMC Urol. 2004 Dec 10; 4:14 | utf-8 | BMC Urol | 2,004 | 10.1186/1471-2490-4-14 | oa_comm |
==== Front
BMC BiolBMC Biology1741-7007BioMed Central London 1741-7007-2-251557420410.1186/1741-7007-2-25Research ArticleSearch for computational modules in the C. elegans brain Reigl Markus [email protected] Uri [email protected] Dmitri B [email protected] Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 11724, USA2 Department of Molecular Cell Biology and Department of Physics of Complex Systems, Weizmann Institute of Science, Rehovot, 76100, Israel2004 2 12 2004 2 25 25 1 12 2003 2 12 2004 Copyright © 2004 Reigl 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
Does the C. elegans nervous system contain multi-neuron computational modules that perform stereotypical functions? We attempt to answer this question by searching for recurring multi-neuron inter-connectivity patterns in the C. elegans nervous system's wiring diagram.
Results
Our statistical analysis reveals that some inter-connectivity patterns containing two, three and four (but not five) neurons are significantly over-represented relative to the expectations based on the statistics of smaller inter-connectivity patterns.
Conclusions
Over-represented patterns (or motifs) are candidates for computational modules that may perform stereotypical functions in the C. elegans nervous system. These modules may appear in other species and need to be investigated further.
==== Body
Background
There is little doubt that neurons are elementary building blocks of the nervous system [1]. It is less clear, however, whether multi-neuron modules (smaller than invertebrate ganglia or vertebrate nuclei and cortical columns) can be meaningfully defined, either anatomically [2] or physiologically [3]. The existence of such multi-neuron modules would greatly simplify the description of nervous system structure and function. An example of such simplification can be found in electrical engineering. An electronic circuit is often represented in terms of modules such as operational amplifiers, logical gates and memory registers rather than as a wiring diagram showing each transistor, resistor and diode. However, unlike electrical engineers who designed these modules themselves, neurobiologists did not design the brain, and evolution rarely leaves records of its experimentation. Therefore, if multi-neuron modules have indeed evolved they need to be discovered.
In this paper, we search for anatomically defined multi-neuron modules in the Caenorhabditis elegans nervous system. We choose C. elegans as a model organism because its wiring diagram is largely known, including the identities of all 302 neurons and most synapses between them [4-6]. Our approach follows the reasoning developed previously in the context of gene regulation and other networks [7,8]. If a certain multi-neuron module performs some stereotypical function it may appear in the nervous system repeatedly. Therefore, search for multi-neuron connectivity patterns that appear more often than by "chance" (compared with the expectations as defined below) may yield these multi-neuron modules. Of course, there may be functionally important modules that appear infrequently and would be missed by our analysis. In the electronic circuit analogy, our approach would discover logical gates in a processor wiring diagram but not a rectifier in a power supply, which is essential but appears only once.
To search for N-neuron modules, we sort all N-neuron combinations into classes defined by their inter-connectivity pattern and count the number of combinations in each class. By comparing these counts with the mean counts from random networks, constructed based on our expectations, we detect significantly over-represented patterns, or motifs. In order to avoid assigning significance to a N-neuron pattern just because it contains N-1-neuron motifs we incorporate the N-1-neuron statistics into the expectations used to search for N-neuron motifs [8]. To do this, we perform our search sequentially, by starting with doublets (or neuronal pairs, N = 2) and then increasing the number N of neurons included in the pattern sequentially up to quintuplets (N = 5).
We look for motifs in the wiring diagram of the C. elegans nerve ring (a large fraction of the nervous system) assembled in two datasets [6]. Datasets 1 and 2 were obtained from serial-sections electron microscopic (EM) reconstructions of two different animals [4]; for details see Methods. The datasets contain the numbers of synapses formed in a subset of C. elegans neurons. Two given neurons may be connected by more than one synapse, which we call the multiplicity of connection. However, the small size of the dataset compels us to use the binary representation of these connections (connected or unconnected). In order to obtain binary connectivity matrices, we threshold the multiplicity of connections at various values Θ: Pairs having less than Θ synapses are considered unconnected while those having at least Θ synapses are considered connected. Such procedure is justified because more than a single synaptic contact may be necessary for an observable physiological effect of one neuron on another. Since we do not know the physiologically relevant count of synapses, we repeat our calculation for 1 ≤ Θ ≤ 7.
Unfortunately, datasets 1 and 2 contain a caveat of synaptic ambiguities, which arises from the limitations of EM in C. elegans. When one pre-synaptic neuron makes contact with two adjacent processes of different neurons (send_joint in Durbin notation [6]), it is not known which of these processes acts as a post-synaptic terminal; both might be involved. We address this ambiguity by performing our analysis in two ways. In the main text we present the results obtained on the datasets that include both send and send_joint synaptic connections. We repeated the analysis on the datasets where send_joint synapses were split equally between the two potential post-synaptic partners. Specifically, we calculated multiplicity of connections by adding send_joint synapses at 50% synaptic strength. In the limit of high multiplicity, this is equivalent to assigning the post-synaptic neuron by chance. We find essentially the same results for this connectivity dataset (see Supplementary Information [Additional file 1]).
Results
Bi-directionally connected doublets (N = 2) are over-represented
We classify all possible doublets (or pairs) of the C. elegans neurons into three classes: unconnected, uni-directionally connected and bi-directionally connected, and compare the number of doublets in each class to that expected in a random network (Figure 1). The random network ensemble consists of connectivity matrices that preserve the numbers of incoming and outgoing synapses for each neuron but not the identities of the partners [9,10]. The motivation behind this choice of the random matrix ensemble and the details of the algorithm are explained in Methods.
We find that the number of doublets in each class deviates from the mean of the random matrix counts, as shown in Figure 1 for a representative threshold Θ = 3. For the purposes of module search, the most interesting finding is the over-representation of the reciprocally connected doublets (pattern #3), for two reasons. First, if a set of neurons were to function as a module it should not consist of two (or more) disconnected subsets. This consideration rules out pattern #1. Second, since our search for modules is aimed at identifying over-represented inter-connectivity patterns we are less interested in under-represented ones. This consideration rules out pattern #2. We note that pattern counts are not independent, but are subject to sum rules. For example, the number of neurons in the network fixes the total doublet count. Also, the total number of connections is equal to the count of pattern #2 plus twice the count of pattern #3. These sum rules place stringent constraints on possible combinations of doublet counts. Yet, for patterns with greater number of neurons (N>2), these constraints become less stringent because the number of patterns increases (see below).
We repeat the above calculations for other datasets and threshold values and consistently find the significant over-representation of bi-directionally connected doublets (data not shown). In C. elegans, such over-representation was reported previously on a qualitative level [4]. Interestingly, an over-representation of bi-directionally connected doublets was also found for pyramidal neurons in mammalian neocortex [11-13]. This suggests that motifs may represent evolutionary conservation or convergence driven by similar computational constraints. Next, we discuss whether C. elegans can provide a clue to the functional significance of the over-representation of reciprocally connected doublets.
Can bilateral (left-right) symmetry of the C. elegans neuronal network account for the over-representation of the reciprocally connected doublets? Indeed, about two thirds of C. elegans neurons have a bilaterally symmetric partner. If connections between these pairs obeyed bilateral symmetry then they could not be uni-directional, creating a bias in favor of bi-directional connections. To see whether this is the case, we calculate the percentage of bi-directional connected doublets, which consist of a bilateral neuron pair. We find that these percentages are small: 7.1% and 5.5% in datasets 1 and 2, respectively. Therefore, bilateral symmetry is not sufficient to explain the observed result.
The over-representation of reciprocally connected doublets in C. elegans has been explained [6] as a consequence of correlation between adjacency and connectivity of neurons. The argument is that, if there is a synapse from neuron A to neuron B, they must be adjacent. If neurons A and B are adjacent then a synapse from B to A is more likely than chance, increasing the probability of a reciprocal connection. Analysis of original EM reconstructions [4] supports this argument [6,14]. Adjacency in this case does not refer to the nearby placement of cell bodies but to the number of EM sections (divided by five) in which the processes of the two neurons are in contact [6,14].
Although correlation between adjacency and connectivity may account for the over-representation of reciprocally connected doublets, why such correlation would exist in C. elegans remains unclear. It could be that the number of neuronal pairs, which can be adjacent, is limited by physical constraints. This would restrict the adjacent pairs only to the ones that need to connect for functional reasons. Indeed, volume exclusion explains neuron dimensions in the cortical column ([15] and references therein). In the C. elegans network, however, the small number of neurons should in principle allow a contact between any pair of neurons. This argument is supported by the observation that many neuronal processes are longer than the distance between the corresponding cell bodies, suggesting that the connection can be made. However, processes tend to run in bundles and make synapses only in their (often varying) neighborhoods [14]. This suggests that other (e.g. developmental) constraints may restrict the number of adjacent neurons. Alternatively, it could be that network functionality requires over-representation of reciprocal connections (or clustering). These issues must be explored in the future.
Several triplet classes (N = 3) are over-represented
We classify all connected triplets in the C. elegans wiring diagram into 13 classes and count the number of triplets in each class. We compare the actual number of triplets in each class to the null-hypothesis random matrix ensemble defined as follows. In order to include the observed over-representation of reciprocally connected doublets, we construct random networks that preserve the numbers of bi-directional and uni-directional connections for each neuron. Figure 2 shows triplet counts for each class relative to the mean of the random matrix ensemble. For threshold Θ = 2 we find that several triplet counts are noticeably different from the mean of the random matrix ensemble, e.g. patterns #10, #12, #14 and, possibly, #15 and/or #16 in Figure 2. Similar results were found for other values of the threshold (within the biologically plausible range, Θ = 1 to 7).
Are these differences between triplet counts in actual and random networks significant? One might answer this question by calculating, for each class, a significance p-value, i.e. the probability of finding a random matrix with deviation from the mean exceeding or equal to that for the actual network. Although such an approach would be correct if over-representation of a single class were examined, it would over-estimate the true significance (i.e. under-estimate the p-value) when many different classes are evaluated simultaneously. This situation is known as multiple hypothesis testing and requires an adjustment of the raw p-values (see Methods).
We chose to perform multiple hypothesis testing adjustment by controlling the family-wise error rate, i.e. the probability of mistakenly reporting at least one non-over-represented pattern, by using the single-step min P procedure [16,17]. The adjusted p-values for every class and threshold represent the probability of finding a random matrix R, in which at least one class i has smaller (or equal) raw p-value than that found for a given class and threshold in the actual network. This measure can be calculated by counting the number of random matrices, which have a smaller (or equal) raw p-value (in at least one class) than that in the actual network for a given class and threshold. By dividing this number of matrices by the total size of the random matrix ensemble, we estimate the multiple hypotheses testing corrected significance measure Pm for each class and threshold, Figure 3 (see Methods).
According to the significance measure, Pm, one of the most consistently over-represented motifs is the feedforward loop (triplet pattern #10), previously noticed in C. elegans [5,18] and other networks [7,8]. For the full list of feedforward loops see Supplementary Information [Additional files 2 and 3]. Could some known feature of neuronal organization account for the observed over-representation of the feedforward loop? We consider two hypotheses:
i. The three-layered feedforward neuronal network is not sufficient to account for over-representation of the feedforward loop
If one views the C. elegans nervous system as a three-layer feedforward network, where sensory neurons synapse mostly on interneurons, and interneurons synapse on other interneurons or motorneurons, this could explain the over-representation of the feedforward loop. We argue that this is not the case for two reasons. First, the feedforward loop is also over-represented among interneurons (Figure 4). Second, the three-layer model of the C. elegans nervous system is overly simplified. For example, there are feedback connections from interneurons to sensory neurons and from motorneurons to interneurons. To evaluate whether detected feedforward loops fit the three-layer feedforward network, we analyze the function of the neurons in these loops. About 40% of the detected feedforward loops either contain all neurons from the same functional group or at least one connection goes from a neuron in a lower layer to a neuron in a higher layer, Table 1. These loops do not fit into this three-layer model, undermining the hypothesis.
ii. The likelihood of connectivity between nearby neurons may partially account for over-representation of the feedforward loop
Since connectivity and adjacency are correlated in C. elegans and other nervous systems one could argue the following [4]. If two neurons have a common synaptic partner, then they are likely to be adjacent to that common partner, and hence to each other. If the two neurons are adjacent they are likely to be connected to each other. Again, adjacency cannot refer to the cell body position: The fraction of over-represented triplets that consist of neurons belonging to the same ganglia is typically less than 30%. Yet this argument could be valid if the adjacency refers to the contacts between neuronal processes (see above) and needs to be verified using original EM reconstructions [4]. The problem with this argument is that it would also predict an over-representation of all strongly connected patterns (#10 to #16), as opposed to the weakly connected patterns (#4 to #9). Yet, strongly connected triplet classes #13 and #11 (the feedback loop) are not over-represented (Figure 3) so further explanation is required.
It is possible that the over-representation of the feedforward loop is a consequence of other factors or their combinations (such as feedforwardness and locality of connectivity combined). But even if these factors are found, the characterization of the network in terms of over-represented motifs remains valid. The over-representation of the feedforward loop still requires a functional explanation just as the bi-directionally connected doublet does. In gene transcription regulation networks, the feedforward loop was proposed to carry out information processing functions such as filtering out fluctuations and responding only to persistent stimuli [7]. Feedforward loop can also carry out other functions [5,18], depending on the polarity of synapses involved and the dynamic response of neurons. Once these factors are established experimentally, motif function can be analyzed theoretically.
In addition to the feedforward loop, we find that two other (both symmetric) patterns are consistently over-represented: pattern #12 and pattern #14 (Figure 3). For the full list of these patterns see Supplementary information [Additional files 2 and 3]. Previous work [8] did not identify these patterns as motifs because of their low absolute count at the only threshold considered (Θ = 5). Again, we ask whether this could be a consequence of the bilateral symmetry of the C. elegans nervous system. Indeed, the bilateral symmetry implies that pairs of bilaterally symmetric neurons are also connected symmetrically, meaning that triplets containing such a pair are likely to be symmetric. However, we find that the fraction of triplets #12 and #14 containing a bilaterally symmetric pair of neurons and an unpaired neuron is rather small (between 10% and 20% in datasets 1 and 2). This suggests that the bilateral symmetry of the nervous system is not sufficient to explain the over-representation of pattern #12 and #14.
Just like in any other screening algorithm, our criteria for outliers are somewhat subjective and the goal is to draw attention to interesting candidates. We limit our discussion to over-represented patterns #10, #12 and #14 because in our judgment they are most robust outliers based on the several criteria used. The reader may judge that some other patterns are over-represented as well. For example, patterns #15 and #16 are significantly over-represented for small thresholds (Figure 3). Because the absolute counts of these patterns in the C. elegans network are small, we cannot verify that they are consistently over-represented. Further work on larger datasets will show whether these patterns may be viewed as motifs.
Several quadruplet classes (N = 4) are over-represented
We classify all connected quadruplets into 199 classes and count the number of quadruplets in each class. Then we compare the actual counts of quadruplets in each class to the mean counts of quadruplets in a random matrix ensemble. In this case, random matrices preserve the numbers of uni-directional and bi-directional connections for each neuron and, in addition, the numbers of triplets (see Methods). Because of the large number of quadruplet classes, we show results (Figure 5) only for patterns selected according to the following criteria: the multiple hypothesis testing corrected significance values Pm must be less than 0.1 for at least one threshold per pattern, while the number of quadruplets in the actual network must be at least 5. The last condition excludes patterns that may appear as over-represented due to very small quadruplet counts.
We find that quadruplet pattern #45 is consistently over-represented [8]. Can we explain this observation by some other known factor? We consider the following two hypotheses:
i. Bilateral symmetry of the nervous system is not sufficient to explain the over-representation of the quadruplet pattern #45
One could propose that symmetric patterns should be over-represented because of the bilateral symmetry of the nervous system. We think that this argument by itself cannot explain the observed over-representation for two reasons. First, the fraction of feedforward quadruplets containing two bilaterally symmetric neuron pairs in motif 45 is rather small (less than 10% in dataset 1 and less than 14.3% in dataset 2). Second, many symmetric patterns are not over-represented, such as, for example, patterns 25, 30, 31, 35, 43, 44 and 65 (Figure 6).
ii. Feedforward structure of the nervous system may partially explain the over-representation of the feedforward quadruplet
One could propose that the feedforward three-layer structure of the nervous system could account for this observation (see over-represented triplets). We find that 14% to 37% of the feedforward quadruplets do not fit into this proposition because either they contain a feedback connection or all neurons belong to the same layer (Table 2). After comparing these percentages to the relative excess values we conclude that the feedforward structure may explain over-representation for some threshold values but not for others.
It is possible that some other factors (in addition to feedforwardness) account for the reported quadruplet over-representation. Just as argued in case of triplets, discovering these factors would be complementary to the characterization of the over-represented motif. It would be particularly interesting to determine the functional role of these motifs. Again, we arbitrarily limit our discussion of over-represented quadruplets to pattern #45. The reader may judge that some other patterns are over-represented and deserve attention (e.g. patterns #36, 50). This is why in Figure 5 we show all the outliers satisfying relatively weak criteria.
We find no over-represented quintuplet classes (N = 5)
We classify all connected quintuplets into 9364 classes (out of 9608 patterns total, i.e. connected and unconnected) and count the actual number of quintuplets in each class. We compare these counts with the mean of the random matrix ensemble. In this case, the random matrices preserve the numbers of uni- and bi-directional connections for each neuron and, in addition, keep the numbers of all triplets and quadruplets in a 10% range of the actual network. We do not find any significantly over-represented quintuplets. This could happen because there are no significantly over-represented quintuplets with a given number of quadruplets. Alternatively, this could happen because specifying the numbers of triplets and quadruplets constrains the number of quintuplets in any random matrix the size of the C. elegans network. Therefore, absence of significantly over-represented quintuplets in C. elegans does not rule out the existence of five-neuron modules that can be detected as motifs by applying our algorithm to larger networks.
Discussion
By comparing counts of multi-neuron patterns in the C. elegans wiring diagram to the mean counts of the appropriate random matrix ensemble, we find several over-represented motifs. First, we find that bi-directionally connected doublets (out of three possible doublet classes) are over-represented, given the number of connections on each neuron is fixed. Second, several triplet classes (out of thirteen possible connected patterns) are over-represented, given the actual number of bi-directional (as well as uni-directional) connections for each neuron. Third, we find that several quadruplet classes (out of 199 connected patterns) are over-represented, given the numbers of triplets are preserved in addition to previously listed constraints. We find no over-represented quintuplet classes. Some of these results, such as the over-representation of the feedforward loop and the feedforward quadruplet, have been reported previously [5,8,18]. The current paper extends and complements previous reports by performing a systematic motif search for various connection multiplicity thresholds and rigorous statistical significance assessment. Also, we consider whether the discovered motifs can be accounted for by previously known facts about the organization of the nervous system. There is no functional explanation for the existence of the motifs. Therefore, the identified motifs are candidates for modules that may perform stereotypical functions in the C. elegans nervous system, and they need to be investigated further.
Although the main motivation for this work, search for modules, led to our focus on over-represented patterns, we also checked for under-representation. For example, previous work indicated that the number of triplets with pattern #11 (or feedback loops) was small [6]. To determine significance, we applied the single-step min P procedure to the absolute value of the deviation of counts from the mean. We found that the feedforward loop is not significantly under-represented, yet many other patterns, such as weakly connected triplets were significantly under-represented (see Supplementary Information [Additional File 1]).
Our motif search algorithm is different from previous attempts to find modules [19]. For example, traditional clustering approaches look for the subsets of nodes, which are connected with their own subset more strongly than with other subsets. In our algorithm, we consider all the connections within a pattern (unlike [20], who considered only some connections within the pattern) but ignore the connections with neurons outside the pattern. One could question the expediency of ignoring multiple possible inputs to the neurons in a module since those inputs could influence the operation of that module. To counter this, we point out that if there were a particularly recurring way to attach an external connection to a given N-neuron motif then it would appear as an N + 1-neuron motif. If, on the other hand, the motif is connected in many different ways in different instances, their significance will be washed out. Thus our approach may hierarchically detect modules with recurring input/output sites, growing them out of smaller patterns. A second justification for looking at N-neuron patterns is that the nervous system is capable of performing many different functions under different circumstances and neurons active in one case may be silent in another. Therefore, in any particular case, many of the anatomical inputs to the module may remain silent and can safely be ignored. This speculation may be verified experimentally by simultaneous monitoring of neuronal activity in different neurons.
The strategy and algorithms we described in this paper can be applied to incompletely mapped networks because a highly significant pattern is also likely to be over-represented in a sub-network. However, the statistical power of our algorithm increases with the knowledge of the wiring diagram. Therefore it was natural to choose the C. elegans nervous system, whose wiring diagram is largely known. Unfortunately, C. elegans has some disadvantages when it comes to the interpretation of the results: the polarity of a synapse (excitatory vs. inhibitory) in C. elegans is often unknown; electrophysiological investigations are still difficult in C. elegans [21]; and the whole network contains only 302 neurons, limiting the statistical power of the approach. Yet we hope that recent technological developments [22] will eliminate the first two disadvantages and allow functional analysis of the discovered modules. Moreover, we expect that our results have implications for understanding nervous system structure and function beyond C. elegans. The modules we identify in C. elegans may be a general property of the nervous system, and, once identified, can be searched for in other species. Finally, the algorithm itself can be applied to other networks [8] once they become available.
As in any other theoretical analysis, we made several simplifications. For example, we assumed that the strength of synaptic connection between a pair of neurons is characterized by its multiplicity (i.e. the number of synapses between that pair). This assumption may be questioned if synapses implementing high-multiplicity connections are weaker than those implementing low-multiplicity connections, as known to happen in nematodes [23]. Yet, this assumption represents a reasonable first step in the systematic quantitative analysis, which may be extended in the future by estimating synaptic strength from the original EM reconstructions. In addition, we ignored the polarity of the synapses and the existence of gap junctions. Yet our results are robust to the inclusion of these factors in the future because if an over-represented class is found, it will remain over-represented even if divided into smaller sub-classes. It would be interesting to see whether the inclusion of the above factors will reveal specific over-represented sub-classes.
Conclusions
We have shown that certain neuronal connectivity patterns are significantly over-represented in the C. elegans nervous system. These patterns, called motifs, are candidates for computational modules that may perform stereotypical functions. It would be interesting to determine what these functions are and whether these motifs appear in other nervous systems.
Methods
Representation of the networks
We transformed the C. elegans synaptic connectivity data into a binary matrix A, called Adjacency Matrix or Connectivity Matrix, in which an entry Aij is 1 if there is a connection from neuron i to neuron j and 0 otherwise. The order in which neurons are assigned to rows in this matrix is not important for our calculations. The multiplicity of synapses between two given neurons is mapped to a binary value by applying a threshold to the data. We assume a synaptic connection of threshold Θ from neuron i to neuron j if neuron i makes at least Θ synapses on to neuron j. Adjacency matrices that we used are available in the Supplementary Information [Additional files 2 and 3].
Detecting & counting patterns
We implemented two strategies for counting the number of triplets, quadruplets and quintuplets in a given connectivity matrix. First, to obtain the count of all N-neuron patterns, we took all different N-neuron subsets and characterized their connectivity. Second, we took all possible N-neuron subsets out of the neighborhood of a neuron x. This neighborhood is defined by all neurons that can be reached from x, if the directed connectivity matrix is made undirected. In both cases it is crucial for the run time of the algorithm to detect the pattern class from these connectivity sub-matrices as quickly as possible. We realized this by defining a function that maps each possible N-neuron sub-matrix to a unique integer value. Then we classified all the sub-matrices based on the function value and a pre-calculated lookup table, which identifies the pattern class from the function value.
Creating random matrices
The number of neurons that receive synaptic input from a given neuron x is called out-degree of x. The number of neurons providing synaptic input to neuron x is called in-degree of x. In the binary matrix representation of a network as described above, the out-degree of a neuron x can be calculated as the sum of row x, the in-degree as the sum of column x.
N = 2. For the first step of our analysis we create random matrices that preserve the in-degree and out-degree of every neuron but change their connection partners. Starting with an empty matrix, our algorithm selects neurons in a random order and connects each with the required number of other neurons, chosen randomly out of the remaining neurons with in-degree and out-degree less than that in the C. elegans network. This choice of random matrices is motivated by the observation that the distribution of in-degrees and out-degrees in C. elegans is significantly different from Poisson, which is expected for a randomly generated matrix without any correlations (Erdõs-Rényi random graph) (Figure 7).
N = 3. We keep the number of incoming and outgoing uni-directional connections as well as the number of reciprocal connections for each neuron the same. One of the implemented algorithms starts with an empty matrix. Then it randomly selects a neuron and does three things. It reconnects all outgoing connections of that neuron to other neurons, as long as their in-degree does not exceed that in the C. elegans network. It reconnects all incoming connections of that neuron to other neurons, as long as their out-degree does not exceed that in the C. elegans network. It reconnects all reciprocal connections of that neuron to other neurons with available unconnected reciprocal connections. We also implemented a second algorithm to verify the robustness of our results. This algorithm [9,10] will randomly pick and swap 2 unidirectional or 2 bi-directional connections (a→b and c→d will be changed to a→d and c→b).
N = 4. For comparing the count of quadruplets, we construct random matrices that keep the same not only in-degree and out-degree of uni-directional and bi-directional connections for each neuron but also the count of the 16 different 3-neuron pattern in the whole matrix. Starting from a random matrix for N = 3 as described above, we use the Simulated Annealing algorithm [24] by swapping two connections of the same type until the count for all triplets in the random matrix matches the real network. Since this swapping operation does not change the degrees of the various connection types for the neuron, the algorithm only has to check if the triplet count in all 16 classes is identical to C. elegans.
N = 5. For the analysis of the quintuplets, we modified the Simulated Annealing algorithm to match the count of all 4-neuron patterns to the real network. With this algorithm we could only find random matrices for which the relative difference between the count of each pattern in the random matrix and the real dataset was less than 10%.
Coin-tossing example of multiple hypothesis testing correction
Here we illustrate the issue of multiple hypothesis testing by considering a classical coin-tossing example. Imagine determining whether a given coin is fair (i.e. yielding heads with probability 1/2) or not by tossing it 100 times and recording the number of heads. If the number of heads is not too different from 50, we expect that the coin is fair. The significance of the deviation in the number of heads from 50 is characterized by the p-value, which is the probability that a fair coin would have that or greater deviation. For example, the probability of getting 62 or more heads is about 1% and the corresponding p-value = 0.01. Now consider testing simultaneously 100 different coins by tossing each 100 times. Analyzing these 100 experiments for outliers reveals that one coin yielded 62 heads. Does this mean that this specific coin is unfair? Not necessarily. Even if all the coins are fair, such a seemingly unlikely result will be observed approximately once when examining 100 coins. In other words, the p-value estimated for a single coin is an underestimation of the true p-value when 100 coins are examined simultaneously. This situation is called multiple hypotheses testing and requires a modification of the p-value definition.
p-Value calculation/multi hypotheses testing correction
Assume the number of N-neuron patterns in the i-th class in the actual network A and a random network R is given by: cN,i (A) and cN,i (R). Then the raw p-value is defined by:
pi = Pr(cN,i (Rk) ≥ cN,i (A), Rk ∈ {R}).
Because we look for over-representation of all connected patterns in parallel (and there are m = 13 patterns for N = 3, m = 199 patterns for N = 4 and m = 9364 patterns for N = 5), there is an increased probability of finding an over-represented pattern by chance. We correct for that by calculating a multiple hypothesis testing corrected p-value for each pattern and threshold. This p-value, Pm, reflects the probability that one random matrix Rk0 out of our random matrix ensemble {R} will have at least one pattern, i, which has smaller (or equal) raw p-value than the given pattern in C. elegans. This is known as the single-step min P procedure and controls for family-wide error rate [16,17]. In mathematical notation the single-step min P adjusted p-values are defined by:
where denotes the complete null hypothesis, pi the probability that the count for pattern i in a random matrix R is greater than the count in C. elegans, and Pj denotes the raw p-value for the ith pattern in a random matrix k0: Pj = Pr(cN,j (R) ≥ cN,j (Rk0)).
To determine Pm for a pattern i we perform the following procedure:
1. For all random matrices (k0 is the index of the random matrix; we usually created n = 1000 of them) out of the ensemble we calculate between and all other random matrices in this ensemble for each pattern i:
.
2. We then derive the raw p-value for as a minimum of these values across all patterns i: .
3. We calculate the probability that for a given pattern i the observed count in a random matrix Rk out of our ensemble {R} is greater than the count in the C. elegans network :
4. Last, we calculate the single-step min P adjusted p-value Pm for a given pattern i as:
In addition, we verified our results with the alternative single-step max T adjusted p-value [16,17] (for figures and explanations see Supplementary Information [Additional file 1]).
Datasets/data sources
We used data from [6], which provides separate connectivity data for the different reconstructions JSH and N2U done by White et al. (1986). We deleted 11 non-neuronal cell or classes from the dataset: CEPshDR, CEPshVL, CEPshVR, GLRDL, GLRDR, GLRL, GLRR, GLRVL, GLRVR, hyp, mu_bod. The classification of the neurons into their function and their location was taken from [20].
Supplementary Material
Additional File 1
A document containing supplementary information and data not presented in the paper. See also
Click here for file
Additional File 2
Description of the files containing triplet lists and used data sets. See also
Click here for file
Additional File 3
Zip file containing files mentioned in Additional file 2
Click here for file
Acknowledgements
We thank Armen Stepanyants and Sen Song for ideas and discussion of statistical issues, and Carlos Brody and Ingrid Ehrlich for comments on the manuscript. This work was initiated at the Aspen Center for Physics and was supported by the Lita Annenberg Hazen Foundation and the David and Lucile Packard Foundation at Cold Spring Harbor Laboratory and by the Minerva Foundation grant to UA.
Figures and Tables
Figure 1 Doublet counts in the C. elegans network compared to the random matrix ensemble. Bi-directionally connected doublets are over-represented in the C. elegans network. Counts shown are for dataset 1, threshold Θ = 3, number of random matrices n = 1000. Other datasets and thresholds give similar results.
Figure 2 Triplet counts in the C. elegans network compared to the random matrix ensemble. Blue squares show triplet counts for the actual network (dataset 1, threshold Θ = 2), red crosses show counts for each random connectivity matrix relative to the mean count for the whole random matrix ensemble. Three framed motifs are discussed in the main text. All matrices in this ensemble (n = 1000) preserve the counts of uni- and bi-directional connections for each neuron. A count for a given pattern is often the same in many matrices resulting in few crosses (e.g. there are only 25 crosses for pattern 12 because the count of this pattern in the random matrix ensemble varies between 6 and 32; 29 and 31 were not observed).
Figure 3 Significance measure Pm of triplet over-representation for different thresholds. The multiple hypothesis testing corrected p-values for triplet patterns show significant over-representation of patterns 10, 12 and 14 (datasets 1 and 2, n = 1000). The significance measure Pm represent the probability of finding a random matrix R, in which at least one class has smaller (or equal) raw p-value than that found for a given class in the actual network.
Figure 4 Significance measure Pm of triplet over-representation among interneurons shows that pattern #10 is significantly over-represented. Multiple hypothesis testing corrected p-values for the triplet over-representation in datasets 1 and 2 (n = 1000). The figure shows that motif #10 is significantly over-represented within interneurons.
Figure 5 Significance measure Pm of selected quadruplet over-representation for different thresholds. Multiple hypothesis testing corrected p-values for the quadruplet over-representation in datasets 1 and 2 (n = 1000). The patterns shown satisfy the following selection criteria: There must be a significant value Pm < 0.1 for at least one Θ and the count of this pattern in C. elegans must be at least 5. All random matrices in the ensemble (n = 1000) preserve the number of uni- and bi-directional connections for each neuron as well as the count in all triplet classes for the whole network.
Figure 6 Examples of symmetric quadruplet patterns that are not over-represented.
Figure 7 Distribution of degree in C. elegans in comparison to Poisson distribution. The distribution of in-degree and out-degree in a random matrix (Erdõs-Rényi random graph) can be approximated by Poisson. We observe that the distribution in C. elegans is significantly different from Poisson (p < 7%). Figure shows in-degree and out-degree for dataset 1, Θ = 1; Dataset 2 and other thresholds give similar results.
Table 1 Feedforward loops that do not fit into consideration of a 3-layer feedforward network.
Theta =
1 2 3 4 5
Dataset 1 49 % 39 % 33 % 38 % 34 %
2 47 % 40 % 41 % 39 % 29 %
The table shows the percentages of feedforward loops in which all three neurons belong to the same functional group or at least one of the three connections is made from an interneuron to a sensory neuron or from a motor neuron to an interneuron.
Table 2 Percentage of quadruplets in pattern #45, which do not fit into the three-layer feedforward network model.
motif i = 45 Theta =
1 2 3 4 5
Dataset 1 37 % 31 % 29 % 25 % 23 %
2 34 % 25 % 16 % 19 % 14 %
These quadruplets contain either all four neurons from the same layer or at least one connection from a motor neuron to an interneuron or from an interneuron to a sensory neuron.
Table 3 Number of connected neurons and the count of the different connection types after applying thresholds to the two datasets.
Source Dataset Threshold Θ = Connected neurons Total connections uni-directional connections Reciprocal connections
Durbin JSH 1 1 179 1152 1011 141
2 175 649 603 46
3 170 463 435 28
4 161 328 313 15
5 140 226 216 10
6 119 163 156 7
7 106 124 119 5
Durbin N2U 2 1 187 1288 1143 145
2 181 728 685 43
3 172 460 445 15
4 162 323 313 10
5 138 206 200 6
6 119 143 138 5
7 97 104 101 3
==== Refs
Ramón y Cajal S Textura del Sistema Nervioso del Hombre y de los Vertebrados Translation: Texture of the Nervous System of Man and the Vertebrates 1999 New-York: Springer 1899
Peters A Sethares C Myelinated axons and the pyramidal cell modules in monkey primary visual cortex J Comp Neurol 1996 365 232 255 8822167
Abeles M Gerstein GL Detecting spatiotemporal firing patterns among simultaneously recorded single neurons J Neurophysiol 1988 60 909 924 3171666
White JG Southgate E Thomson JN Brenner S The structure of the nervous system of the nematode Caenorhabditis elegans Philos Trans R Soc Lond B Biol Sci 1986 314 1 340
Hall DH Russell RL The posterior nervous system of the nematode Caenorhabditis elegans: serial reconstruction of identified neurons and complete pattern of synaptic interactions J Neurosci 1991 11 1 22 1986064
Durbin RM Studies on the Development and Organisation of the Nervous System of Caenorhabditis elegans 1987 Ph.D. Cambridge University
Shen-Orr SS Milo R Mangan S Alon U Network motifs in the transcriptional regulation network of Escherichia coli Nat Genet 2002 31 64 68 11967538 10.1038/ng881
Milo R Shen-Orr S Itzkovitz S Kashtan N Chklovskii D Alon U Network motifs: simple building blocks of complex networks Science 2002 298 824 827 12399590 10.1126/science.298.5594.824
Maslov S Sneppen K Specificity and stability in topology of protein networks Science 2002 296 910 913 11988575 10.1126/science.1065103
Newman ME Strogatz SH Watts DJ Random graphs with arbitrary degree distributions and their applications Phys Rev E Stat Nonlin Soft Matter Phys 2001 64 026118 11497662
Markram H Lübke J Frotscher M Roth A Sakmann B Physiology and anatomy of synaptic connections between thick tufted pyramidal neurones in the developing rat neocortex J Physiol 1997 500 409 440 9147328
Sjöström PJ Turrigiano GG Nelson SB Rate, timing, and cooperativity jointly determine cortical synaptic plasticity Neuron 2001 32 1149 1164 11754844 10.1016/S0896-6273(01)00542-6
Holmgren C Harkany T Svennenfors B Zilberter Y Pyramidal cell communication within local networks in layer 2/3 of rat neocortex J Physiol 2003 551 139 153 Epub 2003 Jun 2017. 12813147
White JG Southgate E Thomson JN Brenner S Factors that determine connectivity in the nervous system of C. elegans Cold Spring Harb Symp Quant Biol 1983 48 pt 2 633 640 6586380
Chklovskii DB Synaptic connectivity and neuronal morphology: Two sides of the same coin Neuron 2004 43 609 617 15339643
Dudoit S Shaffer J Boldrick J Multiple hypothesis testing in microarray experiments Statistical Science 2003 18 71 103 10.1214/ss/1056397487
Westfall PH Young SS Resampling-based multiple testing:examples and methods for p-value adjustment 1993 New York: Wiley
White JG Neuronal connectivity in Caenorhabditis elegans Trends Neurosci 1985 8 277 283 10.1016/0166-2236(85)90102-X
Gerstein GL Bedenbaugh P Aertsen MH Neuronal assemblies IEEE Trans Biomed Eng 1989 36 4 14 2646211 10.1109/10.16444
Achacoso TB Yamamoto WS Neuroanatomy of C elegans for computation 1992 Boca Raton, FL: CRC Press
Goodman MB Hall DH Avery L Lockery SR Active currents regulate sensitivity and dynamic range in C. elegans neurons Neuron 1998 20 763 772 9581767 10.1016/S0896-6273(00)81014-4
Kerr R Lev-Ram V Baird G Vincent P Tsien RY Schafer WR Optical imaging of calcium transients in neurons and pharyngeal muscle of C. elegans Neuron 2000 26 583 594 10896155 10.1016/S0896-6273(00)81196-4
Angstadt JD Donmoyer JE Stretton AO The number of morphological synapses between neurons does not predict the strength of their physiological synaptic interactions: a study of dendrites in the nematode Ascaris suum J Comp Neurol 2001 432 512 527 11268011 10.1002/cne.1116
Metropolis N Rosenbluth AW Rosenbluth MN Teller AH Equation of state calculations by fast computing machines J Chem Phys 1953 21 1087 1092
| 15574204 | PMC539283 | CC BY | 2021-01-04 16:02:56 | no | BMC Biol. 2004 Dec 2; 2:25 | utf-8 | BMC Biol | 2,004 | 10.1186/1741-7007-2-25 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-5-1841556656410.1186/1471-2105-5-184Research ArticleIntegrating partonomic hierarchies in anatomy ontologies Burger Albert [email protected] Duncan [email protected] Yiya [email protected] Richard [email protected] MRC Human Genetics Unit, Western General Hospital, Crewe Road, Edinburgh EH4 2XU, UK2004 26 11 2004 5 184 184 12 12 2003 26 11 2004 Copyright © 2004 Burger 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
Anatomy ontologies play an increasingly important role in developing integrated bioinformatics applications. One of the primary relationships between anatomical tissues represented in such ontologies is part-of. As there are a number of ways to divide up the anatomical structure of an organism, each may be represented by more than one valid partonomic (part-of) hierarchy. This raises the issue of how to represent and integrate multiple such hierarchies.
Results
In this paper we describe a solution that is based on our work on an anatomy ontology for mouse embryo development, which is part of the Edinburgh Mouse Atlas Project (EMAP). The paper describes the basic conceptual aspects of our approach and discusses strengths and limitations of the proposed solution. A prototype was implemented in Prolog for evaluation purposes.
Conclusion
With the proposed name set approach, rather than having to standardise hierarchies, it is sufficient to agree on a suitable set of basic tissue terms and their meaning in order to facilitate the integration of multiple partonomic hierarchies.
==== Body
Background
Introduction
As the bioinformatics emphasis has shifted from gene sequence analysis to functional genomics and proteomics, the need to describe gene function in the context of specific tissues of an organism has increased. Hence, in addition to anatomy ontologies built for medical purposes, e.g. GALEN [1], descriptions of anatomies are now often used to annotate a variety of genetic data, such as gene-expression. (A list of such ontologies for human as well as model organisms, e.g. mouse, Drosophila, zebrafish and C elegans, can be found on the Open Biological Ontologies web site [2].
An ontology model typically consists of concepts and relationships between these concepts. One of the key relationships in anatomy is part-of. It is possible to distinguish between different kinds of part-of, e.g. structural part-of and functional part-of. Each anatomy ontology may define one or more such part-of relationships.
Even for a single type of part-of, there may be more than one correct way to devide the anatomy of an organism into parts and subparts. Hence, multiple valid partonomic (part-of) hierarchies may exist for any organism. This raises the issue of interoperability across such hierarchies: when is a tissue in one hierarchy equivalent to a tissue in another hierarchy, and what are the part-of relationships across these hierarchies?
In general, biologists refer to tissues by their names, unlike computers which can easily work with ID numbers. For example the name "/embryo/limb/forelimb bud/ectoderm" is used to describe the ectoderm of the forelimb bud of the limb of the mouse embryo. Although this name uniquely identifies the tissue, it does so by encoding the particular partonomic hierarchy in the name. This causes problems when trying to work with more than one single hierarchy. This paper discusses a naming scheme that preserves the unique identification property of tissue names, without having to restrict it to a particular hierarchy, thus making it easier to integrate multiple partonomic hierarchies.
There is a large body of work discussing mereology (part-of relationships) in the biomedical literature. For example, Rogers and Rector [3] describe their experience of modelling part-of relationships in human anatomy as part of the GALEN project. Aspects of the Digital Anatomist Foundational Model (FMA) are given in [4]. Partonomies in a 3D model of human anatomy are briefly discussed in [5]. All of these papers distinguish between different kinds of part-of relationships. An example of an anatomy ontology using only one type of part-of can be found in the Edinburgh Mouse Atlas Project (EMAP). Although EMAP also uses derives-from relationships to capture cell lineage information with respect to embryo development, it is significantly less complex than GALEN and the FMA. Such variation of complexity is common and typically reflects the different purposes for which the ontologies were built. The EMAP ontology is used to label spatial data for the developing mouse embryo, specifically gene expression data [6].
We are not aware of any previous work dealing specifically with the integration of multiple part-of anatomy hierarchies. However, ontology alignment and integration in general is an active reserach area and has produced tools that aim at helping with the manual alignment of ontologies as well as with the automation of ontology integration. Examples of such tools include OntoMorph [7], OntoMerge [8] and the PROMPT tools suite [9]. Some work has been carried out in trying to use such tools to systematically merge GALEN and FMA, but the results have been rather limited [10,11]. In this paper we are not trying to argue for a general solution to the ontology integration problem, which as the evidence suggests is very hard to achieve. Instead we approach the problem from our specific application experience and seek to find a specific solution for a more limited domain.
The remainder of the paper is organised as follows. The next section introduces the Edinburgh Mouse Atlas, which forms the basis of the work presented here. Thereafter, the issue of multiple part-of hierarchies is discussed. The next section introduces the developed name set representation, followed by a discussion that covers the implementation of a Prolog prototype system. The proposed approach is then evaluated in the discussion section, followed by our conclusions.
Edinburgh Mouse Atlas
The Edinburgh Mouse Atlas (EMAP) and Gene Expression (EMAGE) Database project [12-16] has developed a digital atlas of mouse development which provides a bioinformatics framework to spatially reference biological data. The core databases contain 3D grey-level reconstructions of the mouse embryo at various stages of development, a systematic nomenclature of the embryo anatomy (the anatomy ontology), and defined 3D regions (domains) of the embryo models which map the anatomy onto the spatial models. Through the 3D domains users can navigate from the spatial representation of the embryo to the ontology and vice versa. Data from an in situ gene expression database is spatially mapped onto the atlas allowing the users to query gene expression patterns using the 3D embryo model and/or the ontology as a reference.
Following the description of mouse embryo development by Theiler [17], the anatomy ontology is organised into 26 developmental stages, referred to as Theiler stages (TS1-TS26). Each stage is primarily organised as a structural part-of tree, or partonomic hierarchy. Figure 1 shows the top 3 levels of the tree at TS6. (The browser shown in the figure is available on-line at the Mouse Atlas web site [12].)
The tissues represented by subnodes of a node in the tree are intended to be non-overlapping (exclusive) and complete, i.e. they describe all distinct parts of the parent tissue. For example, in Figure 1, the trophectoderm consists of the mural trophectoderm and the polar trophectoderm, which are distinct from each other and are the only parts of the trophectoderm at that stage. Although this holds for EMAP, it is not a requirement for the proposed approach. (In this paper, the term 'tissue' is used in a very generic way, meaning both: whole anatomical structures as well as specific tissues.)
Each tissue can be uniquely identified by its full name. A full name is an n-tuple: (t0, t1,...,tn). The path name of the tissue is (t0, t1,..., tn-1). The component name is tn. For example, given the tissue name (using a file directory style notation):
/embryo/branchial arch/3rd arch/branchial pouch/endoderm/dorsal
its full name is:
(embryo, branchial arch, 3rd arch, branchial pouch, endoderm, dorsal)
its path name is:
(embryo, branchial arch, 3rd arch, branchial pouch, endoderm)
and its component name is:
dorsal
Although the ontology covers all parts of the mouse embryo, there may not be a single node representing a particular tissue of interest. For example, there is no single node named (embryo, mesenchyme, trunk mesenchyme, paraxial mesenchyme, somite, sclerotome). However, there is a tissue named (embryo, mesenchyme, trunk mesenchyme, paraxial mesenchyme, somite), which has somite 05 to somite 20 as subparts (somite 05 to somite 20 are part of that tissue), and each of those has a subpart with component name sclerotome. The approach taken in EMAP is to introduce a new tissue node, called a group, with the appropriate subparts identified. Figure 2 shows the anatomy part-of graph for this example (at Theiler stage 14).
Although adding the notion of groups to EMAP is addressing the need for alternative arrangements of the part-of hierarchy, it does also raise a number of new questions. For example, it requires a suitable algorithm to determine appropriate tissues of which the new group should be part of. Also, some restrictions should be put in place to constrain what new links can be added; for example, if a new group contains all parts of some other tissue, then that tissue itself, rather than all of its parts, should be linked to the group. In other words, we require a mechanism that prevents biologists from adding too many part-of links unnecessarily. Let us assume that a new group needs to be introduced that contains leg as one of its parts. In this case the biologist should introduce a single part-of link between the new group and leg, and not multiple part-of links between the new group and hip, knee, lower leg and upper leg (which are the parts a leg consists of). The fact that these are parts of the group should be deduced from the transitivity property of the part-of relationship. These and other considerations seem representative of the more general problem of trying to integrate multiple part-of hierarchies over the same anatomical space. The remainder of this paper describes a possible solution to this problem.
Multiple part-of hierarchies
As previously mentioned, there is more than one way to structure the anatomical part-of hierarchy of an organism. The intersection of these hierarchies may occur at any level; they may share some or all of the their leaf nodes, but may also share intermediate nodes. A particular hierarchy may only deal with part of the organism, e.g. brain or heart, while others, such as EMAP, cover the entire organism.
The central example we use in this paper is that of somites. The somites are a repeating anatomical structure down the back of the animal. They give rise to the vertebrae, muscles of the backbone, skin and other structures. Each somite is divided into 3 parts: dermomyotome, myocoele and sclerotome. The dermomyotome is a group of cells which form the dermal layer of the skin and muscle tissue. The myocoele is a fluid-filled cavity of the somite, and the sclerotome gives rise to the bone of the vertebrae.
Most ontologies require each of their concepts to be uniquely identified by a name. In the context of an anatomical ontology, such as EMAP, it is clearly not enough to simply use the name sclerotome when wanting to refer to the sclerotome of somite 18. In general, the full name of the tissue is required, though in some cases a part of it may be sufficient, e.g. there is only one tissue at Theiler stage 14 that has component name somite 18.
Focusing on the somite part of the anatomy given in Figure 2, we can draw two possible hierarchies, as shown in Figure 3. (somite, somite 05, dermomyotome) in H1 and (somite, dermomyotome, somite 05) in H2 clearly semantically refer to the same mouse embryo tissue, in spite of using different names. Hence, for an anatomy ontology to embody its particular part-of hierarchy in the naming of its tissues is not helpful when it comes to integrating multiple hierachies. The proposal is therefore to avoid this problem by using name sets to identify a particular tissue.
Results
Name set representation of part-of hierarchies
Basic name sets
Each tissue in a part-of hierarchy is identified by the set of component names along the path from the root to the tissue (including the component name of the tissue itself). For example, in H1 the dermomyotome of somites 5 and 20 are represented as {dermomyotome, somite, somite 05} and {dermomyotome, somite, somite 20}, respectively; and in H2 somite 20's dermomyotome is represented as {dermomyotome, somite, somite 20}. Using NS(T) to denote the name set of tissue T, equivalence between two tissues is identified by the equivalence of their name sets:
NS(Ti) = NS(Tj) → Ti = Tj
Let Ti Tj denote that Ti has Tj as a direct subpart, and let Ti Tj denote that Ti has Tj as a subpart (direct or indirect)1, i.e. Ti Tj ... Tk implies Ti Tk, then the part-of relationships can be derived from the name sets as follows:
NS(Ti) ⊂ NS(Tj) → Ti Tj and
Ti Tj ∧ (¬ ∃k·Ti Tk ∧ Tk Tj) → Ti Tj
The first line simply states that Ti has Tj as a subpart, if the name set of the first is a proper subset of the name set of the second. The second line states that Ti has Tj as a direct subpart (or child tissue) if Ti has Tj as one of its subparts, and there are no other subparts of Ti which themselves have Tj as one of their own subparts. In the graph representing the ontology, an arrow is drawn from Ti to Tj if, and only if, Ti Tj.
The name set representation does not explicitly deal with temporal relationships. For example, changes in the anatomy of the developing embryo must be captured explicitly, i.e. if a particular subpart disappears from one developmental stage to the next, this should be reflected in the lack of that subpart in the ontological representation for the latter stage. Furthermore, the given representation does not explicitly distinguish between classes and instances of tissues. For example, while in general it holds that a leg has a lower leg part, this may not be true in specific instances. The proposed representation does not deal with such instance issues; many of the existing model organism anatomy ontologies used in bioinformatics today similarly do not represent information at the instance level.
Rest-of tissues
A "rest-of" tissue is a tissue that represents all parts of that tissue other than those which are explicitly represented in a "sibling" of the rest-of tissue. For example, the embryo mesenchyme marked as T1 in Figure 4 does not include the mesenchyme of the first branchial arch (labeled T3) or any of the other parts of the embryo (not shown in the figure).
Looking at the name set representation of T1 and T3 (in Figure 4), we see that NS(T1) ⊂ NS(T3). Based on the definition from above, T1 T3 follows. This, however, is not true. In other words, the basic name set representation introduced earlier is not sufficient to cope with rest-of tissues.
Positive and negative name sets
To deal with "exclusions" such as required for rest-of tissues, we introduce negative name sets (NSn) in addition to the name sets we already have (and we shall refere to as positive name sets (NSp) from now on). A tissue Tr includes in its negative name set the component name of any "sibling" tissue Ts, if Tshas a subpart with the same component name as Tr. For example, branchial arch is added to the negative name set of T1 because of T3 (from Figure 4).
Part-of relationships can now be derived from the name set representation of tissues as follows:
NSp(Ti) ⊂ NSp(Tj) ∧ NSn(Ti) ∩ NSp(Tj) = ∅ → Ti Tj and
Ti Tj ∧ (¬ ∃ k·Ti Tk ∧ Tk Tj) → Ti Tj
The first line states that Ti has Tj as a subpart, if the positive name set of Ti is a proper subset of the positive name set of Tj, and the intersection of the positive name set of Tj and the negative name set of Ti is empty. The intersection part has been added to enforce the exclusions needed to deal with rest-of cases. The second line's meaning is identical to what it was before.
Returning to the example in Figure 4, T1 is now represented as NSp(T1) = {embryo, mysenchyme} and and NSn(T1) = {branchial arch}, T3 is represented as NSp(T3) = {1st arch, branchial arch, embryo, mesenchyme} and NSn(T3) = {}. Since NSn(T1) ∩ NSp(T3) = {branchial arch}, i.e. non-empty, T3 is not a subpart of T1, as required.
For exclusions to work properly, negative name sets must be propagated to their subparts, as is implicitly the case for positive name sets already. Hence, T2 (in Figure 4) will also include branchial arch in its negative name set, keeping T3 from becoming one of its subparts.
Integration of multiple part-of hierarchies
Assuming that two or more part-of hierarchies are based on the same set of component names, integrating these hierarchies becomes a trivial task. Relationships (identity as well as part-of) between tissues from different hierarchies follow directly from the rules described above. For example, applying these rules to the hierarchies in Figure 3, the integrated part-of hierarchy of Figure 5 can automatically be generated.
Given the integrated name set representation of two or more hierarchies, it is not possible to determine which tissue belongs to which original hierarchy. For example, if asked for the immediate subparts of somite, based on the rules governing the part-of relationship, all of the tissues at the second level of the diagram in Figure 5 would be returned. To address this problem, extra information needs to be captured. This can easily be achieved by adding a view set to each tissue. For example, the view set for somite would be {H1, H2}, as it would be for all leaf node tissues in Figure 5. The intermediate tissue nodes have either {H1} (left part) or {H2} (right part) as their view sets. Thus, recreating one of the original hierarchies simply becomes a matter of filtering the integrated hierarchy using the view sets. In addition to the reconstruction of the original hierarchies, new views on the integrated hierarchy, or even on the original ones, can easily be created using appropriate name set "queries".
Prototype
A prototype of the name set representation for the Mouse Atlas anatomy ontology has been implemented in Prolog; an extension of the prototype we developed for our work on the Formalisation of Mouse Embryo Anatomy [6]. This original prototype included the following two predicates:
tissue(S, T, FN).
• S: stage ID, e.g. 14 for Theiler stage 14;
• T: tissue ID number (accession number);
• FN: full name of tissue represented by the list [N1, N2, N3, ...];
hasPart(TID1, TID2).
• TID2 is an immediate part of TID1, i.e. TID1 TID2;
For the evaluation of the name set representation, we use an extended version of the tissue predicate (view handling is ommitted from the protoype description to keep our examples simple) :
ext_tissue(S, T, FN, NSp, NSpL, NSn, NSnL).
• S, T, FN: as above;
• NSp: positive name set of tissue represented by a list [N1, N2, N3, ...];
• NSpL: length of NSp;
• NSn: negative name set of tissue represented by a list [N1, N2, N3, ...];
• NSnL: length of NSn;
For example, the embryo mesenchyme tissue of Figure 4 is represented as:
ext_tissue(14,705, ["embryo", "mesenchyme"],
["embryo", "mesenchyme"], 2,
["branchial arch", "limb", "organ system"], 3).
The following Prolog clause is used to determine whether Tp Tc is true:
subPart(Tp, Tc) :-
ext_tissue(Sp, Tp,_, NSpp, NSpLp, NSnp,_),
ext_tissue(Sp, Tc,_, NSpc, NSpLc,_,_),
NSpLc > NSpLp,
ord_subset(NSpp, NSpc),
ord_disjoint(NSpc, NSnp).
Predicates ord_subset and ord_disjoint from the Prolog library were used to implement the set theoretic aspects of the representation. Although these predicates support ordered sets, this is not required for our representation (but there were no unordered set predicates in the library). NSpLc > NSpLp is required to enforce proper subset relationships.
The following two Prolog clauses are used to determine whether Tp Tc is true:
not_immediate_subPart(Tp, Tc) :-
subPart(Tp, Tm),
subPart(Tm, Tc).
immediate_subPart(Tp, Tc) :-
subPart(Tp, Tc),
not not_immediate_subPart(Tp, Tc).
The Prolog implementation given is not particularly efficient and there are a number of optimisations that could be put in place. However, as the purpose of the prototype was not to deliver a robust application for end-users, but a reference implementation of the proposed approach for evaluation purposes, it proved entirely sufficient.
The paper makes no claims over the relative merits of different implementation strategies for the proposed approach. Alternatives to Prolog include using a relational database system or an ontology language, such as OWL (more details of OWL available from W3C [18]). The latter is of particular interest as it is gaining wide acceptance in the bioinformatics domain. At the time this work began, tools for developing ontologies using OWL were still in their early stages, and hence, we decided not to use them. In the meanwhile, however, Protege [19] and OilEd [20], have matured sufficiently and do provide appropriate alternative implementation platforms.
Discussion
For evaluation purposes, a number of tests were carried out on the name set representation of the Mouse Atlas anatomy. These are discussed here, together with some general observations about the proposed approach.
The first assumption that must hold is that no two tissues (at any given stage) have the same name set representation. This was tested using
test1 :-
ext_tissue(S, T1,_, NSp,_, NSn,_),
ext_tissue(S, T2,_, NSp,_, NSn,_),
T1 not = T2.
test1 returns no, i.e. no two different tissues with the same name sets were found, as required.
To test whether all part-of relationships can be reconstructed from the name set representation, we used
test2 :- immediate_subpart(T1, T2), not hasPart(T1, T2).
test3 :- hasPart(T1, T2), not immediate_subPart(T1, T2).
Both, test2 and test3 return no, i.e. the name set representation does not lead to any part-of relationships that are not intended (test2), and all existing part-of relationships are found through the name sets (test3), as required.
The smallest form of part-of hierarchy integration is the addition of a new tissue node, which is equivalent to adding a group in EMAP. A recently identified need for a group has been for all (embryo, mesenchyme, trunk mesenchyme, paraxial mesenchyme, somite, myocoele) tissues at Theiler stage 17. Using predicate
immediate_subPart_ns(S, NSp, NSn, T).
• S: stage ID;
• NSp: positive name set of new tissue node (group);
• NSn: negative name set of new tissue node (group);
• T: tissue ID of immediate sub-part of tissue identified by name sets and stage;
we can write a "query" in the form:
immediate_subPart_ns(17, ["embryo", "mesenchyme", "myocoele",
"paraxial mesenchyme", "somite", "trunk mesenchyme"],[], T), tissue(_, T, FN),
writeName(FN), nl, fail.
and obtain the following result:
("embryo", "mesenchyme", "trunk mesenchyme", "paraxial mesenchyme", "somite", "somite 05", "myocoele")
("embryo", "mesenchyme", "trunk mesenchyme", "paraxial mesenchyme", "somite", "somite 06", "myocoele")
...
("embryo", "mesenchyme", "trunk mesenchyme", "paraxial mesenchyme", "somite", "somite 30", "myocoele")
Similarly, using predicate immediate_superPart_ns(), we obtain:
("embryo", "mesenchyme", "trunk mesenchyme", "paraxial mesenchyme", "somite")
immediate_superPart_ns() is analogous, and its Prolog implementation very similar, to immediate_subPart_ns(). Details are, therefore, omitted.
The correctness of these results was confirmed by one of the biologists who created EMAP. Other, similar tests, worked equally well. A constraint put on all of these cases, however, is that the name set of the new group tissue must only contain names that are already used in the existing hierarchy.
This raises the question of how to deal with the introduction of new component names. For example, the addition of a group (embryo, head) cannot automatically be carried out, since the existing hierarchy does not use head in its name sets. For the integration to work, it is first necessary to add head to the appropriate name sets in the existing hierarchy. This can be done at the highest appropriate levels, since sub-parts inherit all name set elements from their super-parts, and may therefore not require as much effort as one initially expects.
For the head example, however, we did identify two additional problems which are likely to be typical in this context. Firstly, some agreement needs to be reached as to what in fact is considered to be part of the newly introduced tissue. In our example: how much of the neck is anatomically considered to be part of the head? The second problem deals with the fact that an existing tissue may need to be divided further in order to obtain the appropriate subparts for the newly introduced tissue. For example, the carotid artery runs from the head into the body of the mouse embryo, i.e. only a part of carotid artery is actually part of the head. Hence, the carotid artery needed to be divided into two subparts, one for the head section of it, one for the rest. In our name set approach, the former contains head in its positive name set, while the latter contains head in its negative name set. Of course, only the head section part becomes part of the head. Neither of these two problems presents any direct consequences for our approach.
When merging ontologies of different granularity, the same principle as before applies: shared component names must be used in a consistent manner. Assuming ontology O1 includes midbrain as one of the parts of the brain, but no further detail, and O2 is a brain anatomy ontology that divides the midbrain into cerebral aqueduct, floor plate, lateral wall, etc., then we would find {brain, central nervous system, embryo, midbrain, mouse, nervous system, organ system} as the positive name set for midbrain in O1, and {brain, cerebral aqueduct, midbrain} as the positive name set in O2, resulting in {brain, central nervous system, cerebral aqueduct, embryo, midbrain, mouse, nervous system, organ system} – the union of these previous two name sets – as the representation of midbrain in the merged ontology. The meaning of the component names in the intersection of the two original names sets, {brain, midbrain} must have been used in a consistent manner for the merger to work, though many of the component names will differ across the ontologies, because of the different levels of granularity, e.g. the terms nervous system and organ system are unlikely to be found in the brain specific ontology. (We omitted the negative name sets from this discussion, but the implications are essentially the same as for the positive name sets.)
Taking a closer look at these "basic tissue terms", called component names thus far, shows that some of them have additional structural complexity and if one wishes to take advantage of the semantics of these complexities, the proposed name set representation would need to be extended. For example, at Theiler stage 18 the tissue (embryo, branchial arch, 1st arch, mandibular component, mesenchyme) has two subparts, called (..., mesenchyme derived from head mesoderm) and (..., mesenchyme derived from neural crest). The naming, hence, reflects lineage relationships between tissues, and the identity of a tissue is partially established by that relationship. Although extensions to the name set representation could be developed to allow the inclusion and subsequent reasoning over such information, it would lead to a semantic overloading of the name sets and for simplicity are, therefore, not considered further – the (component) name is treated as an atomic string describing a tissue, while the lineage relationship is modelled externally to the name sets.
Theoretically, merging two part-of hierarchies can be accomplished by systematically (top-down) adding each tissue from one hierarchy into another, i.e. conceptually the problem can be reduced to iteratively adding "group nodes" as discussed above.
The approach discussed in this paper will not work where there has been no agreement on the basic component terms, and as such is different from already existing work on merging autonomous ontologies. This raises two questions: what is the basis on which these terms should be agreed and what benefits are to be obtained from the proposed solution if such agreement has to be reached before these partonomic hierarchies can be merged. With respect to the first question, if a basic term, for example skin, exists, then it must be possible to dissect the mouse to a level that separates all the corresponding tissue from the rest of the mouse tissues, e.g. separate all skin tissue from the rest of the mouse. Other examples of basic terms are, therefore, head, skeleton, limb and forelimb. At this point scientists are then free to use combinations of these terms (for the positive and negative name sets) to describe the anatomical concepts they are interested in, e.g. {head, skin} to refer to the skin of the head. The different anatomy hierarchies created by different scientists can then be automatically merged using the approach proposed in this paper. Hence, to answer the second question from above, the benefit of our solution lies in the removal of the need for multiple scientists to agree on a single anatomy partonomy where all tissue concepts are defined and their part-of relationships specified. Instead, a much more flexible solution is offered without having to sacrifice the interoperability across multiple data sets annotated with these anatomical concepts.
Essentially, the solution is based on the transitivity property of the structural part-of relationship. As such, one could imagine implementations other than the one based on name sets to achieve the same result. The basic idea, however, would be the same. Using the name set concept makes the solution more directly accessible to biologists, who are more familiar with naming anatomical concepts than using computer generated IDs. We believe that the same approach may be applicable in other ontology areas, which have similarly transitive relationships, but since we have not tested this idea, we shall not elaborate on it in this paper.
Also, the work described here only deals with the integration of hierarchies that are based on the same type of part-of relationships. Some preliminary studies suggest that where there are different types and these types are organised in an is-a hierarchy, the proposed integration mechanism will still work at the level of the common part-of type. For example, let H1 be a part-of hierarchy based on part-of-type-1, and let H2 be a part-of hierarchy based on part-of-type-2. If both, part-of-type-1 and part-of-type-2, are specialised versions of the more general part-of-type-0, i.e. part-of-type-1 is-a part-of-type-0 and part-of-type-2 is-a part-of-type-0, then we can use the proposed approach to integrate H1 and H2. The integrated hierarchy, however, would only support part-of-type-0 semantics. Our work in this area is still in its early phase and beyond the scope of this paper. Further details will be reported elsewhere.
The work presented in this paper has focused on the issue of integrating different partonomic hierarchies in one species, mouse. We note that a similar approach may be useful when trying to integrate partonomic hierarchies across different organisms. This is subject of current research work, however, and will be reported on separately.
Conclusions
Anatomy ontologies play an important role in bio-medical informatics. One of the key relationships modelled in such ontologies is that of part-of. For any given organism, however, there is more than one way to divide it into parts and subparts, thus leading to more than one valid partonomic hierarchy. To be able to interoperate between bioinformatics resources that make use of these anatomy ontologies, the corresponding hierarchies must be reconciled in some way. The paper addresses the problem that unique identifying names for tissues often reflect the partonomic hierarchies in which they are used. Although these names are in fact ordered sets (the order implying a particular hierarchy) of "component names", the order in these sets is not necessary to uniquely identify any tissue. Also, the sets of components in names can be used to derive all part-of relationships in the hierarchy. Based on these observations, we have developed a name set representation which facilitates integration of different partonomic hierarchies. Although this does not eliminate the requirement to agree on a set of suitable basic tissue terms and their meaning, it does remove the need to standardise the partonomic hierarchies. The proposed approach has been tested for the anatomy ontology of the Edinburgh Mouse Atlas. A Prolog prototype was implemented for evaluation purposes.
Note
1Tj is a direct subpart of Ti, if Tj is part of Ti and there is no other tissue Tk such that Tj is part of Tk and Tk is part of Ti. If such a tissue Tk exists, Tj is an indirect subpart of Ti.
Authors' contributions
AB developed the name set representation, implemented the prototype and carried out parts of the evaluation. YY provided input with respect to the current implementation of EMAP and EMAGE in relation to the proposed name set representation. DD carried out part of the evaluation process. RB contributed to the development of the name set representation. DD and RB are overall project leaders of EMAP and EMAGE. All authors have contributed to the writing and/or revision of the paper.
Figures and Tables
Figure 1 Anatomy browser Screenshot of Mouse Atlas anatomy browser showing the top 3 levels of mouse embryo anatomy at developmental stage TS6.
Figure 2 Extract of part-of hierarchy in EMAP at Theiler Stage 14 Diagram illustrating the need for so-called "group" nodes, sclerotome in this example, in the EMAP anatomy part-of hierarchy.
Figure 3 Alternative hierarchies for somite Two possible part-of hierarchies for the somite part of the ontology, and how they relate.
Figure 4 Name set representation and part-of hierarchy The name set representations for selected tissues are presented in the context of the part-of hierarchy.
Figure 5 Merged ontology for somite The part-of hierarchy that results from merging two possible hierarchies for somites.
==== Refs
OpenGALEN
OBO open biological ontologies
Rogers J Rector A GALEN's Model of Parts and Wholes: Experiences and Comparisons In Proceedings of Annual Fall Symposium of American Medical Informatics Association 2000 714 718
Noy F Musen M Mejino J Rosse C Pushing the Envelope: Challenges in a Frame-Based Representation of Human Anatomy Tech Rep SMI-2002-0925pdf, Stanford Medical Informatics 2002
Schubert R Hoehne K Partonomies for interactive explorable 3D-Models of Anatomy In Proceedings of the 1998 AMIA Annual Symposium 1998
Burger A Davidson D Baldock R Formalization of mouse embryo anatomy Bioinformatics 2004 20 259 264 14734318 10.1093/bioinformatics/btg400
Chalupsky H OntoMorph: A translation system for symbolic knowledge In Principles of Knowledge Representation and Reasoning: Proc 7th Int Conf on Data Engineering (ICDE'04) 2004
Dou D McDermott D Qi P Ontology translation by ontology merging and automated reasoning In Proc 13th International Conference on Knowledge Engineering and Knowledge Management (EKAW2002), Workshop on Ontologies 2002
Noy NF Musen MA The PROMPT suite: interactive tools for ontology merging and mapping Int J Hum Comput Stud 2003 59 983 1024 10.1016/j.ijhcs.2003.08.002
Mork P Bernstein P Adapting a Generic Match Algorithm to Align Ontologies of Human Anatomy In Proc 20th Int Conf on Data Engineering (ICDE'04) 2004
Zhang S Mork P Bodenreider O Lessons learned from aligning two representations of anatomy Proc First International Workshop on Formal Biomedical Representation 2004
The Edinburgh Mouse Atlas project
Brune R Bard J Dubreuil C Guest E Hill W Kaufman M Stark M Davidson D Baldock R A Three-Dimensional Model of the Mouse at Embryonic Day 9 Dev Biol 1999 216 457 468 10642785 10.1006/dbio.1999.9500
Davidson D Baldock R Bioinformatics Beyond Sequence: Mapping Gene Function in the Embryo Nat Rev Genet 2001 2 409 418 11389457 10.1038/35076500
Davidson D Bard J Brune R Burger A Dubreuil C Hill W Kaufman M Quinn J Stark M Baldock R The mouse atlas and graphical gene-expression database Semin Cell Dev Biol 1997 8 509 517 9441956 10.1006/scdb.1997.0174
Ringwald M Baldock R Bard J Kaufman M Eppig J Richardson J Nadeau J Davidson D A database for mouse development Science 1994 265 2033 2034 8091224
Theiler K The House Mouse: Atlas of Embryonic Development 1989 Springer Verlag
W3C World Wide Web consortium
Protege
OilEd
| 15566564 | PMC539284 | CC BY | 2021-01-04 16:02:46 | no | BMC Bioinformatics. 2004 Nov 26; 5:184 | utf-8 | BMC Bioinformatics | 2,004 | 10.1186/1471-2105-5-184 | oa_comm |
==== Front
Malar JMalaria Journal1475-2875BioMed Central London 1475-2875-3-491558832510.1186/1475-2875-3-49ResearchAnti-Plasmodium activity of ceramide analogs Labaied Mehdi [email protected] Arie [email protected] Marc [email protected]èze Marc [email protected]ée Stéphane [email protected] Serge L [email protected] Chunbo [email protected] Shimon [email protected] Philippe [email protected] USM0504 Biologie fonctionnelle des protozoaires, Département Régulations, Développement, Diversité Moléculaire, Muséum National d'Histoire Naturelle, Boite postale n°52, 61 rue Buffon, 75231 Paris Cedex 05, France2 Department of Biochemistry, Hebrew University-Hadassah School of Medicine, P.O. Box 12272, Jerusalem, 91120, Israel3 CNRS FRE 2775, Station biologique de Roscoff, 29682 Roscoff, France2004 10 12 2004 3 49 49 30 10 2004 10 12 2004 Copyright © 2004 Labaied et al; licensee BioMed Central Ltd.2004Labaied 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
Sphingolipids are key molecules regulating many essential functions in eukaryotic cells and ceramide plays a central role in sphingolipid metabolism. A sphingolipid metabolism occurs in the intraerythrocytic stages of Plasmodium falciparum and is associated with essential biological processes. It constitutes an attractive and potential target for the development of new antimalarial drugs.
Methods
The anti-Plasmodium activity of a series of ceramide analogs containing different linkages (amide, methylene or thiourea linkages) between the fatty acid part of ceramide and the sphingoid core was investigated in culture and compared to the sphingolipid analog PPMP (d,1-threo-1-phenyl-2-palmitoylamino-3-morpholino-1-propanol). This analog is known to inhibit the parasite sphingomyelin synthase activity and block parasite development by preventing the formation of the tubovesicular network that extends from the parasitophorous vacuole to the red cell membrane and delivers essential extracellular nutrients to the parasite.
Results
Analogs containing methylene linkage showed a considerably higher anti-Plasmodium activity (IC50 in the low nanomolar range) than PPMP and their counterparts with a natural amide linkage (IC50 in the micromolar range). The methylene analogs blocked irreversibly P. falciparum development leading to parasite eradication in contrast to PPMP whose effect is cytostatic. A high sensitivity of action towards the parasite was observed when compared to their effect on the human MRC-5 cell growth. The toxicity towards parasites did not correlate with the inhibition by methylene analogs of the parasite sphingomyelin synthase activity and the tubovesicular network formation, indicating that this enzyme is not their primary target.
Conclusions
It has been shown that ceramide analogs were potent inhibitors of P. falciparum growth in culture. Interestingly, the nature of the linkage between the fatty acid part and the sphingoid core considerably influences the antiplasmodial activity and the selectivity of analogs when compared to their cytotoxicity on mammalian cells. By comparison with their inhibitory effect on cancer cell growth, the ceramide analogs might inhibit P. falciparum growth through modulation of the endogenous ceramide level.
==== Body
Background
Sphingolipids are essential components of eukaryotic cell membranes, predominantly found in the outer leaflet. Sphingosine and ceramide (Figure 1) are the two simplest molecules structurally, which belong to the sphingolipid family. Sphingosine represents the sphingoid backbone, and ceramide has a fatty acid linked in a amide bond to sphingosine. Sphingolipid species have two types of functional groups linked to the 1-position, i.e. sphingomyelin (SPM) (Figure 1) having a phosphorylcholine group, and a variety of glycolipids having either glucose, galactose, galactosyl-sulfate or oligo-glycosides linked to the sphingosine moiety of ceramide.
Figure 1 structures of sphingolipids and analogs
Until recently, sphingolipids were primarily considered to be structural components of membranes. However, data accumulated during the last decade have expanded the view of their biological functions. They are now also considered to be key molecules which regulate many functions essential to eukaryotic cells [1-5]. They are involved, for example, in the regulation of membrane fluidity and are part of discrete membrane microdomains or rafts implicated in signalling and trafficking in cells [4,6-8]. Interest in sphingolipids was strengthened by an increasing body of evidence demonstrating their role as secondary messengers for intracellular signal transduction pathways that regulate many cellular processes. For example, ceramide accumulates in response to several different inducers such as cytokines, cytotoxic agents or to stressful conditions, which lead to cell cycle arrest or to apoptosis [9]. Sphingosine is a protein kinase C inhibitor [10] that inhibits growth or stimulates proliferation, depending upon the cell type [11,12].
Ceramide plays a central role in sphingolipid metabolism [13]. It can be converted into SPM through transfer of the choline phosphate group from phosphatidylcholine or serves as a precursor for complex sphingolipids (cerebrosides which possess sugar residues and gangliosides which contain sialic acid residues in addition to the carbohydrate units).
Moreover, ceramide can be phosphorylated by a distinct kinase and can also be produced by enzymatic hydrolysis of complex sphingolipids. In turn, ceramide can be hydrolyzed to sphingosine and fatty acid by ceramidases.
In contrast to yeast and mammalian cells, the current understanding of sphingolipid metabolism and the biological role of sphingolipids in the development of Plasmodium falciparum, the causative agent of malaria, is still limited. Gerold et al. [14] provided evidence that de-novo synthesis of sphingolipids occurs in the intraerythrocytic stages of the human malaria parasite P. falciparum and can be inhibited by the well established inhibitors of de-novo ceramide biosynthesis, fumonisin B1, cyclo-serine and myriocin [15,16]. However, these compounds are weak inhibitors of parasite growth. Evidence was provided that another pathway for the synthesis of glycosylated sphingolipids exists in P. falciparum [14,17]. The importance of sphingolipid metabolism for parasite development was demonstrated by Haldar's work showing that: (i) The parasite contains two distinct forms of SPM synthase, one sensitive to sphingolipid analogs, d,1-threo-1-phenyl-2-decanoylamino-3-morpholino-1-propanol (PDMP) or d,1-threo-1-phenyl-2-palmitoylamino-3-morpholino-1-propanol (PPMP) (Figure 1), known to inhibit the synthesis of glucosylceramide in mammalian cells [18], and the second insensitive to them [19]; (ii) These analogs blocked the parasite proliferation in culture by preventing the formation of the tubovesicular network (TVN) that extends from the parasitophorous vacuole to the red cell membrane and delivers essential extracellular nutrients to the parasite [20-22]. Neutral magnesium-dependent sphingomyelinase activity was also identified in P. falciparum [23-25], indicating that a sphingomyelin cycle (ceramide-SPM conversion) exists in Plasmodium. Recently, an increase in the intracellular ceramide content and an activation of parasite sphingomyelinase(s) were found to be associated with the parasite death process as induced by artemisinine and mefloquine [26].
Given the importance of sphingolipids in many cellular functions and the central role of ceramide in sphingolipid metabolism, the anti-Plasmodium activity of non-natural analogs of ceramides was investigated on the intraerythrocytic development of P. falciparum. Interestingly, a series of analogs containing a methylene (CH2-NH) linkage between the fatty acid and the sphingoid-analog core showed considerably higher anti-Plasmodium activity than their counterparts with a natural amide (CO-NH) linkage or than PPMP. The methylene analogs irreversibly blocked parasite development in contrast to PPMP whose effect is cytostatic. Their efficiency in inhibiting parasite growth did not correlate with their potential to inhibit parasite SPM synthase activity, indicating that SPM synthase is not their primary target. Possible mechanisms of action are discussed.
Methods
Materials
D,1-threo-1-phenyl-2-palmitoylamino-3-morpholino-1-propanol-HCl (D,1-threo-PPMP) was purchased from Matreya (Pleasant Gap, PA). 6-((N- (7-nitrobenz-2-oxa-1, 3-diazol-4-yl) amino) hexanoyl sphingosine (NBD-C6-ceramide) and N- (4,4-difluoro-5, 7-dimethyl-4-bora-3a, 4a-diaza-s-indacene-3-pentanoyl) sphingosyl phosphocholine (BODIPY-FL-C5-ceramide) were obtained from Molecular Probes, Inc. (Eugene, OR). The compounds of Figure 3 and Figure 4 were synthesized according to the procedure described by Dagan et al [27], using specific starting materials for each analog. The compounds of Figure 2 were synthesized by linking specific fatty acids to the amino group of substituted 1,3-dihydroxy-2-aminophenyl derivatives. The full description of the synthesis of each specific analog will be described in a separate publication.
Figure 2 Anti-P. falciparum activity of ceramide analogs having an amide linkage (series A).
Figure 3 Anti-P. falciparum activity of ceramide analogs having a methylene or a thiourea linkage (series B).
Figure 4 Anti-P. falciparum activity of selected derivatives
P. falciparum culture and synchronization
P. falciparum strains (FcB1/Colombia, K1/Thailand, F32/Tanzania, W2/Indochina) were maintained in continuous culture on human erythrocytes in RPMI medium containing 7% (v/v) heat-inactivated human serum under an atmosphere of 3% CO2, 6% O2, 91% N2, at 37°C, as described by Trager and Jensen [28]. Parasite synchronization was performed successively by treatment with 5% (w/v) sorbitol and by concentration in gelatin solution as previously described [29].
Anti-Plasmodium activity
Drug susceptibility assays were performed using a modification of the semi automated microdilution technique of Desjardins et al. [30]. Stock solutions of test compounds were prepared in DMSO. Drug solutions were serially diluted twofold with 100 μl culture medium in 96-well plates. Asynchronous parasite cultures (100 μl, 1 % parasitemia and 1 % final hematocrite) were added to each well and incubated for 24 hours at 37°C prior to the addition of 0.5 μCi of [3H] hypoxanthine (Amersham, France, 1 to 5 Ci.mmol/ml) per well. After a further incubation of 24 hour, plates were frozen and thawed. Cell lysates were then collected onto glass-filter papers and counted in a liquid scintillation spectrometer. The growth inhibition for each drug concentration was determined by comparison of the radioactivity incorporated in the treated culture with that in the control culture (having the same final % of DMSO) maintained on the same plate. The concentration causing 50% growth inhibition (IC50) and 90% growth (IC90) were obtained from the drug concentration-response curve and the results were expressed as the means ± the standard deviations determined from several independent experiments. The DMSO concentration never exceeded 0.1% (v/v) and did not inhibit the parasite growth.
Cytotoxicity test upon human embryonic cells
A human diploid embryonic lung cell line (MRC-5, Bio-Whittaker 72211D) was used to assess the cytotoxic effects towards eukaryotic host cells. MRC-5 cells were seeded at 5,000 cells per well in 100 μl. After 24 hours, the cells were washed and two-fold dilutions of the drug were added in 200 μl standard culture medium (RPMI medium + 5% fetal calf serum) and maintained for five days under 5% CO2 atmosphere. The final DMSO concentration in the culture remained below 0.1%. Untreated cultures were included as controls. The cytotoxicity was determined using the colorimetric MTT assay according to the manufacturer's recommendations (Cell proliferation kit I, Roche Applied Science, France) and scored as a percentage of reduction in absorption at 540 nm of treated cultures versus untreated control cultures. IC50 values were obtained from the drug concentration-response curve. The results were expressed as the mean ± the standard deviations determined from several independent experiments. The index of selectivity was defined as the ratio of the IC50 value on MRC-5 to that of P. falciparum.
Parasite stage-specific inhibitory effects and reversibility
Synchronized cultures (1–2% parasitemia) at the ring stage (0–10 hours old parasites), the trophozoite stage (25–35 hours old parasites) and the schizonte stage (40–48 hours old parasites) were maintained in the presence of drug concentrations in the vicinity of IC50 values. Aliquots were removed at the indicated times, washed three times with culture medium and maintained in culture in the absence or in the presence of a given drug. Parasite morphology was determined on Giemsa-stained smears defined according to the following criteria: the ring stage, when parasites exhibited a peripheral cytoplasm stained by Giemsa and a unstained intraparasitic vacuole; the trophozoite stage, when parasites showed a fully stained cytoplasm, haemozoin crystals and one nucleus; the schizont stage, when parasites presented several distinctive nuclei. Parasitaemias were determined by counting 3,000 cells for each sample. Controls consisted of parasites incubated with DMSO instead of drugs processed in the same way.
Sphingomyelin synthase activity assays
SPM synthase activity was measured as described by Haldar et al. [31]. Briefly, assays were performed on P. falciparum cultures at the trophozoite stage (20–30 h old parasites). 400 μl of culture (1 × 108 parasites) were incubated for 60 min at 37°C with 10 μM NBD-C6-ceramide and 0 to 500 μM PPMP or AD2646. Cells were then lysed by freezing and thawing of the culture. Lipids were extracted by a modification of the method of Bligh and Dyer [32]. To each sample, three volumes of a CH3OH/CHCl3 mixture (1:2, v:v) were added and the mixture vortexed for one min. Organic and aqueous phases were separated by centrifugation (12,000 × g, five min) and the organic phase was dried. Lipids were dissolved in 15 μl ethanol and analysed by thin layer chromatography on HPTLC plates (Silica gel 60 F254, Merck, Darmstadt, Germany) in CH3OH/CH3Cl3/NH4OH (75:25:4, v:v:v). For qualitative analyses, the fluorescent lipids were detected under UV and for quantitative analyses, the fluorescent lipid spots were scraped, eluted in one ml methanol and quantified at an excitation of 470 nm and an emission of 530 nm in a spectrofluorometer. The percentage of SPM synthase activity for each drug concentration was determined by comparison of the fluorescence quantified in the analog-treated culture with that in the control culture (without drug).
Labelling of infected red blood cells and fluorescence microscopy
Infected erythrocytes treated with or without ceramide analogs were incubated for 30 min, at 37°C, in culture medium containing 10 μM BODIPY-FL-C5-ceramide, washed three times with culture medium without serum and fixed overnight, at 4°C, in 3.7% formaldehyde/0.05% glutaraldehyde. Cells were mounted on poly-L-lysine coated slides and viewed using a Nikon Eclipse TE 300 DV inverted microscope with an 100X oil objective mounted on a piezzo electric device using appropriate fluorescence emission filters. Image acquisition (z-series) was performed with a back illuminated cooled detector (CCD EEV: NTE/CCD-1024-EB, Roper Scientific, France) using a 0.2 μm step. Data acquisition and image deconvolution process were performed with Metamorph software (Universal Imaging Corporation, Roper Scientific, France). The images presented correspond to the maximum intensity projection of the deconvoluted z-series.
Results and Discussion
Anti-Plasmodium activity of non-natural ceramide analogs
Non-natural analogs of ceramides were synthesized comprising two functional groups [27] : 1) A phenyl group substituted on carbon 3 of a sphingoid-like backbone; with the phenyl group replacing the sphingosine acyl chain [33,34] to which were linked nitro or amine groups, or carbon chains of varying lengths; and 2) a fatty acid with an amide (CO-NH) linkage (series A, Figure 2), a methylene (CH2-NH) or a thiourea (CS-NH) linkages (series B, Figure 3) on carbon 2. Analogs in which the alkyl group replaces the amide were investigated because the carbonyl group of ceramide was shown not to be necessary for triggering apoptosis in mammalian cells. In fact, replacement of the carbonyl group of ceramide by a methylene group substantially reduced the time required for cell death [35]. Only D/L-threo enantiomers were investigated on P. falciparum since reports demonstrated that D/L-erythro enantiomers of ceramide analogs were less efficient in inhibiting glucosylceramide synthase in mammalian cells [18] and did not inhibit SPM synthase activity in P. falciparum [19].
Figure 2 and Figure 3 show the IC50 values obtained for the different compounds on the development of the chloroquine-resistant strain FcB1 of P. falciparum in culture (IC50 value for chloroquine = 115 ± 25 nM, n = 3). Interestingly, the nature of the linkage considerably influences the anti-Plasmodium activity. Analogs with amide linkage were found to inhibit parasite growth with IC50 values in the micromolar range (Figure 2). Best IC50 values were similar to that obtained with the ceramide-related compound PPMP (IC50 = 9.0 ± 1.7 μM, n = 3). However, this IC50 value for PPMP differed from the previously reported value (IC50 = 0.85 μM) [19]. The discrepancy may be due to drug susceptibility assay conditions which were performed on synchronized cultures at the ring stage for Lauer et al. [19] and on asynchronous cultures in the present study. Analogs with methylene linkages were more efficient than the amide analogs in killing parasites with IC50 values in the nanomolar range (Figure 3).
For the D-threo nitro phenyl analogs of series A, no particular increase of the inhibitory activity was observed with the increase of the N-acyl chain length (IC50 values ranging from 10.8 to 40.4 μM, Figure 2). For the series B, best activities were observed for N-alkyl chain length of 12–16 carbons (IC50 values ranging from 17 to 42 nM for the series B, Figure 3). In both series, substitution of the nitrophenyl group by an aminophenyl group instead of nitro group decreased the anti-Plasmodium activity significantly (compare compounds AD2495 and AD2623 of series A, Figure 2; and compounds AD2646 and AD2672 of series B, Figure 3).
Increase of the analog hydrophobicity by substitution of the nitro group of the phenyl ring by alkyl chains seems to decrease the anti-Plasmodium activity of compounds of both series (compare compounds AD2583 and AD2603-7, Figure 2 and compounds AD2646 and AD2677-78-80, Figure 3). Surprisingly, in the B series, the anti-Plasmodium activity was restored in compounds with symmetrical alkyl chains of 6–8 carbon length (compounds AD2651 and AD2670, Figure 3). No systematic difference in anti-Plasmodium activity was observed between D-threo and L-threo enantiomer of a same analogue: e.g. the enantiomers AD2646 and AD2645 of the B series showed similar activity (Figure 3). It can also be noted that ceramide analogs containing a thiourea linkage also showed a significant anti-Plasmodium activity (Figure 3, compounds AD2215-17) with, however, a less pronounced inhibitory effect than analogs with a methylene linkage.
Inhibition of parasite growth by the methylene analog AD2646 was observed having similar IC50 values on the P. falciparum strains K1 (IC50 = 45 nM), F32 (IC50 = 21 nM) and W2 (IC50 = 28 nM), suggesting that the drug is not restricted to a specific strain and acts through a conserved mechanism in malarial parasites. Furthermore, analysis of drug combination with antimalarial drugs showed that AD2646 has a non-synergistic and non-antagonistic effect with CQ on the CQ-resistant strain K1, and with mefloquine and with artemether on the FcB1 strain (data not shown). Compound AD2646 (Figure 1) was selected to further investigate the biological effects of methylene analogs on parasite development.
Structure-activity relationship around AD2646 showed that the presence of a nitro group linked to the phenyl is not essential for anti-Plasmodium activity (Figure 4, compare IC50 values of compounds AD2646 and AD2730) nor hydroxylation on carbon 1 (compare compounds AD2730 and AD2724). In contrast, hydroxylation of carbon 3 is important for anti-Plasmodium activity since removal of the hydroxyl group reduced the activity 13.5 times (compare compounds AD2730 and AD2729).
Cytotoxicity on human cells MRC-5 of ceramide analogs in methylene linkage
The cytotoxicity of methylene analogs upon human MRC-5 cells (diploid embryonic lung cell line) was evaluated (Table 1). Derivatives tested showed IC50 values in the micromolar range, from 5 to 8 μM (except for AD2619), which are similar to the IC50 value of PPMP. No major difference of toxicity was observed between D- and L-threo enantiomers (compare AD2646 and AD2645). In contrast to what was observed for P. falciparum, hydroxylation of the sphingosine carbon 3 does not seem important for cytotoxicity since similar IC50 values were measured for AD2646 and AD2729, suggesting different mechanism(s) of action for AD2646 on MRC-5 cells and P. falciparum. AD2646 and 4 derivatives show high selectivity for P. falciparum as illustrated by the high index of selectivity of these compounds ranging from 160 to 624. The index of selectivity was defined as the ratio of the IC50 value on MRC-5 cells to that on P. falciparum. It can be noted that no selectivity was observed for PPMP. A similar range of growth inhibition was measured on P. falciparum (Figure 2) and HL-60 cells [36] with ceramide analogs in amide linkage supporting a weak selectivity of these analogs for P. falciparum.
Table 1 Cytotoxicity of methylene analogs and PPMP on human MRC-5 cells
Compounds IC50 (μM) IC90 (μM) Index of selectivity
AD2646 (-) 4.9 7.5 160
AD2645 (+) 6.1 10.3 161
AD2672 (-) 3.7 5.9 2
AD2730 (-) 6.1 9.9 322
AD2729 (-) 5.8 9.8 22
AD2619 (-) 26.7 42.3 624
PPMP 7.5 12.4 0.8
IC50 and IC90 values are the mean of three independent experiments. The S.E. were within 10% of the mean. (-): D-threo, (+): L-threo. Index of selectivity is defined by the ratio of the IC50 value on MRC-5 cells to that on P. falciparum.
It must be emphasized that the amide linkage of ceramide analogs is not required for activating apoptosis in cancer cells [35]. An increase of cytotoxicity of ceramide analogs in methylene linkage compared to their counterparts in amide linkage was also observed on the human histolytic lymphoma U937 [35] and the human leukaemia HL-60 cells [27] however, with higher IC50 values than that observed with P. falciparum.
Stage-specific inhibitory effects of AD2646 and reversibility
To investigate the cytostatic or cytotoxic effects of AD2646 on the parasite development, cultures at the ring stage (0–10 hours), the trophozoite stage (25–35 hours) and the schizonte stage (40–48 hours) were incubated with 30, 100 or 250 nM of AD2646 for 24.5 hours for the ring stage, for 11 hours for the trophozoite stage, and for 14 hours for the schizonte stage. Aliquots were then taken, washed and incubated in the absence or the presence of drug for a further 13 hours to 24 hours depending upon the parasite stage tested (see Figure 5). Parasitaemia and parasite stages were determined on Giemsa-stained smears at time of aliquot removal and after the subsequent incubation.
Figure 5 P. falciparum stage sensitivity to AD2646. Parasites at the ring (A), trophozoite (B) and schizonte (C) stages were maintained in the presence of 30 nM (square) or 100 nM (triangle) AD2646 for 24 h30, 11 h and 14 h, respectively. Aliquots were then taken, washed and maintained in culture in the absence (open symbol) or in the presence (full symbol) of the same concentration of analog. Controls were cultures maintained in the absence of drug (full circle) and processed as the treated cultures. Parasitemia and parasite morphology were determined on Giemsa-stained smears at the indicated time. Each value is the mean of two independent experiments.
Development of the ring stage was slightly affected by a continuous incubation with 30 nM AD2646. In contrast, when incubated with 100 and 250 nM, parasite growth was irreversibly blocked at the young trophozoite stage and the parasite degenerated. Drug removal after 24 hours of incubation did not allow a recovery of parasite growth (Figure 5A). The trophozoite stages were more sensitive to AD2646 since a continuous incubation with 30 nM completely blocked development. Parasites did not enter into division and then degenerated. Only a partial recovery of parasite growth was observed when drug was removed after 11 hours of incubation. A more marked effect was observed with 100 nM AD2646 with degenerated parasites already observed after only 11 hours. No recovery of parasite growth was then observed after drug removal (Figure 5B). The schizont stage appeared less sensitive than the trophozoite stage since a slight effect was only observed on the parasite development with 30 nM AD2646. However, parasite growth was irreversibly blocked by an incubation with 100 nM AD2646 and parasites degenerated (Figure 5C). Similar results were observed for the methylene analogs AD2651 and AD2670, the trophozoite stage being the most sensitive with a complete inhibition of parasite development for 250 nM (data not shown).
It can be noted that, in contrast to methylene analogs, addition of PPMP to parasite culture led to a preferential and reversible arrest of parasite development at the ring stage. The schizont stage (>30 hours old parasites) was insensitive to this concentration of drugs [14,19]. A cytostatic effect of PPMP on the ring-stage was effectively observed : rings blocked by a 24 hours incubation with 5 μM PPMP recovered to a normal growth after drug removal (data not shown). Blockage of parasite development was associated with the inhibition of a sensitive SPM synthase and TVN formation that delivers extracellular nutrients to the parasite [20-22].
Inhibition of sphingomyelin synthase activity and tubovesicular network formation of P. falciparum by compound AD2646
Figure 6 reproduces the inhibitory effects of PPMP and the methylene analogue AD2646 on the SM synthesis activity of young trophozoite (20–30 hours)-infected erythrocytes maintained in culture. As previously reported [19], no SPM synthase activity was measured in non-infected red blood cells and a biphasic inhibition curve was observed with PPMP in infected erythrocytes. Two pools of SPM synthase activity are present in parasites with respect to their inhibition by the ceramide analogue, one very sensitive to the drug and the second only inhibited by high concentrations of drug. The biphasic inhibition curve that superimposes on the PPMP inhibition curve was also recorded for AD2646 indicating that PPMP and AD2646 inhibit the SPM synthase activity of infected-red blood cells in a similar way.
Figure 6 Inhibition of P. falciparum sphingomyelin synthase activity by AD2646 and PPMP. Trophozoite cultures (20–30 hours aged parasites) were incubated with 0–500 μM PPMP (full square) or AD2646 (open square) and 10 μM NBD-C6-ceramide for 60 min, at 37°C. SPM synthase activity was measured as described by Lauer et al. [19]. The percentage of SPM activity was determined by comparison of the activity measured in control cultures maintained without the analogs. Each value is the mean of triplicate experiments.
In contrast, PPMP and AD2646 have completely different effects on the TVN formation for drug concentrations that block parasite growth. After 24 hours of incubation, ring development was totally inhibited by 5 μM PPMP and no TVN was observed as previously described [20] (Figure 7C). As in controls maintained without drug (Figure 7A), TVN was distinctly observed after 24 hours of incubation of rings with 60 nM AD2646 (Figure 7B). This concentration blocks irreversibly the parasite development indicating that AD2646 has no major effect on TVN formation.
Figure 7 Effects of AD2646 and PPMP on the formation of the tubovesicular network of P. falciparum. Infected erythrocytes at the ring stage were incubated for 24 hours in presence of 60 nM AD2646 (B) or 5 μM PPMP (C). TVN formation in treated cells and untreated cells (A) was evaluated by membrane staining using BODIPY-Fl-C5-ceramide. Arrow: TVN. Bar: 5 μm.
These data do not support the hypothesis of parasite growth inhibition due to an inhibition of the parasite SPM synthase activity as was demonstrated for PPMP [19-22] : 1) The anti-Plasmodium activity of AD2646 does not correlate with its inhibitory activity on the SPM synthase. Although AD2646 and PPMP showed similar inhibitory activity on this enzymic activity in parasites in cultures, AD2646 is about 300 times more efficient in inhibiting parasite development than PPMP; 2) In contrast to PPMP which inhibits the parasite development preferentially and reversibly at the ring stage [19], AD2646 inhibited parasite development preferentially and irreversibly at the trophozoite stage (Figure 5); 3) Inhibition of the SPM synthase activity by PPMP is associated with an inhibition of the TVN formation [19-22]. This was not observed in the presence of AD2646 (Figure 7).
What could be the mechanism(s) of action of ceramide analogs in methylene linkage on P. falciparum?
By their lipidic nature, these analogs might act through a detergent effect that could lead to lysis or modification of the integrity of infected-erythrocyte membranes. This apparently is not the case. No significant lysis of normal erythrocytes was observed after 48 hous of incubation with concentrations of analogs up to 10 μM (data not shown). Furthermore, no preferential lysis of infected-erythrocytes was observed on Giemsa-stained smears of infected cultures maintained 48 hours with 250 nM AD2646, a concentration inhibiting parasite growth totally.
Interestingly, the absence of a fatty acyl carbonyl group (methylene linkage) in our ceramide analogs is a critical factor for the efficacy of their antiplasmodial activity. Sphingolipids preferentially interact with cholesterol in membranes, especially in detergent-resistant microdomains (DRMs or rafts). Rafts have been described in Plasmodium and are involved, at least, in the uptake of erythrocyte raft proteins and maintenance of the parasitophorous vacuole containing the parasite, inside the erythrocyte [37]. This interaction implies : 1) van der Waals interactions between the saturated acyl chain and sphingoid moiety of sphingolipids and the rigid planar tetracyclic rings of cholesterol [38] and 2) hydrogen bonds between the 3-β hydroxyl group of cholesterol and the fatty acyl carbonyl group resulting from the amide linkage with the sphingoid moiety [39]. The amide-linked fatty acid function seems to have a profound stabilizing effect on cholesterol-sphingolipid interactions [40]. It could be hypothesized that in a membrane context, methylene analogs might have a destabilizing effect on the cholesterol-sphingolipid interactions and, in consequence, modifications of membrane properties. Indeed, P. falciparum growth is characterized by a setting up of new permeabilities of the infected-erythrocyte membrane [41]. Although the biochemical nature of these new permeabilities is still unknown, they have been characterized from an electrophysiological point of view and involve a malaria-induced anion channel [42,43]. The effect of ceramide analogs was investigated on the properties of this channel. A 24 hours-incubation of infected-erythrocytes with 250 nM AD2646 or 10 μM PPMP did not modulate significantly the induced channel activity measured in the whole-cell configuration of the patch-clamp technique (S. Egee, unpublished data), suggesting that these ceramide analogs do not inhibit parasite growth through modifications of infected-erythrocyte membrane permeabilities.
Ceramide is at the parting of different ways of sphingolipid metabolism. Analogs have the potential to inhibit different ceramide-metabolizing enzymes and then might have a pleiotropic effect. Ceramide analogs in amide linkage were described as potent inhibitors of alkaline ceramidase in HL60 human myeloid leukemic cells [44,45]. Methylene analogs inhibit the biosynthesis of SPM and glycosphingolipids in HL60 cells, and acid ceramidase in vitro [10]. When applied to cancer cells, such analogs induced an elevation of the endogenous level of ceramide with the consequent effects of growth suppression and cell death by apoptosis [44,45]. In contrast to what was observed for cancer cells [27], preliminary results suggest that the ceramide analog AD2646 induced non-apoptotic death of P. falciparum. Parasites exposed to 1 μM AD2646 for up to 36 hours failed to exhibit characteristic apoptosis, as determined by terminal deoxynucleotidyl transferase DNA fragmentation assay and DNA fragmentation using both gel electrophoresis and fluorescence microscopy methods, although the nucleus appeared highly condensed (M. Dellinger, unpublished data). Apoptosis in P. falciparum is still controversial although some characteristics of apoptosis has been described in Plasmodium [46]. Recently, an increase in the intracellular ceramide content and an activation of parasite sphingomyelinase(s) were found to be associated with a non-apoptotic parasite death process as induced by artemisinine and mefloquine [26]. The hypothesis that AD2646 induced parasite death through modulation of endogenous ceramide level, as observed for cancer cells, is under investigation.
Authors's contribution
ML and PG carried out the in vitro inhibition assays on P. falciparum and MRC-5 cells. MG and MD performed the fluorescence microscopy and apoptosis investigations on P. falciparum, respectively. SE and ST carried out electrophysiological studies on the malaria-induced anion channel. AD, CW and SG participated in the design and synthesis of ceramide analogs. All authors read and approved the final manuscript.
Acknowledgment
This work was supported by the French Ministry of Research and New Technologies and by a grant from the Israel Science Foundation (No 607/02). We thank Prof. Joseph Schrével, Dr Isabelle Florent, Dr Stefan H.I.Kappe and Dr James Trotter for reading the manuscript and their fruitful comments.
==== Refs
Ohanian J Ohanian V Sphingolipids in mammalian cell signalling Cell Mol Life Sci 2001 58 2053 2068 11814056
van Blitterswijk WJ van der Luit AH Veldman RJ Verheij M Borst J Ceramide: second messenger or modulator of membrane structure and dynamics? Biochem J 2003 369 199 211 12408751 10.1042/BJ20021528
Jenkins GM The emerging role for sphingolipids in the eukaryotic heat shock response Cell Mol Life Sci 2003 60 701 710 12785717 10.1007/s00018-003-2239-0
Brown DA London E Functions of lipid rafts in biological membranes Annu Rev Cell Dev Biol 1998 14 111 136 9891780 10.1146/annurev.cellbio.14.1.111
Dickson RC Lester RL Sphingolipid functions in Saccharomyces cerevisiae Biochim Biophys Acta 2002 1583 13 25 12069845
Lai EC Lipid rafts make for slippery platforms J Cell Biol 2003 162 365 370 12885764 10.1083/jcb.200307087
Simons K Ikonen E Functional rafts in cell membranes Nature 1997 387 569 572 9177342 10.1038/42408
Alonso MA Millan J The role of lipid rafts in signalling and membrane trafficking in T lymphocytes J Cell Sci 2001 114 3957 3965 11739628
Hannun YA Functions of ceramide in coordinating cellular responses to stress Science 1996 274 1855 1859 8943189 10.1126/science.274.5294.1855
Hannun YA Loomis CR Merrill AHJ Bell RM Sphingosine inhibition of protein kinase C activity and of phorbol dibutyrate binding in vitro and in human platelets J Biol Chem 1986 261 12604 12609 3462188
Spiegel S Milstien S Sphingolipid metabolites: members of a new class of lipid second messengers J Membr Biol 1995 146 225 237 8568838
Zhang H Buckley NE Gibson K Spiegel S Sphingosine stimulates cellular proliferation via a protein kinase C- independent pathway J Biol Chem 1990 265 76 81 2294122
Luberto C Hannun YA Sphingolipid metabolism in the regulation of bioactive molecules Lipids 1999 34 S5 11 10419081
Gerold P Schwarz RT Biosynthesis of glycosphingolipids de-novo by the human malaria parasite Plasmodium falciparum Mol Biochem Parasitol 2001 112 29 37 11166384 10.1016/S0166-6851(00)00336-4
Wang E Norred WP Bacon CW Riley RT Merrill AHJ Inhibition of sphingolipid biosynthesis by fumonisins. Implications for diseases associated with Fusarium moniliforme J Biol Chem 1991 266 14486 14490 1860857
Riley RT Plattner RD Fermentation, partial purification, and use of serine palmitoyltransferase inhibitors from Isaria (= Cordyceps) sinclairii Methods Enzymol 2000 311 348 361 10563339
Couto AS Caffaro C Uhrig ML Kimura E Peres VJ Merino EF Katzin AM Nishioka M Nonami H Erra-Balsells R Glycosphingolipids in Plasmodium falciparum. Presence of an active glucosylceramide synthase Eur J Biochem 2004 271 2204 2214 15153110 10.1111/j.1432-1033.2004.04150.x
Shayman JA Lee L Abe A Shu L Inhibitors of glucosylceramide synthase Methods Enzymol 2000 311 373 387 10563341
Lauer SA Ghori N Haldar K Sphingolipid synthesis as a target for chemotherapy against malaria parasites Proc Natl Acad Sci U S A 1995 92 9181 9185 7568097
Lauer SA Rathod PK Ghori N Haldar K A membrane network for nutrient import in red cells infected with the malaria parasite Science 1997 276 1122 1125 9148808 10.1126/science.276.5315.1122
Elmendorf HG Haldar K Plasmodium falciparum exports the Golgi marker sphingomyelin synthase into a tubovesicular network in the cytoplasm of mature erythrocytes J Cell Biol 1994 124 449 462 8106545 10.1083/jcb.124.4.449
Haldar K Intracellular trafficking in Plasmodium-infected erythrocytes Curr Opin Microbiol 1998 1 466 471 10066518 10.1016/S1369-5274(98)80067-2
Hanada K Mitamura T Fukasawa M Magistrado PA Horii T Nishijima M Neutral sphingomyelinase activity dependent on Mg2+ and anionic phospholipids in the intraerythrocytic malaria parasite Plasmodium falciparum Biochem J 2000 346 Pt 3 671 677 10698693 10.1042/0264-6021:3460671
Lauer SA Chatterjee S Haldar K Uptake and hydrolysis of sphingomyelin analogues in Plasmodium falciparum-infected red cells Mol Biochem Parasitol 2001 115 275 281 11420114 10.1016/S0166-6851(01)00281-X
Hanada K Palacpac NM Magistrado PA Kurokawa K Rai G Sakata D Hara T Horii T Nishijima M Mitamura T Plasmodium falciparum phospholipase C hydrolyzing sphingomyelin and lysocholinephospholipids is a possible target for malaria chemotherapy J Exp Med 2002 195 23 34 11781362 10.1084/jem.20010724
Pankova-Kholmyansky I Dagan A Gold D Zaslavsky Z Skutelsky E Gatt S Flescher E Ceramide mediates growth inhibition of the Plasmodium falciparum parasite Cell Mol Life Sci 2003 60 577 587 12737317 10.1007/s000180300049
Dagan A Wang C Fibach E Gatt S Synthetic, non-natural sphingolipid analogs inhibit the biosynthesis of cellular sphingolipids, elevate ceramide and induce apoptotic cell death Biochim Biophys Acta 2003 1633 161 169 14499735
Trager W Jensen JB Human malaria parasites in continuous culture Science 1976 193 673 675 781840
Grellier P Valentin A Millerioux V Schrevel J Rigomier D 3-Hydroxy-3-methylglutaryl coenzyme A reductase inhibitors lovastatin and simvastatin inhibit in vitro development of Plasmodium falciparum and Babesia divergens in human erythrocytes Antimicrob Agents Chemother 1994 38 1144 1148 8067753
Desjardins RE Canfield CJ Haynes JD Chulay JD Quantitative assessment of antimalarial activity in vitro by a semiautomated microdilution technique Antimicrob Agents Chemother 1979 16 710 718 394674
Haldar K Uyetake L Ghori N Elmendorf HG Li WL The accumulation and metabolism of a fluorescent ceramide derivative in Plasmodium falciparum-infected erythrocytes Mol Biochem Parasitol 1991 49 143 156 1775154 10.1016/0166-6851(91)90137-U
Bligh EG Dyer WJ A rapid method of total lipid extraction and purification Can J Med Sci 1959 37 911 917 13671378
Vunnam RR Radin NS Analogs of ceramide that inhibit glucocerebroside synthetase in mouse brain Chem Phys Lipids 1980 26 265 278 6445239 10.1016/0009-3084(80)90057-2
Abe A Radin NS Shayman JA Wotring LL Zipkin RE Sivakumar R Ruggieri JM Carson KG Ganem B Structural and stereochemical studies of potent inhibitors of glucosylceramide synthase and tumor cell growth J Lipid Res 1995 36 611 621 7775872
Karasavvas N Erukulla RK Bittman R Lockshin R Zakeri Z Stereospecific induction of apoptosis in U937 cells by N-octanoyl- sphingosine stereoisomers and N-octyl-sphingosine. The ceramide amide group is not required for apoptosis Eur J Biochem 1996 236 729 737 8612651 10.1111/j.1432-1033.1996.00729.x
Bielawska A Crane HM Liotta D Obeid LM Hannun YA Selectivity of ceramide-mediated biology. Lack of activity of erythro-dihydroceramide J Biol Chem 1993 268 26226 26232 8253743
Lauer S VanWye J Harrison T McManus H Samuel BU Hiller NL Mohandas N Haldar K Vacuolar uptake of host components, and a role for cholesterol and sphingomyelin in malarial infection Embo J 2000 19 3556 3564 10899110 10.1093/emboj/19.14.3556
Kan CC Ruan ZS Bittman R Interaction of cholesterol with sphingomyelin in bilayer membranes: evidence that the hydroxy group of sphingomyelin does not modulate the rate of cholesterol exchange between vesicles Biochemistry 1991 30 7759 7766 1868053
Sankaram MB Thompson TE Interaction of cholesterol with various glycerophospholipids and sphingomyelin Biochemistry 1990 29 10670 10675 2176878
Bittman R Kasireddy CR Mattjus P Slotte JP Interaction of cholesterol with sphingomyelin in monolayers and vesicles Biochemistry 1994 33 11776 11781 7918394
Kirk K Membrane transport in the malaria-infected erythrocyte Physiol Rev 2001 81 495 537 11274338
Desai SA Bezrukov SM Zimmerberg J A voltage-dependent channel involved in nutrient uptake by red blood cells infected with the malaria parasite Nature 2000 406 1001 1005 10984055 10.1038/35023000
Egee S Lapaix F Decherf G Staines HM Ellory JC Doerig C Thomas SL A stretch-activated anion channel is up-regulated by the malaria parasite Plasmodium falciparum J Physiol 2002 542 795 801 12154179 10.1113/jphysiol.2002.022970
Bielawska A Greenberg MS Perry D Jayadev S Shayman JA McKay C Hannun YA (1S,2R)-D-erythro-2-(N-myristoylamino)-1-phenyl-1-propanol as an inhibitor of ceramidase J Biol Chem 1996 271 12646 12654 8647877 10.1074/jbc.271.40.24800
Selzner M Bielawska A Morse MA Rudiger HA Sindram D Hannun YA Clavien PA Induction of apoptotic cell death and prevention of tumor growth by ceramide analogues in metastatic human colon cancer Cancer Res 2001 61 1233 1240 11221856
Al-Olayan EM Williams GT Hurd H Apoptosis in the malaria protozoan, Plasmodium berghei: a possible mechanism for limiting intensity of infection in the mosquito Int J Parasitol 2002 32 1133 1143 12117496 10.1016/S0020-7519(02)00087-5
| 15588325 | PMC539285 | CC BY | 2021-01-04 16:37:27 | no | Malar J. 2004 Dec 10; 3:49 | utf-8 | Malar J | 2,004 | 10.1186/1475-2875-3-49 | oa_comm |
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-1-411557595810.1186/1742-4690-1-41ResearchHTLV-1 and -2 envelope SU subdomains and critical determinants in receptor binding Kim Felix J [email protected] Nicolas [email protected] Edith N [email protected] Carine [email protected] Marc [email protected] Jean-Luc [email protected] Institut de Génétique Moléculaire de Montpellier (IGMM), CNRS-UMR5535, IFR122 1919 Rte de Mende, F-34293 Montpellier Cedex 5, France2 Current address: Memorial Sloan-Kettering Cancer Center 1275 York Ave, New York, NY, 10021, USA2004 2 12 2004 1 41 41 13 9 2004 2 12 2004 Copyright © 2004 Kim et al; licensee BioMed Central Ltd.2004Kim 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
Human T-cell leukemia virus (HTLV) -1 and -2 are deltaretroviruses that infect a wide range of cells. Glut1, the major vertebrate glucose transporter, has been shown to be the HTLV Env receptor. While it is well established that the extracellular surface component (SU) of the HTLV envelope glycoprotein (Env) harbors all of the determinants of interaction with the receptor, identification of SU subdomains that are necessary and sufficient for interaction with the receptor, as well as critical amino acids therein, remain to be precisely defined. Although highly divergent in the rest of their genomes, HTLV and murine leukemia virus (MLV) Env appear to be related and based on homologous motifs between the HTLV and MLV SU, we derived chimeric HTLV/MLV Env and soluble HTLV-1 and -2 truncated amino terminal SU subdomains.
Results
Using these SU constructs, we found that the 183 and 178 amino terminal residues of the HTLV-1 and -2 Env, respectively, were sufficient to efficiently bind target cells of different species. Binding resulted from bona fide interaction with the HTLV receptor as isolated SU subdomains specifically interfered with HTLV Env-mediated binding, cell fusion, and cell-free as well as cell-to-cell infection. Therefore, the HTLV receptor-binding domain (RBD) lies in the amino terminus of the SU, immediately upstream of a central immunodominant proline rich region (Env residues 180 to 205), that we show to be dispensible for receptor-binding and interference. Moreover, we identified a highly conserved tyrosine residue at position 114 of HTLV-1 Env, Tyr114, as critical for receptor-binding and subsequent interference to cell-to-cell fusion and infection. Finally, we observed that residues in the vicinity of Tyr114 have lesser impact on receptor binding and had various efficiency in interference to post-binding events.
Conclusions
The first 160 residues of the HTLV-1 and -2 mature cleaved SU fold as autonomous domains that contain all the determinants required for binding the HTLV receptor.
==== Body
Background
Human T-cell leukemia virus type 1 (HTLV-1) has been found primarily in CD4+ and CD8+ T-lymphocytes in vivo [1-3], whereas CD8+ T-lymphocytes are thought to be the in vivo reservoir of HTLV-2 [4]. However, the in vitro tropism of HTLV-1 and -2, as determined using HTLV envelope-pseudotyped virions or envelope-induced cell fusion assays, appears to be ubiquitous [5-7]. Indeed, we recently showed that Glut1, the ubiquitous vertebrate glucose transporter, serves as a receptor for HTLV-1 and -2 envelope glycoprotein (Env) [8]. While the precise organization and properties of the receptor-interacting Env domains has not been reported, we found that the amino terminal two-thirds of the HTLV-1 extracellular surface component (SU) are sufficient to confer HTLV-1 tropism to an ecotropic Friend murine leukemia virus (F-MLV) Env [9]. A cell fusion interference assay performed with this HTLV/F-MLV Env chimera and the parental Env confirmed that this 215 amino acid Env domain, harbors HTLV-1 receptor-binding determinants [9].
The corresponding domain in MLV Env SU – located upstream of a conserved K/R L L T/N L V Q motif in the SU of the HTLV-1 and F-MLV Env [9,10] – is well characterized and comprises two main functional regions: an amino terminal sequence harboring the receptor-binding determinants, VRA, VRB and VRC [11-13], and a proline-rich region (PRR), starting at the first proline residue of the GPRVPIGP sequence [11,14] and flanked by two highly conserved GXDP [15] and CXXC [16] motifs (Figure 1). In the ecotropic and amphotropic (Ampho) MLV Env, the PRR is a putative hinge region implicated in conformational changes, triggered after receptor binding, and subsequent fusion [17,18]. In the central region of the HTLV SU, a short sequence (Env residues 180 to 205) harbors high proline content and could be a homologue of the MLV PRR.
Figure 1 Homologous modular domains in HTLV and MLV envelopes. Friend-MLV (F-MLV) Env and HTLV-1 Env are schematically represented as open and solid boxes, respectively. Boxes represent, from left to right, the signal peptide which comprises the first 34 and 20 amino acid residues of F-MLV and HTLV Env, respectively, the extracellular surface component (SU) and the transmembrane component (TM) including the carboxy terminal R peptide in F-MLV, which is cleaved in the mature Env glycoprotein [64, 65]. Env landmark positions are indicated and the MLV proline-rich regions (PRR) and the HTLV SU PRR homologue (PRRH) are delineated by vertical lines within the SU at the positions indicated by solid arrowheads. The PRR and PRRH start at the first proline (P) residue downstream of the conserved GXDP motif. Env sequences represented in the figure are obtained from F-MLV strain 57 (accession number CAA26561); P-MLV, F-MCF polytropic MLV (AAA46483); X-MLV, NZB xenotropic MLV (AAA46531); A-MLV, amphotropic MLV strain 4070A (AAA46515); HTLV-2 (NP_041006); and HTLV-1, MT2 strain (VCLJMT). Residue numbering starts from the first methionine of the Env signal peptides. Proline residues and homologous motifs are noted in bold. Amino acid sequence alignments were performed using the Clustal program in the Megalign alignment software package (DNAStar) with manual adjustments.
Several studies using synthetic peptides and neutralizing antibodies against the HTLV Env have shown that determinants within this proline rich region homologue (PRRH) are involved in interference to Env-mediated syncytium formation [19-21]. The PRRH had been thought to encode the receptor-binding domain, as based on cell-to-cell fusion assays [19,22-24]. However, although PRRH synthetic peptides can block HTLV Env-mediated syncytia formation, they have no effect on HTLV SU binding [25] and infection [26]. Indeed, we and others have shown that Env receptor binding per se, as well as interference to receptor-binding, cell-to-cell fusion, syncytium formation, and infection involve several distinct cell surface-associated parameters [27-29]. In the present report, we produced soluble forms of wild-type and mutant HTLV-1 and 2 SU amino terminal subdomains and tested their receptor-binding abilities. We also tested their ability to specifically interfere with HTLV Env cell surface binding, Env-mediated cell-to-cell fusion, and retroviral infection. By testing these essential parameters of Env-mediated dissemination, we delineated the Env receptor-binding domain (RBD) to the first 160 residues of the mature HTLV-1 and -2 SU, excluding the PRRH, and we identified a conserved tyrosine residue at position 114 of HTLV-1 Env as a critical determinant for HTLV Env receptor binding.
Results
Motif conservation and similar modular organization of HTLV and MLV SU, and identification of a proline-rich region homologue (PRRH) in the HTLV SU
As shown in Figure 1, our alignment of the MLV and HTLV SU reveals several notable motif conservations outlining a similar modular organization of the MLV SU and HTLV SU. A (K/R)LL(T/N)LVQ motif, highly conserved between the F-MLV and HTLV-1 SU, is located immediately downstream of the PRR and its PRRH counterpart, respectively. Another highly conserved motif between MLV and HTLV, GXDP, is found immediately upstream of the PRR/PRRH (Figure 1). These two motifs compelled us to notice the PRRH, between the PSQ and KLLTLVQ sequences in HTLV-1, and between the PTQ and KILKFIQ sequences in HTLV-2 (Figure 1). As counted from the first and last proline in the delineated sequence, the PRRH has a proline content of 30.8% and 30.4% for HTLV-1 and -2, respectively. This is slightly lower than the 35.3%, 36%, 36%, and 35.6% proline content for the ecotropic, polytropic, xenotropic, and amphotropic MLV Env, respectively (Figure 1). The presence of a PRRH in the HTLV SU appeared to be characteristic of their MLV-like modular organization, since HTLV SU average proline content outside of the PRRH does not exceed 11%.
Functional, soluble HTLV Env-receptor binding determinants
MLV SU receptor binding determinants are all located upstream of the PRR [11,30]. To test whether the HTLV Env receptor binding determinants are also located upstream of the potential PRRH, we constructed a chimeric Env and several soluble HTLV-1 and -2 SU amino terminal subdomains. The chimeric HTLV/MLV Env, H1183FEnv, comprises the 183 amino terminal residues of the HTLV-1 SU ending with the PSQL residues fused to the PIGP sequence of the F-MLV PRR (Figure 2A). In this Env chimera the receptor-binding domain (first 269 residues) of the F-MLV Env was replaced with the potentially corresponding domain of the HTLV-1 Env SU (Figure 2A). The chimeric H1183FEnv construct – which lacks the HTLV PRRH but has the MLV PRR – was properly expressed in transfected cells and was revealed on immunoblots with an anti-MLV SU polyclonal antibody (Figure 3A). Accordingly, an anti-HTLV-1 monoclonal antibody raised against a PRRH epitope did not bind this chimeric Env (data not shown).
Figure 2 Schematic representation of HTLV/MLV Env chimeras and HTLV SU amino terminal subdomains. Env landmark positions are indicated and SU landmark sequences and positions are indicated by arrowheads. Open arrowheads indicate the position of construct borders. (A) HTLV/MLV Env chimeras. The H1215FEnv and H1183FEnv HTLV/MLV Env chimeras were obtained by replacing the 329 and 269 amino terminal residues of the F-MLV Env (open boxes) with the amino terminal 215 and 183 amino acid residues of the HTLV-1 Env (solid boxes), respectively. The H1215FEnv chimera, previously described and formerly designated HHproFc [9], has been renamed here for sake of nomenclature homogeneity. (B) Soluble HTLV-1 (H1) and HTLV-2 (H2) SU amino terminal subdomains, H1215SU, H2211SU, H1179SU, and H2178SU were constructed as fusion proteins with a carboxy terminal hemagglutinin (HA) or rabbit immunoglobulin Fc (rFc) tag. All amino acid residue numbering starts from the first methionine of the HTLV-1 or -2 Env signal peptide, the amino terminal 20 and 21 aa residues, respectively.
Figure 3 Intracellular expression of HTLV-1 Env chimeras and soluble SU subdomains. Cell extracts (A, B) or culture supernatants (C) were prepared from 293T cells transfected with either full length Env (A) or soluble SU subdomains (B, C) expression vectors as depicted in figure 2. Membranes were probed with either (A) an anti-MLV SU antiserum to detect F-MLV and H1183FEnv uncleaved Env precursor proteins (F-MLV Prgp85 and H1183Fenv Pr, respectively) indicated by arrowheads, and cleaved SU (F-MLV SUgp70 and H1183FEnv SU, respectively) indicated by circles, or (B, C) an anti-rabbit IgG antiserum to detect carboxy terminal rFc-tagged soluble subdomains, including the Ampho-MLV SU subdomain (A397SU).
HTLV-1 and -2 SU amino terminal subdomains with or without their respective PRRH were constructed as fusion proteins with either an influenza hemagglutinin (HA) or rabbit immunoglobulin Fc (rFc) carboxy terminal tag (Figure 2B). The H1215SU and H2211SU subdomains comprise the first 215 and 211 residues, counting from the first methionine in the signal peptide through the KLLTLVQ of HTLV-1 and KILKFIQ of HTLV-2 Env, respectively (Figure 2B). The H1179SU and H2178SU, comprising the amino terminal 179 and 178 amino acids of the HTLV-1 and -2 Env, respectively, exclude the PRRH sequence (Figure 2B).
Cell lysates and cell culture supernatants were analyzed to evaluate intracellular expression and secretion of functional SU amino terminal domains in transfected-cell cultures, respectively. H1215SU and H2211SU, containing the PRRH sequence, and H2178SU lacking this PRRH were all efficiently expressed in transfected cells (Figure 3B). It is noteworthy, however, that recovery of tagged H1179SU molecules was largely inefficient because the vast majority of this protein was cleaved (data not shown). In contrast, no significant cleavage was observed with the other soluble domains released in the medium (not shown) (Figure 3C). As expected for immunoadhesins, H1215SU, H2211SU, and H2178SU rFc-tagged domains were detected as dimers under non-reducing conditions (not shown). Immunoblots of cell extracts revealed two forms of intracellular H1215SU and H2211SU (Figure 3B); this was likely due to variable glycosylation of these subdomains. However, a single secreted, soluble form of each of these amino terminal subdomains was detected in cell culture supernatants (Figure 3C).
A truncated Ampho-MLV SU-rFc fusion protein that comprises the amino terminal 397 residues of the Ampho-MLV Env fused to a carboxy terminal rFc tag was constructed (A397SU) and used as a heterologous control. A single form of this truncated SU was efficiently expressed in transfected cells (Figure 3B), and abundantly secreted in cell culture medium (Figure 3C).
HTLV-1 and -2 SU subdomains with HTLV receptor binding properties
The amino terminal subdomains were tested for their ability to bind to HTLV receptor-presenting cells by flow cytometry. Using this cell surface binding assay, all of the soluble HTLV SU subdomains bound to the A23 hamster fibroblast cell line (Figure 4) as well as to all other cell lines tested, including 293T (human kidney fibroblasts), NIH3T3 and NIH3T3TK- (murine fibroblasts) [29], HeLa (human ovarian carcinoma cells), D17 (canine fibroblast), Jurkat (suspension human T cell line), activated primary human T cells, and numerous other cell lines and primary cell types that are thought to express the HTLV receptor. As expected from our previous work [31], none of these soluble HTLV SU subdomains showed detectable binding on resting T lymphocytes. Notably, binding of the HTLV SU to these cells occurred whether they formed or not syncytia in the presence of HTLV Env [29] and data not shown). Binding by H2178SU was similar to H2211SU, demonstrating that the first 158 residues of the mature HTLV-2 SU, without the 20 amino acids of the amino terminal signal peptide, are sufficient for cell surface binding, and therefore that the PRRH is not required for receptor binding (Figure 4A).
Figure 4 HTLV-1 and -2 SU subdomains interfere with HTLV Env SU cell surface binding. (A) Conditioned medium from control 293T cells (open histograms) or from 293T cells expressing soluble rFc-tagged HTLV-1 H1215SU, HTLV-2 H2211SU and H2178SU, or Ampho-MLV A397SU subdomains (filled histograms), were incubated with A23 hamster cells for 30' at 37°C and binding was assessed by flow cytometry following addition of a secondary FITC-conjugated anti rabbit IgG antibody. Similar results were obtained in binding assays performed using all cell lines described in the text. (B) To assess binding interference, target 293T cells were transfected with the indicated Env construct and subsequently incubated with the HA-tagged H2178SU domain (filled histograms). Binding was detected by FACS following incubation with an anti HA 12CA5 mouse mAb and a FITC-conjugated anti mouse IgG antibody. Open histograms represent background levels of fluorescence. SU constructs are schematically represented below each graph by solid (HTLV), open (F-MLV) or grey (Ampho-MLV) boxes.
To determine whether cell surface binding of these soluble SU domains corresponded to bona fide binding to the HTLV receptor, we performed an Env-specific binding interference assay. In this assay, transfection of the above described chimeric Env and SU subdomains into 293T cells resulted in interference to cell surface binding by the soluble HA-tagged H2178SU subdomain (Figure 4B). Indeed, nearly complete interference was observed when cells were transfected with the amino terminal subdomain constructs, in the presence and absence of PRRH sequences (H1215SU and H2211SU versus H1183FEnv and H2178SU) (Figure 4B). This effect was specific as HTLV SU binding was not inhibited by a heterologous A397SU domain (Figure 4B). Therefore, we showed that the first 163 and 158 residues, with a cleaved signal peptide, of the mature HTLV-1 and HTLV-2 SU, respectively, contained the entire HTLV Env RBD. These data also showed that HTLV-1 and 2 cross-interfered, consistent with the fact that they recognize the same cell surface receptor for infection [8,32].
Interference to HTLV Env-mediated cell-to-cell fusion by HTLV SU amino terminal subdomains
Viral envelope interference occurs when cell surface receptors are occupied by receptor-interacting Env components [33-35]. Since interference to the different Env-mediated functions involves distinct components [27-29], we also tested the abilities of the H1183FEnv and the HTLV SU amino terminal subdomains to interfere with HTLV Env-mediated cell fusion. Interference to cell fusion was measured using a quantitative HTLV envelope cell fusion interference assay (CFIA), as previously described [9].
HTLV-1 Env-induced cell fusion was significantly diminished upon expression of the H1215SU subdomain in target cells, 12% ± 2% of control fusion (P < 0.001), consistent with previous observations using the H1215FEnv chimera [9]. Significant interference to cell fusion was also observed with the H1183FEnv chimera, which lacked a PRRH, down to 26% ± 4% of control fusion (P < 0.001) (Figure 5). The corresponding HTLV-2 SU subdomains produced a nearly identical cell fusion interference profile: interference by the H2211SU isolated domain, in which the PRRH was maintained, resulted in 15% ± 3% of control cell fusion levels, while the H2178SU subdomain, lacking the HTLV PRRH, inhibited HTLV-1 Env-induced cell fusion to 24% ± 6% of control levels (P < 0.001) (Figure 5). It is noteworthy that similar data were obtained when comparing cell fusion interference by H1215FEnv and H1183FEnv. These effects were specific to HTLV SU amino terminal domains as A397SU did not interfere with HTLV-1 Env-mediated cell fusion (83% ± 11% of control fusion) (Figure 5). Furthermore, no interference was observed when these truncated HTLV SU fragments and chimeric Env were tested against heterologous, fusogenic control Env such as AΔR Env, FΔR, XenoΔR and VSVG (data not shown). Altogether, these results confirmed our findings that receptor-binding determinants are present within the first 183 and 178 amino acids of the HTLV-1 and -2 Env, respectively. They also indicated that the PRRH (H1215SU and H2211SU), although unnecessary for receptor binding, modulates the efficiency of interference to HTLV Env-induced cell-to-cell fusion (P < 0.03).
Figure 5 HTLV-1 and -2 SU subdomains interfere with HTLV Env-mediated cell fusion. Cell-to-cell fusion assays were performed by cocultivating fusogenic HTLV-1 Env-expressing cells with target cells expressing the Env derivatives indicated and schematically represented below each histogram. HTLV-1 Env-mediated cell fusion in the presence of target cells transfected with empty vector (Mock) yielded 200 to 1000 blue foci in 4 independent experiments and these levels were defined as 100% cell fusion. Cell fusion levels in the presence of HLTV SU mutants or the A397SU control Ampho-MLV SU subdomain is shown as percent of control. Mean fusion percentages were determined from three to four independent experiments. Error bars represent the standard error of the mean.
Interference to HTLV Env-mediated infection by HTLV SU amino terminal subdomains
Interference, as described above, was based on the inhibition of cell-to-cell fusion induced by fusogenic Env expressed in the absence of other viral proteins. We further evaluated the abilities of the Env chimeras and soluble subdomains to specifically interfere with HTLV Env-mediated infection. HTLV Env-pseudotyped MLV virions, MLV(HTLV), were produced to infect 293T target cells. Because these recombinant cell-free virions are not competent for replication, this viral pseudotype infection assay tests a single round of infection, and does not measure replication and subsequent exponential viral dissemination. Therefore, relative infection values are expressed in linear rather than logarithmic scales.
Infection of mock-transfected target cells, devoid of interfering Env domains, resulted in a mean infection value of 9905 ± 1117 infectious units per ml (iu/ml), and this was taken as 100% control infection (Figure 6). Similar values, 8803 ± 1871 iu/ml or 89% ± 19% of control infection, were obtained upon infection of target cells expressing a heterologous SU subdomain, A397SU (Figure 6). Expression of the H1183FEnv and H1215FEnv chimeric Env in target cells significantly reduced MLV(HTLV) infection to 324 ± 98 iu/ml, 3.3% ± 1% of control infection, and to 307 ± 129 iu/ml, 3.1% ± 1.3% of control infection, respectively (Figure 6 and data not shown). Similarly, the H2178SU and H2211SU subdomains diminished MLV(HTLV) infection to 191 ± 56 iu/ml and 215 ± 122 iu/ml, 1.9% ± 0.6% and 2.2% ± 1.3% of control infection, respectively (Figure 6). The specificity of interference to infection by HTLV Env constructs was assessed by their lack of interference abilities toward Ampho-MLV Env-pseudotyped virions, MLV(Ampho) (data not shown). Thus, for both HTLV-1 and -2, the amino terminal domain upstream of the PRRH was sufficient for specific interference to HTLV Env-mediated infection. Furthermore, in contrast to the cell fusion interference assays described above, the PRRH did not detectably influence MLV(HTLV) infection.
Figure 6 HTLV-1 and -2 SU subdomains interfere with infection by HTLV envelope-pseudotyped virions. 293T cells (5 × 105) expressing the indicated interfering Env derivatives were infected with cell-free HTLV-2 Env-pseudotyped virions MLV(HTLV) carrying a LacZ reporter gene. Infected cells were detected 2 days later by X-gal staining. Infection values are represented as percent of control infection, i.e., relative to infection of mock (pCDNA3.1) transfected target cells, calculated as infectious units per ml of virus containing supernatant (i.u./ml). Data are representative of at least three independent experiments performed in duplicate. Error bars represent the standard error of the mean.
Because HTLV dissemination appears to occur mostly via cell-to-cell contact, we also tested envelope interference to infection by HTLV-1 SU amino terminal domains using a cell-to-cell transmission interference assay. In this assay, cells harboring interfering chimeric Env and soluble subdomains were cocultured with cells producing MLV(HTLV) virions. Transfection of either chimeric Env or soluble subdomains into HeLa target cells decreased MLV(HTLV) infection to levels similar to those observed in the cell fusion interference assay presented in figure 5 (data not shown).
Identification of residues within the HTLV SU amino terminal domain that modulate receptor binding and HTLV Env-mediated interference
Two key residues contained in the HTLV SU RBD and conserved between HTLV-1 and -2, arginine 94 (Arg94) and serine 101 (Ser101) for HTLV-1 Env which correspond to Arg90and Ser97 in HTLV-2 Env, have been shown to alter cell-to-cell fusion and infection when mutated [36,37]. To determine whether mutations of these residues had an effect on receptor binding, we generated H1215SU subdomains with either Arg94 or Ser101 mutated to Ala, yielding the mutant H1(R94A)SU and H1(S101A)SU subdomains, respectively. We also evaluated mutations of Asp106, mutant H1(D106A)SU, and Tyr114, mutant H1(Y114A)SU, both residues found to be highly conserved between all human and simian T cell leukemia viruses (unpublished observations). Surprisingly, cell surface binding profiles of H1(R94A)SU and H1(S101A)SU mutants were not significantly altered when compared to binding by the parental H1215SU, whereas the H1(D106A)SU mutant presented reduced binding to HTLV receptor-bearing cells and the H1(Y114A)SU mutant showed a nearly complete abrogation of cell surface binding (Figure 7A). Loss of binding observed with the two latter mutants was not due to decreased soluble SU fragment production, as assessed by immunoblotting of transfected-cell culture media (Figure 7A). Moreover, equivalent binding profiles were obtained when the same mutations were introduced into the HTLV-2 soluble RBD H2178SU (data not shown). Altogether, these experiments demonstrated that Tyr114, and to a lesser extent Asp106, are key residues involved in HTLV Env receptor binding.
Figure 7 HTLV-1 SU amino terminal domain mutants. (A) H1215SU constructs were generated with the following SU amino terminal point mutations; R94A, S101A, D106A and Y114A. The abilities of these soluble H1215SU constructs to bind 293T cells were assessed by flow cytometry (gray histograms). The levels of expression of the various soluble SU subdomains are shown under each histogram. The abilities of the H1215SU mutants to interfere with (B) HTLV Env-induced cell fusion and (C) MLV(HTLV) pseudotype infection was assayed as described in Figs. 5 and 6. Data are representative of at least three independent experiments performed in duplicate. Error bars represent the standard error of the mean.
We next tested the abilities of these mutants to interfere with HTLV Env-mediated cell fusion and infection, using the assays described above. As mentioned above, all wild-type and mutant HTLV SU subdomains were produced and secreted with a similar efficiency (Figure 7A). Expression of the H1(D106A)SU and H1(Y114A)SU mutants, with decreased capacities to bind the HTLV receptor, correlated with decreased interference to HTLV Env-mediated cell fusion and infection. Indeed, H1(Y114A)SU, which had nearly undetectable level of binding, showed the lowest levels of interference and thus allowed the highest levels of HTLV Env-mediated cell fusion and infection (56% ± 16% and 46% ± 10%, respectively) (Figure 7). Nevertheless, levels of fusion and infection were lower than that observed when the heterologous A397SU was used as a negative control of interference (83% ± 11% and 89% ± 19% for cell fusion and infection, respectively). Thus, overexpression of mutant HTLV SU fragments with highly decreased receptor binding abilities can still exert, albeit to a significantly lesser extent, interference to HTLV Env-mediated cell fusion and infection.
We found that similar levels of interference to HTLV Env-mediated cell fusion and infection were observed when either the parental H1215SU or the mutant H1(S101A)SU were expressed in target cells (Figure 7B and 7C). This is consistent with the capacity of this mutant to bind target cells at levels similar to that of wild type H1215SU. However, interference to HTLV Env-mediated cell fusion and infection did not always correlate with cell surface binding profiles. While the H1(R94A)SU mutant inhibited cell fusion and infection, its effects were significantly lower than those of the wild-type H1215SU (56% ± 8% and 32% ± 2.3%, respectively) (Figure 7B,7C). Thus, although neither Arg94 nor Ser101 of the HTLV-1 SU appears to play a direct role in binding, Arg94 modulates HTLV Env-mediated fusion and infection (Figure 7), likely via post-binding effects rather than binding per se. In conclusion, Tyr114 appeared as the main determinant identified so far for HTLV Env binding, whereas the effects previously described with Arg94 and Ser101 are most likely associated with post-binding events.
Discussion
Here, we report the generation of MLV Env with chimeric HTLV/MLV SU and truncated HTLV-1 and -2 amino terminal SU subdomains that can be expressed in and secreted from eukaryotic cell lines in functional, soluble form. Using these constructs, we demonstrated that the amino terminal 163 and 158 residues (i.e., expunged of their Env signal peptide) of the mature HTLV-1 and -2 Env SU, respectively, were sufficient to exert both HTLV receptor binding and efficient interference to diverse HTLV Env-mediated functions, including binding, cell-to-cell fusion and cell-free as well as cell-to-cell infection. Although the PRRH sequence comprising amino acid residues 180 to 215 of the HTLV-1 Env and 176 to 211 of the HTLV-2 Env was previously thought to be a receptor binding site, our data preclude a major role for this region in the binding properties described above. Indeed, whereas a synthetic peptide composed of amino acids 197 to 216 and located within the HTLV-1 PRRH, has been reported to interfere with HTLV Env-induced syncytia formation [22], this peptide was later shown to compete neither with receptor binding of the entire HTLV-1 Env SU [38], nor with infection [26]. It is therefore likely that the effects reported for PRRH-derived peptides, as measured by syncytia formation, are solely due to post-receptor binding events. However, we identified Tyr114 of the HTLV-1 Env, which corresponds to Tyr110 of the HTLV-2 Env, as a key residue in HTLV Env binding and for all the aforementioned HTLV Env-mediated functional assays. We could not detect binding of H1(Y114A)SU by flow cytometry, while this mutant exerted residual, albeit significantly decreased, interference to HTLV Env-mediated cell fusion and infection. Altered folding outside of the binding domain per se, rather than direct alteration of the receptor-binding site, could also account for the lack of binding of this mutant. However, we favor the latter hypothesis, since the H1(Y114A)SU mutant was properly folded and transported to the plasma membrane and secreted in the medium as efficiently as wild type RBD, thus arguing against gross misfolding of this mutant. Accordingly, Tyr114 appears to be conserved in all known human and simian T cell leukemia viruses strains, which share the same receptor.
The receptor-binding site in MLV RBD is composed of a combination of several cysteine loops located upstream of the PRR [11,39] which is linked to a conserved anti-parallel β core [13]. The isolation of an F-MLV SU amino terminal subdomain allowed crystallization of MLV RBD and the modeling of the RBD cysteine loop arrangement [13]. The precise organization of cysteine loops, likely to harbor the receptor binding determinants, within the HTLV SU amino terminus remains to be established. Nevertheless, the identification of Tyr114 as a key HTLV-1 RBD residue points at this determinant as a very likely receptor-binding core. This, together with previous works relying on syncytia formation and cell-to-cell transmission [36,37], will help to distinguish between bona fide receptor binding determinants and determinants involved at a post-binding level.
Another recently identified determinant, the Pro-His-Gln SU motif conserved among gammaretroviruses such as MLV and feline leukemia viruses (FeLV), has been determined to play a major role in viral entry during post-binding events [40]. The mechanism of this effect involves a direct interaction of MLV SU soluble forms with Env attached SU carboxy terminus [41-46]. This interaction between the SU amino and carboxy termini leads to the T cell-restricted tropism of a natural isolate of FeLV, FeLV T, in which the SU Pro-His-Gln motif is mutated. Indeed, FeLV T is restricted in cat to T cells because they naturally express an endogenous soluble FeLV RBD-related factor called FeLIX that trans-complements the lack of the SU Pro-His-Gln motif in the FeLV T Env and restores its post-binding defect [47]. Despite the HTLV-1 and F-MLV SU homologous modular organization and the assignment of several common motifs between the two latter SU, no obvious Pro-His-Gln motif homologue is present in the HTLV SU amino terminus. Whether a FeLIX-like molecule that interacts with HTLV Env exists in human T cells remains to be addressed. Furthermore, the fact that the Pro-His-Gln has been shown to play a major role in transactivation of viral infection in several gammaretroviruses which are efficiently infectious as cell-free virions [42,44,48], raises the question whether the apparent lack of such a motif in the HTLV simple oncovirus-like SU is linked to the relative inefficiency of HTLV Env-mediated infection by cell-free virions. The HTLV SU subdomains described here should prove to be valuable in addressing such questions.
The recent identification of Glut1, the ubiquitous glucose transporter of vertebrates [49], as a receptor for HTLV Env [8] adds an additional similarity between the Env of HTLV, a deltaretrovirus, and that of gammaretroviruses. All these virus Env recognize multimembrane-spanning metabolite transporters [50,51]. This and the common modular organization of the HTLV and MLV SU raise questions regarding the origin of the HTLV Env. It has previously been reported that envelopes of invertebrate retroviruses may have been "captured" from other viruses [52-54]. As HTLV and MLV have strongly divergent overall genomic organizations, "envelope capture" from related ancestor genes might account for the close relationship between the Env of these phylogenetically distant viruses [10].
Conclusions
We have generated truncated domains of the HTLV Env amino terminus, upstream of residues 183 and 178 of the HTLV-1 and -2 Env, respectively, that were sufficient to bind target cells of different species through interaction with the HTLV Env receptor. We also identified a tyrosine at position 114 and 110 in HTLV-1 and -2 Env, respectively, as a key determinant for this binding. In addition to their use for further exploration of the mechanisms involved in HTLV entry, the tagged HTLV-1 and -2 RBD subdomains described here are novel tools for the detection of Glut1 cell surface expression and intracellular trafficking. Indeed, we tracked intracellular expression of EGFP-tagged HTLV SU subdomains by time-lapse microscopy, and found that they are preferentially routed toward cell-cell contact areas (unpublished observations), where Glut1 is particularly abundant [55] and our unpublished observations). Furthermore, those HTLV SU derivatives could be of particular importance in view of the key roles played by Glut1 in various biological processes, including T cell survival and activation [31,56], tumor genesis [57,58], and neuronal activity [59]. Interestingly, soluble HTLV SU subdomains inhibit Glut1-mediated glucose transport, and accordingly, expression of mutants with diminished receptor binding ability resulted in less pronounced inhibition [8] and data not shown). Thus, these HTLV SU derivatives could also be used as glucose transport inhibitors. These data demonstrate the potential for the novel and broad utility of these reagents in the study of HTLV infection as well as biological processes involving glucose transport and metabolism.
Materials and methods
Construction of chimeric Env and HTLV-1 and -2 SU subdomains
To exchange the PRR and PRRH regions, we introduced an allelic MfeI restriction site in the HTLV-1 and F-MLV Env. Introduction of this site in F-MLV resulted in the substitution of a glutamine and leucine (QL) dipeptide for the parental arginine and valine (RV) residues of the GPRVPIGP motif, at the start of the MLV Env PRR. Introduction of the MfeI site in the PSQL motif of the HTLV-1 SU maintained the parental QL residues, at the start of the HTLV Env PRRH. By exchanging domains at the MfeI sites, we derived the H1183FEnv chimera containing the amino terminal 183 residues of the HTLV Env followed by the F-MLV PRR. In this chimera, the PSQL/PIGP hybrid sequence is generated at the exchange border, and the PRRH of HTLV is replaced by the F-MLV PRR (Figure 2A). In contrast, the entire PRRH of HTLV-1 is present in the H1215FEnv chimera – this Env chimera has been previously described and designated HHproFc [9]. The H1183FEnv and H1215FEnv chimeras, as well as the parental HTLV-1 and F-MLV Env, were inserted in an allelic fashion into the previously described pCEL retroviral Env expression vector [60]. The HTLV-2 Env expression vector, pCSIX/H2, was constructed by inserting the HindIII – EcoRI fragment from pHTE-2 (a gift from M-C Dokhelar) encompassing the HTLV-2 env gene, the pX region and the 3' LTR into pCSI (CMV promoter, SV-40 intron) [61] at the HindIII and EcoRI restriction sites.
The H1215SU, H2211SU, H1179SU, and H2178SU subdomains, corresponding to the HTLV-1 and -2 SU amino terminus with and without their respective PRRH, were generated by PCR and subcloned into the pCSI expression vector as fusion proteins harboring a carboxy terminal rFc or HA tag (Figure 2B). The H1(R94A)SU, H1(S101A)SU, H1(D106A)SU, and H1(Y114A)SU substitution mutants were generated by oligonucleotide-directed PCR mutagenesis on the H1215SU vector and subcloned into the pCSI expression vector. All PCR-generated DNA fragments were sequenced using an ABI Prism 310 sequencer. Cloning details are available upon request.
Protein expression and immunoblots
Approximately 5 × 105 293T cells per 35 mm well were transfected with 5 μg of vectors using a calcium-phosphate-Hepes buffered saline (HBS) transfection protocol. Transfection medium was replaced with 3 ml of fresh culture medium twenty hours post-transfection. Forty-eight hours post-transfection cell culture medium (supernatant) was recovered and filtered through a 0.45 μm pore-size membrane to remove cell debris. Twenty μl were directly analyzed by SDS-PAGE (15% polyacrylamide gel), and the rest was aliquoted and stored at -20°C for later use in binding assays (see below). Cell extracts were collected 48 h post-transfection in 1 ml of cell lysis buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 0.1% sodium dodecyl sulfate [SDS], 1% Nonidet P-40, 0.5% deoxycholate, and a cocktail of mammalian protease inhibitors [Sigma]) and clarified by two successive centrifugations at 13,000 rpm for 10 min at 4°C in a microcentrifuge. Approximately 20 μl of each extract, adjusted after normalization for protein concentration using the Bradford assay (Sigma), were subjected to electrophoresis on SDS-15% acrylamide gels, followed by transfer onto nitrocellulose (Protran; Schleicher & Schuell). Membranes were blocked in phosphate-buffered saline (PBS) containing 5% powdered milk and 0.5% Tween 20, probed with a 1:1000 dilution of a goat anti-RLV gp70 polyclonal antibody (Viromed) followed by a horseradish peroxidase-conjugated anti-goat immunoglobulin (for detection of chimeric Env), or goat anti-rabbit-IgG-horseradish peroxidase-conjugated immunoglobulins (for detection of rFc-tagged SU subdomains). Immunoblots were subsequently washed three times with PBS-0.1% Tween 20 and revealed by chemiluminescence (ECL+, Amersham).
Binding and binding interference assays
Binding assays were performed as previously described [31]. Briefly, 5 × 105 target cells were detached with a PBS-EDTA solution, collected by centrifugation, incubated for 30' at 37°C with 300 μl of rabbit Fc-tagged soluble HTLV-1, HTLV-2, or Ampho-MLV truncated SU, washed, labeled with an anti-rabbit-IgG FITC-conjugated antibody, and analyzed on a FACSCalibur (Becton Dickinson). Data analysis was performed using the CellQuest software (Becton Dickinson). For interference studies, 293T cells were transfected with 4 μg of Env or Env SU subdomain expression vectors (carboxy terminal rFc-tagged forms) using the calcium-phosphate-HBS method. Under these conditions, transfection efficiencies ranged from approximately 80 to 90% of the target cells. Twenty-four and 48 hours post-transfection, cells were collected and transfected 293T cells expressing the different interfering HTLV or Ampho-MLV domains were incubated with a challenging HA-tagged soluble HTLV-2 SU amino terminal subdomain (H2178SU-HA). Cells were stained using a primary 12CA5 anti HA antibody followed by an anti-mouse-IgG FITC-conjugated antibody before detection by flow cytometry.
Envelope interference to cell fusion assay
Briefly, the HTLV/MLV Env chimera, H1183FEnv, was used to interfere with challenging HTLV Env. The interfering non-fusogenic H1183FEnv and truncated HTLV SU subdomains were transiently transfected into HeLaCD4LTRLacZ, a cell line highly susceptible to HTLV Env-induced fusion that contains a stably integrated Tat-dependent LacZ expression vector [62]. These transfectants were cocultured with Tat-expressing NIH3T3(TK-) cells (NIH3T3(TK-)Tat) that were transiently transfected with the challenging HTLV Env. The NIH3T3(TK-)Tat cell line is resistant to HTLV-Env-induced syncytia formation, despite its ability to express the HTLV receptor and to bind HTLV Env, and thus can be used to precisely monitor fusion of the HeLaCD4LTRLacZ target cells [9,29]. H1183FEnv Env and truncated HTLV SU subdomains plasmid DNA (2 to 3 μg) was transfected into HeLaCD4LTRLacZ cells, while challenging, fusogenic HTLV-1 Env plasmid (1 μg) was transfected into NIH3T3(TK-)Tat. The interfering Env or SU subdomain-presenting cells were detached 24 hours post-transfection and 1–2 × 105 cells were cocultured for 24 hours with 1–2 × 105 challenging HTLV-1 Env-presenting NIH3T3(TK-)Tat cells. Subsequently, the cocultured cells were fixed and stained for β-galactosidase expression as described previously [60]. Transfection efficiencies of the HeLaCD4LTRLacZ target cells were approximately 50%. Mock transfections were performed with similar amounts of control plasmid DNAs. Env interference was measured by the decreased number of blue foci and was expressed as percent blue foci of control fusion (mock-transfected target cells). Data are represented as mean interference (± standard deviation), and statistical significance of interference levels was determined using a pairwise Student's t test.
Envelope interference to infection assay
MLV(Ampho) and MLV(HTLV) pseudotyped virions were produced after transfection of 106 293T cells with 5 μg pCSI/Ampho or pCSIX/H2, respectively, 5 μg pCL/Gag-Pol [29] and 10 μg of pCLMFG-LacZ [63], using a calcium-phosphate-HBS transfection protocol. Supernatants were recovered 48 hours post transfection and filtered through 0.45 μm pore-size membrane to remove cell debris, and stored at -80°C. The pCLMFG-LacZ plasmid is a retroviral expression vector that provides a packageable RNA coding for the LacZ gene marker. pCSI/Ampho is an expression vector encoding the Ampho-MLV Env, and the HTLV-2 Env expression vector, pCSIX/H2, is described above.
Virion-containing supernatants were used to infect target 293T cells expressing the chimeric Env or HTLV RBD subdomains. Transfection efficiencies of target 293T cells were >80% in all experiments. Infections were performed 36–48 hours post-transfection on cultures grown in 12 well plates (Costar) at 37°C, medium was changed 24 hours later, and confluent cell monolayers were fixed, stained for β-galactosidase activity before counting blue foci. Interference to infection was determined by infecting transfected target cells with approximately 100 and 1000 iu. Infection was evaluated as described above, and the number of LacZ-positive blue colonies counted was normalized by multiplying by the appropriate dilution factor. The resulting infection values were analyzed as iu/ml of virus containing supernatant. Subsequently the relative infection levels in cells expressing the HTLV SU domains were compared to those of mock transfected cells and were expressed as percentages of control infection (% control).
List of abbreviations used
HTLV Human T-cell leukemia virus
SU envelope extracellular surface component
Env envelope glycoprotein
MLV murine leukemia virus
F-MLV Friend-MLV
RBD receptor-binding domain
PRR proline-rich region
PRRH proline rich region homologue
Ampho amphotropic
HA influenza hemagglutinin
rFc rabbit immunoglobulin constant fragment
A397SU Ampho-MLV Env fused to a carboxy terminal rFc tag
CFIA cell fusion interference assay
iu/ml infectious units per ml
Arg94arginine 94
Ser101serine 101
Tyr114tyrosine 114
FeLV feline leukemia viruses
HBS Hepes buffered saline
PBS phosphate-buffered saline
SDS sodium dodecyl sulfate
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
FJK designed and realized or supervised most of the experiments and co-wrote the manuscript. NM participated to some molecular constructions, set up, realized and analyzed most binding assays and FACS analyses and participated to the redaction of the manuscript. ENG set up and performed the cell-to-cell transmission assay and performed the corresponding experiments, CV constructed some of the RBD point mutants and tested them, MS initiated the project, co-participated in the design of the study, co-coordinated its realization and co-wrote the manuscript, and JLB realized some of the molecular constructs, performed some of the experiments, co-participated in the design of the study, co-coordinated its realization and co-wrote the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We thank N. Taylor for helpful discussion and critical reading of the manuscript, G. Labesse for his help in protein sequence analyses, R.K. Naviaux for the gift of pCL-Eco and pMFG-LacZ plasmids, J.A. Young for the rabbit Fc plasmid, J.-C. Dantonel for the anti-HA antibody, F. Carbonell for technical assistance, and all the members of our laboratory for insightful discussion. FJK was supported by an award from the Philippe Foundation and successive fellowships from the Agence Nationale pour la Recherche contre le SIDA (ANRS), the Association pour la Recherche contre le Cancer (ARC), and the Fondation de France. NM is supported by a graduate student fellowship from the MRT. JLB and MS are supported by the Institut National de la Santé et de la Recherche Médicale (INSERM). This work was supported by grants from ARC (ARC Nos. 5989 and 3424), Fondation de France (Nos. 2291 and 2138) and Association Française contre les Myopathies (AFM No.7706) to MS.
==== Refs
Richardson JH Edwards AJ Cruickshank JK Rudge P Dalgleish AG In vivo cellular tropism of human T-cell leukemia virus type 1 J Virol 1990 64 5682 5687 1976827
Hanon E Stinchcombe JC Saito M Asquith BE Taylor GP Tanaka Y Weber JN Griffiths GM Bangham CR Fratricide among CD8(+) T lymphocytes naturally infected with human T cell lymphotropic virus type I Immunity 2000 13 657 664 11114378 10.1016/S1074-7613(00)00065-0
Nagai M Brennan MB Sakai JA Mora CA Jacobson S CD8(+) T cells are an in vivo reservoir for human T-cell lymphotropic virus type I Blood 2001 98 1858 1861 11535522 10.1182/blood.V98.6.1858
Wang TG Ye J Lairmore MD Green PL In vitro cellular tropism of human T cell leukemia virus type 2 AIDS Res Hum Retroviruses 2000 16 1661 1668 11080807 10.1089/08892220050193119
Sutton RE Littman DR Broad host range of human T-cell leukemia virus type 1 demonstrated with an improved pseudotyping system J Virol 1996 70 7322 7326 8794391
Okuma K Nakamura M Nakano S Niho Y Matsuura Y Host range of human T-cell leukemia virus type I analyzed by a cell fusion-dependent reporter gene activation assay Virology 1999 254 235 244 9986790 10.1006/viro.1998.9530
Trejo SR Ratner L The HTLV receptor is a widely expressed protein Virology 2000 268 41 48 10683325 10.1006/viro.2000.0143
Manel N Kim FJ Kinet S Taylor N Sitbon M Battini JL The Ubiquitous Glucose Transporter GLUT-1 Is a Receptor for HTLV Cell 2003 115 449 459 14622599 10.1016/S0092-8674(03)00881-X
Kim FJ Seiliez I Denesvre C Lavillette D Cosset FL Sitbon M Definition of an amino-terminal domain of the human T-cell leukemia virus type 1 envelope surface unit that extends the fusogenic range of an ecotropic murine leukemia virus J Biol Chem 2000 275 23417 23420 10851227 10.1074/jbc.C901002199
Kim FJ Manel N Battini JL Sitbon M Emergence of vertebrate retroviruses and envelope capture Virology 2004 318 183 191 14972546 10.1016/j.virol.2003.09.026
Battini JL Heard JM Danos O Receptor choice determinants in the envelope glycoproteins of amphotropic, xenotropic, and polytropic murine leukemia viruses J Virol 1992 66 1468 1475 1310758
Battini JL Danos O Heard JM Receptor-binding domain of murine leukemia virus envelope glycoproteins J Virol 1995 69 713 719 7815534
Fass D Davey RA Hamson CA Kim PS Cunningham JM Berger JM Structure of a murine leukemia virus receptor-binding glycoprotein at 2.0 angstrom resolution Science 1997 277 1662 1666 9287219 10.1126/science.277.5332.1662
Koch W Hunsmann G Friedrich R Nucleotide sequence of the envelope gene of Friend murine leukemia virus J Virol 1983 45 1 9 6296423
Gallaher WR Ball JM Garry RF Martin-Amedee AM Montelaro RC A general model for the surface glycoproteins of HIV and other retroviruses AIDS Res Hum Retroviruses 1995 11 191 202 7742034
Sitbon M d'Auriol L Ellerbrok H André C Nishio J Perryman S Pozo F Hayes SF Wehrly K Tambourin P Galibert F Chesebro B Substitution of leucine for isoleucine in a sequence highly conserved among retroviral envelope surface glycoproteins attenuates the lytic effect of the Friend murine leukemia virus Proc Natl Acad Sci USA 1991 88 5932 5936 2062871
Andersen KB A domain of murine retrovirus surface protein gp70 mediates cell fusion, as shown in a novel SC-1 cell fusion system J Virol 1994 68 3175 3182 7512160
Lavillette D Maurice M Roche C Russell SJ Sitbon M Cosset FL A proline-rich motif downstream of the receptor binding domain modulates conformation and fusogenicity of murine retroviral envelopes J Virol 1998 72 9955 9965 9811733
Tanaka Y Zeng L Shiraki H Shida H Tozawa H Identification of a neutralization epitope on the envelope gp46 antigen of human T cell leukemia virus type I and induction of neutralizing antibody by peptide immunization J Immunol 1991 147 354 360 1711082
Tanaka Y Tanaka R Terada E Koyanagi Y Miyano-Kurosaki N Yamamoto N Baba E Nakamura M Shida H Induction of antibody responses that neutralize human T-cell leukemia virus type I infection in vitro and in vivo by peptide immunization J Virol 1994 68 6323 6331 8083972
Londos-Gagliardi D Jauvin V Armengaut MH Astier-Gin T Goetz M Huet S Guillemain BJ Influence of amino acid substitutions on antigenicity of immunodominant regions of the HTLV type I envelope surface gylcoprotein: a study using monoclonal antibodies raised against relevant peptides AIDS Res Hum Retroviruses 1999 15 909 920 10408728 10.1089/088922299310629
Sagara Y Inoue Y Shiraki H Jinno A Hoshino H Maeda Y Identification and mapping of functional domains on human T-cell lymphotropic virus type 1 envelope proteins by using synthetic peptides J Virol 1996 70 1564 1569 8627675
Delamarre L Pique C Pham D Tursz T Dokhelar MC Identification of functional regions in the human T-cell leukemia virus type I SU glycoprotein J Virol 1994 68 3544 3549 8189493
Delamarre L Rosenberg AR Pique C Pham D Callebaut I Dokhelar MC The HTLV-I envelope glycoproteins: structure and functions J Acquir Immune Defic Syndr Hum Retrovirol 1996 13 S85 91 8797709 10.1097/00042560-199600001-00015
Jassal SR Pohler RG Brighty DW Human T-cell leukemia virus type 1 receptor expression among syncytium-resistant cell lines revealed by a novel surface glycoprotein-immunoadhesin J Virol 2001 75 8317 8328 11483777 10.1128/JVI.75.17.8317-8328.2001
Jinno A Haraguchi Y Shiraki H Hoshino H Inhibition of cell-free human T-cell leukemia virus type 1 infection at a postbinding step by the synthetic peptide derived from an ectodomain of the gp21 transmembrane glycoprotein J Virol 1999 73 9683 9689 10516085
Chung M Kizhatil K Albritton LM Gaulton GN Induction of syncytia by neuropathogenic murine leukemia viruses depends on receptor density, host cell determinants, and the intrinsic fusion potential of envelope protein J Virol 1999 73 9377 9385 10516046
Siess DC Kozak SL Kabat D Exceptional fusogenicity of Chinese hamster ovary cells with murine retroviruses suggests roles for cellular factor(s) and receptor clusters in the membrane fusion process J Virol 1996 70 3432 3439 8648675
Kim FJ Manel N Boublik Y Battini JL Sitbon M Human T-cell leukemia virus type 1 envelope-mediated syncytium formation can be activated in resistant Mammalian cell lines by a carboxy-terminal truncation of the envelope cytoplasmic domain J Virol 2003 77 963 969 12502812 10.1128/JVI.77.2.963-969.2003
Ott D Rein A Basis for receptor specificity of nonecotropic murine leukemia virus surface glycoprotein gp70SU J Virol 1992 66 4632 4638 1321266
Manel N Kinet S Battini JL Kim FJ Taylor N Sitbon M The HTLV receptor is an early T-cell activation marker whose expression requires de novo protein synthesis Blood 2003 101 1913 1918 12393496 10.1182/blood-2002-09-2681
Sommerfelt MA Weiss RA Receptor interference groups of 20 retroviruses plating on human cells Virology 1990 176 58 69 1691887 10.1016/0042-6822(90)90230-O
Rein A Interference grouping of murine leukemia viruses: a distinct receptor for the MCF-recombinant viruses in mouse cells Virology 1982 120 251 257 6285602 10.1016/0042-6822(82)90024-1
Rein A Schultz A Different recombinant murine leukemia viruses use different cell surface receptors Virology 1984 136 144 152 6330990 10.1016/0042-6822(84)90255-1
Chesebro B Wehrly K Different murine cell lines manifest unique patterns of interference to superinfection by murine leukemia viruses Virology 1985 141 119 129 2983494 10.1016/0042-6822(85)90188-6
Rosenberg AR Delamarre L Preira A Dokhelar MC Analysis of functional conservation in the surface and transmembrane glycoprotein subunits of human T-cell leukemia virus type 1 (HTLV-1) and HTLV-2 J Virol 1998 72 7609 7614 9696862
Delamarre L Rosenberg AR Pique C Pham D Dokhelar MC A novel human T-leukemia virus type 1 cell-to-cell transmission assay permits definition of SU glycoprotein amino acids important for infectivity J Virol 1997 71 259 266 8985345
Brighty DW Jassal SR The synthetic peptide P-197 inhibits human T-cell leukemia virus type 1 envelope-mediated syncytium formation by a mechanism that is independent of Hsc70 J Virol 2001 75 10472 10478 11581416 10.1128/JVI.75.21.10472-10478.2001
Battini JL Danos O Heard JM Definition of a 14-amino-acid peptide essential for the interaction between the murine leukemia virus amphotropic envelope glycoprotein and its receptor J Virol 1998 72 428 435 9420242
Bae Y Kingsman SM Kingsman AJ Functional dissection of the Moloney murine leukemia virus envelope protein gp70 J Virol 1997 71 2092 2099 9032341
Barnett AL Cunningham JM Receptor binding transforms the surface subunit of the mammalian C-type retrovirus envelope protein from an inhibitor to an activator of fusion J Virol 2001 75 9096 9105 11533173 10.1128/JVI.75.19.9096-9105.2001
Barnett AL Davey RA Cunningham JM Modular organization of the Friend murine leukemia virus envelope protein underlies the mechanism of infection Proc Natl Acad Sci U S A 2001 98 4113 4118 11274436 10.1073/pnas.071432398
Barnett AL Wensel DL Li W Fass D Cunningham JM Structure and mechanism of a coreceptor for infection by a pathogenic feline retrovirus J Virol 2003 77 2717 2729 12552012 10.1128/JVI.77.4.2717-2729.2003
Lavillette D Ruggieri A Russell SJ Cosset FL Activation of a cell entry pathway common to type C mammalian retroviruses by soluble envelope fragments J Virol 2000 74 295 304 10590117
Lavillette D Boson B Russell SJ Cosset FL Activation of Membrane Fusion by Murine Leukemia Viruses Is Controlled in cis or in trans by Interactions between the Receptor-Binding Domain and a Conserved Disulfide Loop of the Carboxy Terminus of the Surface Glycoprotein J Virol 2001 75 3685 3695 11264358 10.1128/JVI.75.8.3685-3695.2001
Lavillette D Ruggieri A Boson B Maurice M Cosset FL Relationship between SU subdomains that regulate the receptor-mediated transition from the native (fusion-inhibited) to the fusion-active conformation of the murine leukemia virus glycoprotein J Virol 2002 76 9673 9685 12208946 10.1128/JVI.76.19.9673-9685.2002
Anderson MM Lauring AS Burns CC Overbaugh J Identification of a cellular cofactor required for infection by feline leukemia virus Science 2000 287 1828 1830 10710311 10.1126/science.287.5459.1828
Farrell KB Ting YT Eiden MV Fusion-defective gibbon ape leukemia virus vectors can be rescued by homologous but not heterologous soluble envelope proteins J Virol 2002 76 4267 4274 11932392 10.1128/JVI.76.9.4267-4274.2002
Mueckler M Hresko RC Sato M Structure, function and biosynthesis of GLUT1 Biochem Soc Trans 1997 25 951 954 9388579
Overbaugh J Miller AD Eiden MV Receptors and entry cofactors for retroviruses include single and multiple transmembrane-spanning proteins as well as newly described glycophosphatidylinositol-anchored and secreted proteins Microbiol Mol Biol Rev 2001 65 371 389 11528001 10.1128/MMBR.65.3.371-389.2001
Tailor CS Lavillette D Marin M Kabat D Cell surface receptors for gammaretroviruses Curr Top Microbiol Immunol 2003 281 29 106 12932075
Malik HS Henikoff S Eickbush TH Poised for contagion: evolutionary origins of the infectious abilities of invertebrate retroviruses Genome Res 2000 10 1307 1318 10984449 10.1101/gr.145000
Rohrmann GF Karplus PA Relatedness of baculovirus and gypsy retrotransposon envelope proteins BMC Evol Biol 2001 1 1 11244578 10.1186/1471-2148-1-1
Pearson MN Rohrmann GF Transfer, Incorporation, and Substitution of Envelope Fusion Proteins among Members of the Baculoviridae, Orthomyxoviridae, and Metaviridae (Insect Retrovirus) Families J Virol 2002 76 5301 5304 11991958 10.1128/JVI.76.11.5301-5304.2002
Virgintino D Robertson D Benagiano V Errede M Bertossi M Ambrosi G Roncali L Immunogold cytochemistry of the blood-brain barrier glucose transporter GLUT1 and endogenous albumin in the developing human brain Brain Res Dev Brain Res 2000 123 95 101 11020555 10.1016/S0165-3806(00)00086-9
Rathmell JC Vander Heiden MG Harris MH Frauwirth KA Thompson CB In the absence of extrinsic signals, nutrient utilization by lymphocytes is insufficient to maintain either cell size or viability Mol Cell 2000 6 683 692 11030347 10.1016/S1097-2765(00)00066-6
Smith TA Facilitative glucose transporter expression in human cancer tissue Br J Biomed Sci 1999 56 285 292 10795374
Younes M Lechago LV Somoano JR Mosharaf M Lechago J Wide expression of the human erythrocyte glucose transporter Glut1 in human cancers Cancer Res 1996 56 1164 1167 8640778
Magistretti PJ Pellerin L [Functional brain imaging: role metabolic coupling between astrocytes and neurons] Rev Med Suisse Romande 2000 120 739 742 11094539
Denesvre C Sonigo P Corbin A Ellerbrok H Sitbon M Influence of transmembrane domains on the fusogenic abilities of human and murine leukemia retrovirus envelopes J Virol 1995 69 4149 4157 7769674
Battini JL Rasko JE Miller AD A human cell-surface receptor for xenotropic and polytropic murine leukemia viruses: possible role in G protein-coupled signal transduction Proc Natl Acad Sci U S A 1999 96 1385 1390 9990033 10.1073/pnas.96.4.1385
Dragic T Charneau P Clavel F Alizon M Complementation of murine cells for human immunodeficiency virus envelope/CD4-mediated fusion in human/murine heterokaryons J Virol 1992 66 4794 4802 1629956
Naviaux RK Costanzi E Haas M Verma IM The pCL vector system: rapid production of helper-free, high-titer, recombinant retroviruses J Virol 1996 70 5701 5705 8764092
Ragheb JA Anderson WF pH-independent murine leukemia virus ecotropic envelope-mediated cell fusion: implications for the role of the R peptide and p12E TM in viral entry J Virol 1994 68 3220 3231 8151784
Rein A Mirro J Haynes JG Ernst SM Nagashima K Function of the cytoplasmic domain of a retroviral transmembrane protein: p15E-p2E cleavage activates the membrane fusion capability of the murine leukemia virus Env protein J Virol 1994 68 1773 1781 8107239
| 15575958 | PMC539286 | CC BY | 2021-01-04 16:36:37 | no | Retrovirology. 2004 Dec 2; 1:41 | utf-8 | Retrovirology | 2,004 | 10.1186/1742-4690-1-41 | oa_comm |
==== Front
Nutr JNutrition Journal1475-2891BioMed Central London 1475-2891-3-221558829110.1186/1475-2891-3-22ResearchSoy versus whey protein bars: Effects on exercise training impact on lean body mass and antioxidant status Brown Erin C [email protected] Robert A [email protected] Ari [email protected] Steven T [email protected] Department of Sport & Exercise Sciences, The Ohio State University, Columbus, Ohio, USA2 Department of Human Nutrition, The Ohio State University, Columbus, Ohio, USA3 DrSoy Inc., Irvine, California, USA2004 8 12 2004 3 22 22 26 8 2004 8 12 2004 Copyright © 2004 Brown et al; licensee BioMed Central Ltd.2004Brown 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
Although soy protein may have many health benefits derived from its associated antioxidants, many male exercisers avoid soy protein. This is due partly to a popular, but untested notion that in males, soy is inferior to whey in promoting muscle weight gain. This study provided a direct comparison between a soy product and a whey product.
Methods
Lean body mass gain was examined in males from a university weight training class given daily servings of micronutrient-fortified protein bars containing soy or whey protein (33 g protein/day, 9 weeks, n = 9 for each protein treatment group). Training used workouts with fairly low repetition numbers per set. A control group from the class (N = 9) did the training, but did not consume either type protein bar.
Results
Both the soy and whey treatment groups showed a gain in lean body mass, but the training-only group did not. The whey and training only groups, but not the soy group, showed a potentially deleterious post-training effect on two antioxidant-related related parameters.
Conclusions
Soy and whey protein bar products both promoted exercise training-induced lean body mass gain, but the soy had the added benefit of preserving two aspects of antioxidant function.
==== Body
Background
Many male exercisers avoid soy protein because there is a perception that it is inferior to proteins like whey for supporting lean boss mass gain. This perception persists even though there are no studies comparing whey and soy for effects on lean body mass gain. Soy may actually help promote lean body mass gain by the antioxidants associated with soy protein. Antioxidants are agents, either consumed in the diet or made by the body, which work against molecular damage due to oxidant reactions caused by free radicals, which are reactive molecules with an unpaired electron [1]. Soy protein isolate contains a mixture of antioxidants including isoflavones, saponins, and copper, a component of a number of antioxidant enzymes [2]. Body free radical production seems to be particularly high during exercise, and the resulting oxidant stress appears to contribute to muscle damage and fatigue [3]. This damage and fatigue could conceivably limit progress in exercise training by slowing muscle recovery between exercise workouts. This could limit lean body mass gain during an exercise program.
If soy protein can promote lean body mass gain at least as well as whey, there may be one advantage to consuming soy protein. Soy protein contains antioxidants which may not only help with lean body mass gain, but which can also promote other aspects of health. Antioxidant actions are thought to work against the onset and severity of many diseases and health problems [1]. This may be particularly important during exercise training, which in some cases, depletes antioxidant capacities and/or increases oxidant stress [i.e. [4,5]]. This may explain why high degrees of chronic exercise can be detrimental. For example, some athletes show increases in histochemical muscle lesions as well as high cancer mortality, which have been linked to prolonged periods of exercise [6,7]. However, this area has been controversial since some studies suggest that long term exercise training produce body adaptations which increase antioxidant defenses [i.e. [8,9]]. Either way, soy protein antioxidants could conceivably exert beneficial effects during exercise training, either by restricting antioxidant depletion or by enhancing antioxidant capacity increases.
The present study compared a soy protein product to a whey protein product in subjects undergoing a 9 week weight training program. Subjects were evaluated for lean body mass gain and changes in antioxidant status. The latter was done using one measurement of a component of antioxidant capacity and one for a component of oxidant stress. The former was based on an assay called plasma antioxidant status which assesses the ability to scavenge certain chemically generated radicals. The oxidant stress parameter was plasma myeloperoxidase, a measure of neutrophil activation, which is associated with increased secretion of superoxide radical [1].
Methods
Subjects
This study was approved by the Human Subjects Review Committee for Biomedical Sciences at The Ohio State University. All subjects signed an informed consent form. Male subjects, aged 19–25, were recruited from the Sport, Fitness and Health Program courses at The Ohio State University to participate in the present 9-week study. All subjects were considered experienced weightlifters with at least 1 year or more experience in strength training, which was confirmed by a questionnaire. Subjects were reported to be non-smokers, non-vegetarians, not currently taking supplements of any kind, and having no major health problems (i.e., diabetes, cardiovascular disease, etc.). All subjects had a body mass index (BMI) of less than 30.
Strength Training Program
At the start of the study, each subject was put on a common strength training program to strictly follow for the duration of the 9 week study. Subjects were given either workout 1 or workout 2. The two workouts were identical with the exception of exercise order and were designed to prevent subjects in the strength training classes from having to perform the same exercises at the same time. Midway through the program, subjects with workout 1 were given workout 2 and vice versa in order to maintain consistency.
The strength training protocol was 3 sets of 4–6 repetitions for 14 exercises so that strength was the variable being maximized. The following exercises were performed to work all major muscle groups: 1) chest press; 2) chest fly; 3) incline press; 4) lat pull-down; 5) seated row; 6) military press; 7) lateral raise; 8) preacher curl; 9) bicep curl; 10) supine tricep extension; 11) seated tricep extension; 12) leg press; 13) calf raise; and 14) abdominal crunches.
Protein Treatments
Subjects were randomly assigned in a double-blind manner to either a soy, whey, or control group. The controls did the exercise program but did not consume a protein product (n = 9/each group). The soy protein product was DrSoy® Bars, which contained 11 grams of protein and an assortment of micronutrients. The whey bars were made using the same recipe as the DrSoy® Bars except that whey protein was substituted for soy protein. The products were supplied to study personnel in plain wrappers with different colors for each product. The color code was unknown to the subjects and study personnel who were in contact with the subjects. Each subject was instructed to consume 3 bars per day for the 9-week training period. This was in addition to the subjects' self-selected diet. Subjects were instructed not to change eating patterns during the course of the study. The time of the day when the bars were consumed was recorded daily in the subject's fitness log so that compliance could be monitored.
Measurements
Lean body mass was analyzed by hydrostatic weighing. Each subject performed at least 3 efforts and an average reading was taken. Blood was drawn into heparin tubes before and after the 9 week treatment period on a day when the subjects did not exercise. Blood was spun at 3000 × g and the plasma was stored at -70°C until analysis. Unfortunately, a problem during blood processing made some plasma samples unavailable for analysis. Plasma was analyzed for free radical scavenging capacity using the Total Antioxidant Status Assay Kit from Calbiochem-Novachem Corp. (San Diego, CA). Plasma myeloperoxidase was analyzed using an ELISA kit from Calbiochem-Novachem.
Statistical analysis
Statistical analysis was done by the Jump 3.1 program (SAS Institute, Cary, NC), with significance at p < 0.05. For each parameter and treatment group, values prior to the 9 week treatment were compared to values after treatment by paired, 2-tailed Student's t-test. In addition, for lean body mass, the changes in values for soy treatment were compared to the change in values for the other two groups by Tukey test.
Results
Baseline subject characteristics are given in Table 1. Exercise training plus soy or whey treatments each produced a statistically significant increase in lean body mass, but the training alone did not (Figure 1). A comparison of the change in lean body mass for the soy group versus the change in the whey group did not show a significant difference (Figure 2). Plasma radical scavenging capacities fell in the whey and training alone groups, while the myeloperoxidase values rose in those same two groups (Figures 3 and 4). The values were unchanged in the soy group (Figures 3 and 4).
Table 1 Subject characteristics.
WHEY SOY CONTROL (Training Alone)
AGE 20.36 ± 0.34 21.67 ± 0.24 20.44 ± 0.63
HEIGHT (cm) 180 ± 1.55 179 ± 1.30 178 ± 1.81
WEIGHT (kg) 81 ± 2.81 79 ± 2.49 79 ± 0.48
LBM (kg) 67 ± 1.96 66 ± 2.30 67 ± 1.65
Values are means ± SEM.
Figure 1 Lean body mass pre- and post-treatment. Values are % lean body mass (kg) ± SEM from 9 subjects per group. *Significantly different from pre-treatment values (paired t-test, p < 0.05)
Figure 2 Percent change lean body mass. Values are % change in lean body mass ± SEM. *Different letters indicate significantly differences between groups (Tukey test, p < 0.05)
Figure 3 Plasma antioxidant status. Values are mM of trolox equivalents ± SEM (N = 5 for control and whey, 8 for soy) *Significantly different from pre-treatment values (paired t-test, p < 0.05)
Figure 4 Plasma myeloperoxidase. Values are mg/L ± SEM (N = 5 for control and whey, 8 for soy) *Significantly different from pre-treatment values (paired t-test, p < 0.05) **Significantly different from pre-treatment values (paired t-test, p < 0.01)
Discussion
In this study, soy and whey were both effective at increasing lean body mass with exercise training, but the soy had the added advantage of inhibiting two negative effects of training on antioxidant status. The percent change in the radical scavenging capacity (total antioxidant status) seen with training alone and training plus whey was substantial compared to the differences typically seen for these types of measurements[11-13].
The lean body mass data seen here contradicts the common, but unconfirmed notion that soy is inferior to whey for promoting lean body mass gain. It should be noted, however, that the general trend for this study may or may not be duplicated for other study designs. For example, the time frame used here, 9 weeks, is not overly long for seeing lean body mass gain, which may explain why the training alone did not produce an effect on lean body mass gain. Thus, the effects of soy or whey on lean body mass gain versus training alone may be more pronounced than in longer studies. It should also be noted that the training program used here emphasized low exercise repetitions in subjects not used to this type of training. In addition, this study included only subjects that were still relatively early in their training experience, and placed no restriction on Calorie intake. These design considerations were geared toward gaining bulk and power. The effects of whey or soy on lean body mass might be different in a design that emphasizes higher repetitions or Calorie restriction in other types of subjects. In addition, it can be noted that the current study diet intervention used bars which included added micronutrients. Thus, this study did not determine if the effects of the soy or whey protein required co-administration of micronutrients.
It is not known whether the negative effects of training seen here for antioxidant status in the whey plus training alone groups would continue upon longer training. The current state of knowledge concerning exercise training effects on antioxidant defenses does not present a clear pattern [i.e. [4,5,8,9]], possibly because of the highly variable circumstances involved in different studies such as training intensity, types of exercise done, types of antioxidant measures used, fitness level of the subjects, length of training, and dietary patterns of the subjects. These variables may help explain why some studies find training-induced declines in antioxidant defense while others find no change or even an increase. Nonetheless, the present study suggests that soy protein intake can promote antioxidant function during training which could be helpful no matter what the effects of training by itself.
Another unresolved issue is whether the effects on lean body mass seen here for the two proteins were due to increased total protein intake or other factors. In regard to the former, the data regarding the amount and type of protein intake necessary to produce optimal strength training gains is conflicting. While a diet meeting the current RDA for protein intake (0.8 g/kg body mass) may be sufficient for the sedentary individual, recent studies suggest dietary protein exceeding that of the RDA is needed for muscle hypertrophy [14,15]. One of the difficulties in deriving an exact protein recommendation for exercisers is that total energy intake has not been consistent in the studies. In some studies, total energy intake was low, which can cause an abnormally high percentage of energy output to be derived from protein [15,16]. In the present study, a 3 day diet record gave no indication that Calorie intake was low (data not shown).
If soy and whey promotion of lean body mass gain was not due to increased total protein intake, which remains uncertain, then other factors were responsible. In the case of soy protein, there are associated antioxidants [2]. As presented in the Introduction, this could conceivably help indirectly with lean body mass gain. In the case of whey, the content of essential amino acids, especially those with sulfur, may be conducive to promoting lean body mass gain [i.e. [17,18]].
In summary, soy and whey protein bars both supported lean body mass gain in conjunction with a short term power-based weight training program, but only the soy bar prevented a training-induced drop in antioxidant capacities.
Competing interests
Author AB owns the company that produces the soy bars used in the study.
Authors' contributions
ECB planned and carried out specifics of the intervention. RAD conceived the general aims of the study and chose the blood measurements. AB invented the protein bars and planned specifics of the nutrition intervention. STD planned the general aspects of the exercise intervention.
Acknowledgements
The authors thank Joshua Selsby and Kristi Seifker for excellent technical assistance.
==== Refs
Kehrer J Free radicals as mediators of tissue injury and disease Crit Rev Toxicol 1993 23 21 48 8471159
DiSilvestro RA Antioxidant actions of soya Food Indust J 2001 4 210 220
Clarkson PM Antioxidants and physical performance Crit Rev Food Sci Nutr 1995 35 131 141 7748472
Schippinger G Wonisch W Abuja PM Fankhauser F Winklhofer-Roob BM Halwachs G Lipid peroxidation and antioxidant status in professional American football players during competition Eur J Clin Invest 2002 32 686 692 12486869 10.1046/j.1365-2362.2002.01021.x
Bergholm R Makimattila S Valkonen M Liu ML Lahdenpera S Taskinen MR Sovijarvi A Malmberg P Yki-Jarvinen H Intense physical training decreases circulating antioxidants and endothelium-dependent vasodilatation in vivo Atherosclerosis 1999 145 341 349 10488962 10.1016/S0021-9150(99)00089-1
Karlsson J Antioxidants and Exercise 1997 Champaign IL: Human Kinetics
Polednak AP College athletes, body size, and cancer mortality Cancer 1976 38 382 387 947530
Selamoglu S Turgay F Kayatekin BM Gonenc S Yslegen C Aerobic and anaerobic training effects on the antioxidant enzymes of the blood Acta Physiol Hung 2000 87 267 273 11428751
Robertson JD Maughan RJ Duthie GG Morrice PC Increased blood antioxidant systems of runners in response to training load Clin Sci 1991 80 611 618 1647925
Grisham MB Jones HP Chow CC Superoxide and inflammation Cellular Antioxidant Defense Mechanisms 1988 3 Boca Raton: CRC Press 123 142
DiSilvestro RA Blostein-Fujii A Watts B Low phytonutrient, semipurified liquid diets depress plasma total antioxidant status in renal dialysis patients Nutr Res 1999 19 1173 1177 10.1016/S0271-5317(99)00078-0
Rossi AL Blostein-Fujii A DiSilvestro RA Soy beverage consumption by young men: increased plasma total antioxidant status and decreased acute, exercise-induced muscle damage J Nutraceuticals Funct Med Foods 2000 3 33 44 10.1300/J133v03n01_03
Dasgupta A Decreased total antioxidant capacity and elevated lipid hydroperoxide concentrations in sera of epileptic patients receiving phenytoin Life Sci 1997 61 437 443 9244370 10.1016/S0024-3205(97)00401-3
Lemon PWR Is increased dietary protein necessary or beneficial for individuals with a physically active lifestyle? Nutr Rev 1996 54 S169 S175 8700446
Lemon PWR Proctor DN Protein intake and athletic performance Sports Med 1991 12 313 325 1763249
Welle S Matthews DE Campbell RG Nair KS Stimulation of protein turnover by carbohydrate overfeeding in men Am J Physiol 1989 257 E413 E417 2782403
Walzem RL Dillard CJ German JB Whey components: millennia of evolution create functionalities for mammalian nutrition: what we know and what we may be overlooking Crit Rev Food Sci Nutr 2002 42 353 375 12180777 10.1080/10408690290825574
Lands LC Grey VL Smountas AA Effect of supplementation with a cysteine donor on muscular performance J Appl Physiol 1999 87 1381 1385 10517767
| 15588291 | PMC539287 | CC BY | 2021-01-04 16:39:29 | no | Nutr J. 2004 Dec 8; 3:22 | utf-8 | Nutr J | 2,004 | 10.1186/1475-2891-3-22 | oa_comm |
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-3-411554832310.1186/1475-925X-3-41ResearchThreshold intensity factors as lower boundaries for crack propagation in ceramics Marx Rudolf [email protected] Franz [email protected] Per-Ole [email protected] Department of Prosthetic Dentistry, Section of Dental Materials, University Hospital of the University of Technology, 52074 Aachen, Germany2004 17 11 2004 3 41 41 22 9 2004 17 11 2004 Copyright © 2004 Marx et al; licensee BioMed Central Ltd.2004Marx 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
Slow crack growth can be described in a v (crack velocity) versus KI (stress intensity factor) diagram. Slow crack growth in ceramics is attributed to corrosion assisted stress at the crack tip or at any pre-existing defect in the ceramic. The combined effect of high stresses at the crack tip and the presence of water or body fluid molecules (reducing surface energy at the crack tip) induces crack propagation, which eventually may result in fatigue. The presence of a threshold in the stress intensity factor, below which no crack propagation occurs, has been the subject of important research in the last years. The higher this threshold, the higher the reliability of the ceramic, and consequently the longer its lifetime.
Methods
We utilize the Irwin K-field displacement relation to deduce crack tip stress intensity factors from the near crack tip profile. Cracks are initiated by indentation impressions. The threshold stress intensity factor is determined as the time limit of the tip stress intensity when the residual stresses have (nearly) disappeared.
Results
We determined the threshold stress intensity factors for most of the all ceramic materials presently important for dental restorations in Europe. Of special significance is the finding that alumina ceramic has a threshold limit nearly identical with that of zirconia.
Conclusion
The intention of the present paper is to stress the point that the threshold stress intensity factor represents a more intrinsic property for a given ceramic material than the widely used toughness (bend strength or fracture toughness), which refers only to fast crack growth. Considering two ceramics with identical threshold limits, although with different critical stress intensity limits, means that both ceramics have identical starting points for slow crack growth. Fast catastrophic crack growth leading to spontaneous fatigue, however, is different. This growth starts later in those ceramic materials that have larger critical stress intensity factors.
==== Body
Background
Slow crack growth is most suitably described in a v (crack velocity) versus KI (stress intensity factor) diagram. Slow crack growth in ceramics is attributed to corrosion assisted stress at crack tips or at any defect pre-existing in the ceramic [1]. The combined presence of body fluid molecules (mainly water), which reduce the surface energy at the crack tip, and the presence of high stresses are the reasons for subcritical crack growth (SCCG) in ceramics.
The presence of stress intensities above a critical value (KI > KIc) initiates fast catastrophic crack growth, followed by the deterioration of a dental or a body restoration machined from ceramics. The presence of stress intensities above a threshold value (KI > KI0) initiates SCCG in ceramics, followed by a slow, however continuous, erosion of the strength of a restoration which also may result in final fatigue. In an early stage of ceramic research it was believed that this lower limit for SCCG is very close to zero. In the mean time, however, one has learned that for most ceramic materials the lower limit for SCCG is significantly larger than zero. Indeed, it may even be just below KIc.
The threshold limit KI0 corresponds to a crack equilibrium at null crack velocity. Therefore, it allows a safety range of clinical use. The higher the value of KI0, the higher the reliability, and hence the lifetime of a restoration. Bio-components should be designed to work in a region of the v-KI-diagram where the upper border line of that region corresponds to the threshold limit.
In the present paper we preferentially focus on those ceramics that are important in dental research. Note, however, that alumina and zirconia have meaning in both fields of application (dentistry and medicine). We use soda lime glass as a well characterized standard and silicon nitride as important in the general field of ceramics.
There are several methods available and in the literature extensively described how the threshold limit can be measured. The feasibility of these measurement procedures is mostly demonstrated with the help of soda lime glass as a brittle solid model.
In principle, the proper test for existence of a threshold lies in the observation of reversibility of crack growth. The threshold can be regarded as a Griffith quiescent point, where forward and backward fluctuations just balance, i.e., the mean velocity of the crack tip becomes zero. The forward and backward fluctuations take place over discrete energy barriers definable as G = W = 2γ, where G is the energy release rate, W is the Dupré work of adhesion, and γ is the surface energy. If G <W the crack should retract and heal; otherwise it should repropagate [2]. On the basis of this assumption, the authors in [2] (see also [3]) calculate equations prescribing the v G characteristics (crack velocity versus mechanical energy release rate; equivalent to v - KI crack velocity versus stress intensity factor) at specified chemical concentrations and temperatures, which can describe observed v-G dependencies.
One common experimental method to determine the threshold limit is to measure slow crack growth rate down to velocities as low as 10-14 m/s. Then one can extrapolate from the vertical branch of the function to the zero velocity limit on the stress intensity factor axis KI, with the intersection KI equal to KI0 [4-7].
Another method to determine the aforesaid threshold limit is the "interrupted static fatigue test" (ISF-test) [8]. For a bending experiment, the applied stress is chosen such that a significant fraction of samples fails in a "hold period". Samples that do not fail during this static phase are then fractured by the usual four point bending technique. The threshold is calculated either from the applied stress intensity factor at which 50% of samples fail during the stress hold, or by using the factor applied to the weakest specimen during the stress hold as calculated for various hold times. Once the value of the stress intensity factor becomes independent of hold time, it is equivalent to the threshold [9].
Another method uses a side grooved specimen with a crack propagating along its length, and under a bending condition similar to four point bending. The crack velocity can be obtained from the rate of load relaxation at constant displacement and the initial crack length. Having established the v - K diagram, the threshold is determined as described above. For further details refer to [10].
Other methods may be characterized by the phrase "decay of residual stress" [11]. Here, the threshold limit can be calculated from the residual stress factor attributed to this decay of residual stress.
The current method of measurement used, however, is based on indentation cracking, analogous to other studies also utilizing flaw initiation for starting the test [11-13]. After this start, however, the subsequent procedure is different. A follow up of the decay of residual stress intensities near the crack tip is done over a period of about one year, determining Ktip via the COD for different times after indentation [14].
Methods
Using a micro-hardness testing machine, a Vickers indentation is made on the carefully polished surface of a sample of the ceramic to be investigated. Radial cracks emanate from each of the four indentated corner sources.
To determine the stress intensity present at the crack tip due to the indentation, the near crack tip profile is determined using a scanning microscope (ESEM: "Environmental Scanning Electron Microscope"). A specific feature of this technique is that it is carried out at a moderate vacuum (p ≈ 10-1 mbar). Hence, there is no longer need to sputter the samples with a gold or carbon layer. Our initial attempts to measure the crack opening displacement (COD) showed that sputtering resulted in blurring the crack banks or even partly hiding the crack. Thus, we abandoned those attempts and started again when the ESEM was available. Before the availability of the ESEM it was nearly impossible to precisely measure crack profiles at submicrometer resolution which, however, is mandatory.
Images of the crack profiles (Fig. 1) were digitally stored and analyzed by imaging software (Paint Shop Pro, V. 6, Jasc Software, Eden Prairie, Maine, USA).
Figure 1 Example of a Vickers indentation. Only one of four corners is shown (length of diagonal 115 μm). With the crack tip as a starting point (x = 0) the crack width 2*u(x) is measured at the distance x (COD after Irwin [15]; ceramic material for this example: Empress 1). The residual tensions cause crack growth over a long time interval until, at the end of the crack, Ktip is equal to KI0. Crack tip shielding by secondary effects (micro structural elements which toughen material as the crack extends) may slightly distort results (measured KI0 then lower than true KI0). Insert: idealized COD.
The measured profiles can be attributed to the crack opening displacement (COD) near the crack tip [14]. The near crack tip profiles for stress-free crack surfaces are usually represented by the Irwin K-field displacement relation [15], with 2u being the total COD, x the distance from the crack tip, and the plane strain Young's modulus E' = E/(1-v2);(v = Poisson's constant) being.
We assume that there is no crack shielding. Then, in equilibrium, the currently acting crack tip stress intensity factor Ktip is balanced by the toughness of the material KIc (mode I loading [15]):
and by re-arrangement:
If data are taken sufficiently close to the crack tip (x ≤ 20 μm), a linear relationship is experimentally observed between u(x)2 and x. Ktip can then be calculated from a regression analysis as the slope of a straight line, provided E' is known (see below).
The residual stresses close to the crack tip initiated by the indentation impression gradually decay over time t, and one anticipates that they slowly fade away eventually approaching zero. Hence Ktip= Ktip(t) and it is plausible to assume Ktip(t→∞) ≈ KI0. Therefore, in the present work, because of slow crack growth, we take the threshold value of the stress intensity factor as the time limit of the slowly decreasing Ktip value. Provided that a suitable high resolution scanning microscope is at hand, there is no need of sputtering the samples, and the presently utilized method is very simple. A potential shortcoming, however, is that this method may need many months or even years until the residual stresses are relaxed and the threshold value is reached.
The authors concede that they have chosen to consider a somewhat ideal situation since the assumption Ktip(t→∞) = KI0 assumes ideal behavior. In real ceramics, especially polycrystalline and composite materials, the crack tip may be shielded from residual load by micro structural elements, which toughen the material in the region just before the crack tip [2]. This behavior is reminiscent to R-curve behavior.
We carried out ESEM analyses of crack profiles after 1 hour and then after up to 420 days, at 5 dates distributed over the whole time interval (Fig. 3). After indentation and between two measurements the samples were stored at normal lab environmental conditions (21°C, 65 % humidity).
We determined the threshold stress intensity of the following ceramics (Soda lime glass and veneering ceramics as reference): Al2O3, coarse grained, load of indention 9,9 kg, Young's modulus 350 GPa (Frialit-Degussit, Mannheim/Ludwigshafen, Germany), Cerec Mark II, 4 kg, 69 GPa, HiCeram, 6,9 kg, 107 GPa, VMK 95, 4 kg, 91 GPa (all three Vita, Bad Säckingen, Germany), Cercon Base, 7,9 kg, 210 GPa, CergoGold, 4 kg, 70 GPa (both Degudent-Dentsply, Hanau, Germany), Dicor, 2 kg, 74 GPa (Corning Glass Works, Corning, USA), Empress 1, 5,9 kg, 67 GPa, Empress 2, 5,9 kg, 96 GPa (both Ivoclar, Schaan, Liechtenstein), Lava, 8 kg, 210 GPa (3M-Espe, Seefeld, Germany), Soda lime glass, 2 kg, 73 GPa (Saint Gobain, Aachen, Germany), Si3N4, 6 kg, 289 GPa and hipped 5%Y2O3-Zirkon, 8 kg, 210 GPa.
The constitution of the soda lime glass was SiO2 72.65, Al2O3 0.28, MgO 3.98, CaO 8.84, Na2O 13.79, K2O 0.19, other 0.27.
Results
As examples, Fig. 1 shows a crack starting at the corner of a Vickers indentation (right hand) and Fig. 2 shows a plot representing data for "Cerec Mark II" two days after indentation, as a function of distance from crack tip x (2 μm <x < 23 μm), analyzed with the help of Eq. 1'. A linear relationship is observed, from which the value of Ktip (t = 48 h) = 0,90 MPa√m was easily and precisely deduced. Fig. 3 shows all Ktip values determined in an analogous manner for nine examples out of the thirteen investigated ceramics. The gradual decrease of Ktip(t) due to decaying stress intensities at the crack tip becomes apparent. The manner in which Ktip(t) decreases suggests an exponential relationship, as the decrease appears to be linear on a logarithmic scale. The truncation of the measurements after about 104 hours (for reasons of feasibility) appears somewhat arbitrarily, and it cannot be excluded that a further decay, although very small, may have been missed. Note that due to the apparent exponential relationship, the overestimation of the threshold value KI0 due to the truncation after 104 hours becomes smaller and smaller with time. We plan to do further measurements after another interval of 104 hours (417 days). Considering the mathematical aspect, 105 hours (11+ years) would make more sense; but such a long interval is obviously not practicable. As already mentioned, this time constraint is a decided disadvantage of our current method to determine the threshold value.
Figure 2 Regression analysis representing data for "Cerec Mark II" two days after indentation, analyzed with the help of Eq. 1 (u(x)2 as a function of distance from crack tip x (2 μm <x < 23 μm)). A linear relationship is observed.
Being aware of the these limitations, and having in mind the neglected possible crack tip shielding as discussed above, we identify KI0= Ktip(t→∞). Fig. 4 displays all KI0 values in comparison with their KIc counterparts.
Figure 3 Ktip(t) values of nine out of the thirteen ceramics investigated. The gradual decrease of Ktip(t) with time due to decaying stress intensity at the crack tip becomes apparent.
Figure 4 KI0 threshold values (hatched columns) in comparison with their counterpart critical stress intensities, KIc (unfilled columns). Refer also to [22]. In the available literature, values for reference: Al2O3 (KI0 = 2.5 ± 0.2 MPa√m); ZrO2 (KI0 = 3.1 ± 0.2 MPa√m, both values after [4]); Soda lime glass (KI0 = 0,42 MPa√m, after [11]).
Discussion
KIc is the lower limit for (fast) catastrophic crack growth. Stress intensities exceeding this limit cause fast crack growth at supersonic velocity, and eventually result in destruction of ceramic components. This kind of destruction, however, is not the most common or important, since it can be avoided by strictly limiting the stress intensities existing throughout a component by a suitable shape of construction.
KI0 is the upper limit of stress intensities for absence of crack growth and the lower limit for (slow) subcritical crack growth (SCCG). Limiting stress intensities such that they stay always below KI0 means infinite life time for a component, since SCCG becomes irrelevant. Hence, the most favorable characteristic stress intensity values are obvious: KIc as high as possible and KI0 as close as possible to KIc. Such a selection minimizes the extension of the interval in which subcritical crack growth can take place, and it maximizes resistance to catastrophic crack growth due to overloading. Fig. 5 gives a ranking of all ceramics currently tested, based on threshold values related to the corresponding critical values KI0/KIc. Favorable ceramics within their class of toughness are situated at the right hand side of the chart. Note, however, that a perfect ceramic material dependent on the focused area of application has not only a favorable (threshold/critical) stress strength relationship but also a high KIc value.
Figure 5 Ranking of all ceramics as imposed by their ratio "threshold value to critical value" (KI0/KIc). Dicor: see [23].
At first glance zirconia may seem to be a ceramic material superior to alumina, since it has a critical stress intensity factor (Fig. 4: 9.4 ± 1.5 MPa·√m) which is about three times larger than this of alumina. Values in the literature for zirconia are up to about 8 MPa·√m [16], compared with 5.4 MPa·√m and [5]: 5.0 ± 0.2 MPa·√m [17] for alumina. Naturally, this is a significant advantage when operations near the critical stress of a material are involved. However, in practical applications, stresses having an intermediate level are more common, thus initiating SCCG instead of catastrophic crack growth. Then, if the threshold stress intensities of two ceramics are equal, they are both subject to SCCG at the same rate. Apparently. zirconia vs alumina is an example for such a situation (Fig. 4): meaning that both ceramics have equal potential for SCCG. The different behavior of these ceramics is solely rendered to stress bearing capabilities near catastrophic crack growth. At such stresses near KIc zirconia, of course, has properties superior to alumina.
It becomes apparent that at moderate stresses alumina and zirconia may be equally suitable choices, and other criteria may become important for favoring the one or the other material. Such reasons may be the ease of shaping, questions of color, ease of veneering, esthetic considerations, availability, and other circumstances.
There is one other aspect to be considered when comparing zirconia and alumina. The exponents n of SCCG of both ceramics are high (in principle meaning slow SCCG), and the answer to the question of which of the materials has the larger exponent depends on whether static or cyclic behavior is addressed: nstatic = 39 vs 104 and ncyclic = 28 vs 16 for Al2O3 and Y-PSZ, respectively [16]. These parameters show that lifetimes are shortened and crack growth rates are significantly accelerated by cyclic loading compared to static loading.
Zirconia is known to be sensitive to humidity, which is a particular important issue when prosthetic and orthopedic applications are considered. It is known that yttria stabilized zirconia ceramics can be destabilized during the process of steam sterilization. This is due to hydrothermal transformation, resulting in surface roughening of the zirconia ceramic femoral heads. These femoral heads may also undergo slow degradation during long term implantation in the human body. This low temperature degradation does not become significant before several years, but it does raise the question of the use of zirconia for load bearing systems [4]. In conclusion, it can be stated that SCCG of Y-TZP is activated by the influence of water [18,19], however, there is some controversy about this effect [20]. An analogous statement holds for MgO-partially stabilized zirconia (PSZ) [21].
Note that concerning the sensitivity to humidity, there is a notable difference between ceramics for dental or for orthopedic applications. Ceramics for dental applications are often veneered by a different ceramic, which means that there is a protective shield against humidity attacking from outside of the ceramic tooth (but not from inside or from the marginal region).
Fig. 6 displays an example of a zirconia ceramic material developed for dental applications and which was formerly used. The sensitivity to humidity becomes apparent.
Figure 6 Example (linear Weibull plot) for a zirconia based ceramic material developed for dental applications. Samples handled at 60 % relative humidity (lab environmental conditions; diamonds) vs samples stored in aqua dest for 10 days (triangles). The sensitivity to humidity is obvious. The bending strength due to water storage decreases from σ63% = 1,346 MPa to 1,003 MPa (about 25 %).
There are some other examples of ceramics for which a large difference in the critical stress intensities is observed whereas the threshold values are very similar. For these ceramics an analogous argument holds, as given above for alumina vs zirconia. From Fig. 4, for Empress 1 or Empress 2 (e.g.) the following values are measured: KIc = 1.17 ± 0.08 MPa·√m or KIc = 2.48 ± 0.22 MPa·√m, respectively; and KI0 = 0.83 ± 0.16 MPa·√m or KI0 = 0.94 ± 0.12 MPa·√m, respectively. Again, the critical stress intensity values are largely different, the threshold values, however, are nearly identical. Compare also "Al2O3" with "Lava" and "Cercon".
Authors' contributions
RM conceived in the study, designed the study and drafted the manuscript. POW and FJ carried out the experimental work. All authors read and approved the final manuscript. All authors contributed equally to this work.
Acknowledgements
We thankfully acknowledge the proposal of Prof. Rödel, Darmstadt to utilize the COD method for determination of threshold values.
==== Refs
Lawn B Fracture of brittle solids 1991 Cambridge: University Press
Wan KT Lathabai S Lawn BR Crack Velocity Functions and Thresholds in Brittle Solids J Eur Ceram Soc 1990 6 259 268 10.1016/0955-2219(90)90053-I
Cook RF Influence of crack velocity thresholds on stabilized nonequilibrium fracture J Appl Phys 1989 65 1902 1910 10.1063/1.342902
De Aza AH Chevalier J Fantozzi G Schehl M Torrecillas R Crack growth resistance of alumina, zirkonia and zirkonia toughened alumina ceramics for joint prostheses Biomaterials 2002 23 937 945 11774853 10.1016/S0142-9612(01)00206-X
Drouin JM Cales B Chevalier J Fantozzi G Fatigue behaviour of zirconia hip joint heads – experimental results and finite element analysis J Biomed Mater Research 1997 34 149 155 9029293 10.1002/(SICI)1097-4636(199702)34:2<149::AID-JBM2>3.0.CO;2-Q
Kocer C Collins RE Measurement of Very Slow Crack Growth in Glass J Am Ceram Soc 2001 84 2585 2593
Krell A Pippel E Woltersdorf J Burger W Subcritical crack growth in Al2O3 with submicron grain size J Eur Ceram Soc 2003 23 81 89 10.1016/S0955-2219(02)00072-9
Sglavo VM Green DJ Threshold Stress Intensity Factor in Soda-Lime Silicate Glass by Interrupted Static Fatigue Test J Eur Ceram Soc 1996 16 645 651 10.1016/0955-2219(95)00176-X
Sglavo VM Green DJ The interrupted static fatigue test for evaluating threshold stress intensity factor in ceramic materials: A numerical analysis J Eur Ceram Soc 1995 15 777 785 10.1016/0955-2219(95)00048-Y
Evans AG A method for evaluating the time-dependent failure characteristics of brittle materials – and its application to polycrystalline alumina J of material science 1972 7 1137 1146 10.1007/BF00550196
Salomonson J Zeng K Rowcliffe D Decay of residual stress at indentation cracks during slow crack growth in soda-lime glass Acta mater 1996 44 543 546 10.1016/1359-6454(95)00199-9
Cook RF Liniger EG Kinetics of Indentation Cracking in Glass J Am Ceram Soc 1993 76 1096 1105
Sglavo VM Green DJ Indentation Determination of Fatigue Limits in Silicate Glasses J Am Ceram Soc 1999 82 1269 1274
Seidel J Rödel J Measurement of Crack Tip Toughness in Alumina as a Function of Grain Size J Am Ceram Soc 1997 80 433 438
Irwin GR Fracture In Handbuch der Physik 1958 6 Berlin: Springer
Munz D Fett T Ceramics – Mechanical Properties, Failure Behaviour, Materials Selection 1999 Berlin: Springer
Rhee YW Kim HW Deng Y Lawn BR Brittle Fracture versus Quasi Plasticity in Ceramics: A simple Predictive Index J Am Ceram Soc 2001 84 561 565
Chevalier J Olagnon C Fantozzi G Calès B Crack propagation behavior of Y-TZP Ceramics J Am Ceram Soc 1995 78 1889 1894
Dauskardt RH Yu W Ritchie RO Fatigue Crack Propagation in Transformation-Toughened Zirconia Ceramic J Am Ceram Soc 1987 70 C-248 C-252
Shimizu K Oka M Kumar P Kotoura Y Yamamuro T Makinouchi K Nakamura T Time-dependent changes in the mechanical properties of zirconia ceramic J Biomed Mater Res 1993 27 729 734 8408102
Becher PF Subcritical crack growth in partially stabilized ZrO2 (MgO) J of Mat Science 1986 21 297 300
Marx R Fischer H Weber M Jungwirth F Rissparameter und Weibullmodule: unterkritisches Risswachstum und Langzeitfestigkeit vollkeramischer Materialien Dtsch Zahnärztl Z 2001 56 90 98
Fischer J Krämer V Kappert HF Qualitative Untersuchungen zur Oberflächen-Kristallisation dentaler Glaskeramik (Dicor) Dtsch Zahnärztl Z 1989 44 891 893
| 15548323 | PMC539288 | CC BY | 2021-01-04 16:37:32 | no | Biomed Eng Online. 2004 Nov 17; 3:41 | utf-8 | Biomed Eng Online | 2,004 | 10.1186/1475-925X-3-41 | oa_comm |
==== Front
Cardiovasc UltrasoundCardiovascular Ultrasound1476-7120BioMed Central London 1476-7120-2-261558142810.1186/1476-7120-2-26ResearchTransthoracic coronary flow reserve and dobutamine derived myocardial function: a 6-month evaluation after successful coronary angioplasty Cicala Silvana [email protected] Maurizio [email protected] Pasquale [email protected]'Errico Arcangelo [email protected] Pasquale [email protected] Moira [email protected] Giancarlo [email protected] Divitiis Oreste [email protected] Cardioangiology Unit, Department of Clinical and Experimental Medicine, Federico II University Hospital Naples, Italy2 Division of Cardiology, "Villa dei Fiori" Hospital Naples, Italy2004 6 12 2004 2 26 26 4 11 2004 6 12 2004 Copyright © 2004 Cicala et al; licensee BioMed Central Ltd.2004Cicala 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 percutaneous transluminal coronary angioplasty (PTCA), stress-echocardiography and gated single photon emission computerized tomography (g-SPECT) are usually performed but both tools have technical limitations. The present study evaluated results of PTCA of left anterior descending artery (LAD) six months after PTCA, by combining transthoracic Doppler coronary flow reserve (CFR) and color Tissue Doppler (C-TD) dobutamine stress.
Six months after PTCA of LAD, 24 men, free of angiographic evidence of restenosis, underwent standard Doppler-echocardiography, transthoracic CFR of distal LAD (hyperemic to basal diastolic coronary flow ratio) and C-TD at rest and during dobutamine stress to quantify myocardial systolic (Sm) and diastolic (Em and Am, Em/Am ratio) peak velocities in middle posterior septum. Patients with myocardial infarction, coronary stenosis of non-LAD territory and heart failure were excluded. According to dipyridamole g-SPECT, 13 patients had normal perfusion and 11 with perfusion defects. The 2 groups were comparable for age, wall motion score index (WMSI) and C-TD at rest. However, patients with perfusion defects had lower CFR (2.11 ± 0.4 versus 2.87 ± 0.6, p < 0.002) and septal Sm at high-dose dobutamine (p < 0.01), with higher WMSI (p < 0.05) and stress-echo positivity of LAD territory in 5/11 patients. In the overall population, CFR was related negatively to high-dobutamine WMSI (r = -0.50, p < 0.01) and positively to high-dobutamine Sm of middle septum (r = 0.55, p < 0.005).
In conclusion, even in absence of epicardial coronary restenosis, stress perfusion imaging reflects a physiologic impairment in coronary microcirculation function whose magnitude is associated with the degree of regional functional impairment detectable by C-TD.
Percutaneous coronary angioplastyCoronary flow reserveColor Tissue DopplerStress-echo
==== Body
Introduction
Percutaneous transluminal coronary angioplasty (PTCA) has deeply modified the effective management of coronary artery disease [1]. Coronary artery restenosis is unfrequent when PTCA is associated to coronary stenting application which is able to enlarge the lumen area stenosis [2,3]. However, even in absence of coronary artery restenosis, the results of revascularization can be suboptimal because of coronary microvessel dysfunction subsequent to the procedure [4,5]. This issue may be intriguing for management of patients undergone PTCA.
The non-invasive assessment after PTCA is usually performed by gated single photon emission computerized tomography (g-SPECT) [6,7] and by stress echocardiography [7,8]. However, both these tools present technical limitations, g-SPECT having a low specificity [9] and semi-quantitative echocardiographic wall motion analysis low sensitivity [10].
In the last years, great interest has been developed about new echocardiographic techniques as Doppler-derived coronary flow reserve (CFR) [11,12] and color Tissue Doppler (C-TD) [13,14]. The first tool provides reliable information about coronary microvascular function in absence of epicardial coronary stenosis [15] while C-TD is able to quantify left ventricular (LV) myocardial performance both at rest and during pharmacological stress [13,14].
On these grounds, aim of the present study was to assess C-TD derived myocardial performance, both at rest and during pharmacologic stress, in relation to the function of coronary microcirculation determined by non invasive CFR after successful PTCA of LAD.
Methods
Study population
Among 30 patients who had undergone PTCA with stenting for significant LAD stenosis between September and October 2000, 24 patients (age = 50–64 years) free of coronary angiographic evidence of LAD restenosis 6 months after the procedure, entered the study and performed non-invasive test screening in the same period of the invasive assessment (±7 days). The informed consent of all patients and approval of Institutional Committee were obtained. Patients were excluded for acute and previous myocardial infarction (according to ECG at rest), concomitant coronary stenosis of right coronary artery and/or circumflex artery, congestive heart failure, valvular heart disease, primary cardiomyopathy, atrial fibrillation, inadequate quality echocardiograms. On the basis of g-SPECT dipyridamole induced perfusion defects, the study population was divided into 2 groups: without and with perfusion defects.
Procedures
Patients underwent dipyridamole gated myocardial perfusion g-SPECT acquisition, transthoracic echocardiography, C-TD (both at baseline and during dobutamine stress) and non-invasive CFR determination by dipyridamole test. All echocardiographic measurements were analyzed without knowledge of the clinical data. According to the rules of the Institutional Committees, all patients withdrew cardiac drugs at least 2 days before the performance of the non invasive assessment.
Dipyridamole g-SPECT
Single day rest/dipyridamole g-SPECT was performed according to the standard methods by injecting patients with technetium 99m (99mTc) tetrophosmin 8 mCi (296 MBq) at rest and 24 mCi (888 MBq) after dipyridamole infusion by volumetric pump (dose of 0.14 mg/kg/min in 4 minutes) [16,17]. A single stress SPECT corresponds to a dose exposure of about 500 chest x-ray. Qualitative assessment of reconstructed gated images was obtained on mid-short axis slices, vertical and horizontal long axis slices.
Transthoracic Echocardiography
Standard echocardiographic examinations were performed using a System FiVe, Vingmed Sound AB machine (GE, Horten, Norway), by a 2.5 MHz transducer equipped with second harmonic capability. M-mode echocardiographic analysis was performed according to the criteria of the American Society of Echocardiography [18] and LV mass indexed for height powered to 2.5 [19]. LV end-diastolic and end-systolic volumes were estimated according to the Simpson method [20] and LV ejection fraction derived.
Stress protocols
Dobutamine stress protocol was performed according to the standard method [21] using low and high-dose (up to 40 μg/Kg/min) by using the System FiVe Vingmed machine. C-TD of posterior septum was recorded at rest and during each dobutamine stage. CFR assessment was performed by HDI 5000 ultrasound machine (ATL Ultrasound, Bothell, Washington, USA), using a high-frequency (7 MHz) transducer. The visualization of the distal portion of the left anterior descending artery and the recording of PW-Doppler derived coronary blood flow velocities performed at baseline and after dipyridamole infusion (0.56 mg/kg over 4 minutes) according to the standards of our laboratory [14]. Blood pressure and a 12-lead ECG were recorded at rest and at the end of each stage of both dobutamine and dipyridamole tests.
C-TD dobutamine stress echocardiography analysis
Echocardiographic images were recorded on S-VHS videotapes and digitally stored on magneto-optical disk for subsequent analysis. Images were evaluated by 2 experienced observers. Baseline and stress wall motion analysis was performed by 2 experienced readers blinded to the other data. Regional wall motion was assessed with a 16-segment model of the left ventricle and semiquantitatively graded from 1 to 4 as follows: 1 = normal; 2 = hypokinesia; 3 = akinesia; and 4 = dyskinesia. A wall motion score index (WMSI), obtained dividing the sum of each segment scores by the number of the segments, was assessed both at baseline and at high-dose dobutamine.
C-TD acquisition of posterior septal wall was performed in real time, superimposed on 2-D images, at baseline and at the end of each dobutamine infusion stage. C-TD imaging was stored in digital format and analyzed off line on cine-loop as previously described [14]. The region of interest was the middle segment of the posterior septum where myocardial velocity profile was obtained. The middle posterior septum for measurements of C-TD was chosen since the perfusion of this myocardial segment is provided by a branch of LAD, where also CFR was determined. The reproducibility of C-TD of our laboratory has been reported, the intra- and inter-observer variability being <3% and <6% for all the measurements both at rest and at high-dose dobutamine [22].
CFR Analysis
Methods and reproducibility (intra-observer and inter-observer variability 1.9% and 4.2% respectively) of our laboratory in measuring coronary blood flow reserve has been described [22]. By placing sample volume on the color signal, spectral Doppler of LAD flow showed the characteristic biphasic flow pattern with a larger diastolic and a smaller systolic component. Diastolic peak velocities were measured at baseline and after dipyridamole, by averaging the highest 3 spectral Doppler signals for each measurement. CFR was defined as the ratio of hyperemic to basal diastolic peak velocities. All images were recorded on a magneto-optical disk and analyzed off-line by 2 independent observers, blinded to the other data.
Statistical Analysis
The analyses were performed by SPSS for Windows release 8.0 (Chicago, Illinois, USA). Data are presented as mean value ± SD. Analysis of variance was used to assess intergroup differences. Linear regression analyses and partial correlation test was done using Pearson's method. Differences were considered significant at p < 0.05.
Results
Characteristics of the study population
The characteristics of the study population and both heart rate and blood pressure at baseline and at high-dose dobutamine are listed in Table 1. The 2 groups were comparable for heart rate and blood pressure values both at rest and at stress dobutamine peak. Of note, the prevalence of arterial hypertension, diabetes mellitus, hypercholesterolemia and smoke was not different between groups and no patients of both groups presented g-SPECT derived myocardial perfusion defects at rest (data not reported in Table).
Table 1 Characteristics of the study population
Variable Normal Perfusion n = 13 Perfusion Defect n = 11 P
Age (years) 55.9 ± 4.1 58.4 ± 3.1 NS
Body mass index (Kg/m2) 26.1 ± 1.1 26.3 ± 0.8 NS
Baseline Systolic BP (mm Hg) 147.0 ± 7.5 149.2 ± 11.6 NS
Baseline Diastolic BP (mm Hg) 85.1 ± 7.5 85.5 ± 9.1 NS
Baseline Heart rate (bpm) 74.8 ± 5.9 73.7 ± 6.9 NS
DOB Systolic BP (mm Hg) 151.1 ± 7.5 151.9 ± 10.2 NS
DOB Diastolic BP (mm Hg) 83.2 ± 6.3 83.4 ± 7.3 NS
DOB Heart rate (bpm) 139.6 ± 5.4 141.0 ± 5.5 NS
BP = Blood Pressure, DOB = Dobutamine
Echocardiographic analysis
The comparisons of echocardiographic measurements and CFR between the 2 groups are reported in Table 2. Because of higher septal and posterior wall thickness, patients with myocardial perfusion defects after PTCA had greater LV mass index (p < 0.05). LV ejection fraction was comparable between the 2 groups.
Table 2 Standard Doppler echocardiographic and CFR analysis
Variable Normal Perfusion Perfusion Defect P
Septal wall thickness (mm) 10.1 ± 1.4 11.2 ± 0.4 <0.02
Posterior wall thickness (mm) 10.2 ± 1.4 10.6 ± 0.5 NS
LV internal diastolic diameter (mm) 54.7 ± 2.6 56.7 ± 3.5 NS
LV internal systolic diameter (mm) 39.2 ± 2.8 39.6 ± 2.9 NS
2-D Ejection Fraction (%) 54.8 ± 6.0 54.9 ± 3.4 NS
LV mass index (g/m 2.7) 49.6 ± 10.7 57.9 ± 8.0 <0.05
LV = left ventricular
Dobutamine test and Color TD analysis
WMSI was comparable between the 2 groups at baseline (1.07 ± 0.10 versus 1.15 ± 0.11) whereas it was higher at low-dose dobutamine (1.07 ± 0.11 versus 1.17 ± 0.12) and at high-dose dobutamine (1.07 ± 0.12 versus 1.20 ± 0.14) (both p < 0.05) in patients with SPECT-derived perfusion defects than in controls. Positive dobutamine stress-echo involving LAD territory (and in particular mid-septal region) was observed in 5/11 patients (45.4%) with SPECT perfusion defects.
C-TD diastolic measurements of mid-septum (Em, Am, Em/Am ratio) were similar between the two groups at rest (Em/Am ratio = 1.04 ± 0.1 and 1.03 ± 0.3 in patients with and without perfusion defects respectively, NS) and at low dose dobutamine (Em/Am ratio = 1.00 ± 0.1 and 1.13 ± 0.4 respectively, NS) while Em/Am ratio was mildly different at high-dose dobutamine (0.83 ± 0.2 and 0.70 ± 0.2 respectively, p < 0.05). Sm peak velocities were lower in patients with perfusion defects at low- (p < 0.05) and at high-dose dobutamine (p < 0.01) and were significantly lower also in patients with perfusion defects showing stress induced wall motion abnormalities in comparison with patients with perfusion defects but no change of wall motion during dobutamine infusion (Figure 1).
Figure 1 In the left panel comparison of Sm peak velocity of middle posterior septum of patients without and with SPECT perfusion defects at rest, at low and at high-dose dobutamine. In the right panel comparison of Sm peak velocity of middle posterior septum during dobutamine stress echocardiography in patients with perfusion defects having or not dobutamine-induced wall motion abnormalities.
CFR analysis
Coronary diastolic peak velocities were similar at rest between the two groups (21.5 ± 5.2 cm/s in patients without perfusion defects and 22.0 ± 19 cm/s in patients with perfusion defects, NS) but significantly different after dipyridamole (60.0 ± 16.1 cm/s versus 49.1 ± 10.8 cm/s respectively, p < 0.05). Thus, CFR was 2.87 ± 0.6 in patients without defects and 2.11 ± 0.4 in patients with perfusion defects (p < 0.002). Of note, analyzing the group with SPECT-derived perfusion defects, the patients with stress inducible wall motion abnormalities had lower CFR (1.91 ± 0.1) than those without change of WMSI during dobutamine infusion (2.28 ± 0.4) (p = 0.06).
Relationship between CFR and Dobutamine stress measurements
In the overall population, CFR was negatively related to WMSI at low-dose (r = -0.46, p < 0.02) and high-dose dobutamine (r = -0.50, p < 0.01). Among C-TD Doppler measurements, CFR was positively related to Sm peak velocity of middle septum at low dose (r = 0.39, p < 0.05) and high-dose dobutamine (r = 0.55, p < 0.0005) (Figure 2) while the relation of Sm at baseline (r = 0.12) did not achieve the statistical significance. No relation of CFR was found with C-TD diastolic measurements of middle septum at any stage of dobutamine stress. Figure 3 shows a patient with SPECT derived normal perfusion: CFR is >2 and middle septal Sm peak velocity has a significant increment from baseline to high-dose dobutamine (Δ = +9). Figure 4 displays a patient with a perfusion defect: CFR is reduced and middle septal Sm increase from baseline to high-dose dobutamine is lower (Δ = +5).
Figure 2 Positive association between CFR and C-TD derived Sm peak velocity of middle septum at high-dose dobutamine. Full circles indicate patients with SPECT-derived myocardial perfusion defects; empty circles indicates patients without perfusion defects.
Figure 3 CFR and Sm peak velocity of middle septum at high-dose dobutamine in a patient with SPECT derived normal perfusion. The upper panels show coronary artery flow velocity in the LAD at baseline and with a normal increase with dipyridamole (DIP). In the lower panels, myocardial systolic velocity (Sm) shows a normal increase at high-dose dobutamine.
Figure 4 CFR and septal Sm peak velocity at high-dose dobutamine in a patient with SPECT derived perfusion defect. The upper panels display a reduced CFR. In the lower panels, the increase of septal Sm from baseline to high-dose dobutamine is low.
Discussion
The present study used new ultrasound tools, as off-line quantitative C-TD [13,14] and Doppler-derived CFR [15], to evaluate long-term effects of PTCA on the relation between myocardial performance and coronary microvascular function, in the absence of angiographic coronary artery restenosis. According to dipyridamole g-SPECT, the population was divided into 2 groups, the first without and the second one with perfusion defect in the LAD territory, comparable for resting LV ejection fraction. Our findings show that, six months after successful PTCA of LAD, patients with myocardial perfusion defects present both lower CFR and reduced peak dobutamine myocardial systolic function of wall involved by LAD perfusion (i.e., middle posterior septum) in comparison with the control group and that CFR is positively related to stress peak Sm velocity measured at middle septum in the overall population.
CFR and perfusion defect after PTCA
According to the study design, we intentionally selected patients without angiographic evidence of post-PTCA LAD restenosis. Worthy of note, the incidence of coronary restenosis was very low 6 months after the procedure, in accord with previous experiences about the combined use of stenting and PTCA [23]. Nevertheless, 11 our patients without restenosis showed dipyridamole g-SPECT LAD perfusion defects. Myocardial hypoperfusion may occur in patients without overt epicardial coronary artery stenosis having coronary microvessel damage [24,25] and coronary microvascular dysfunction corresponds to a reduced CFR in the absence of epicardial coronary stenosis [26]. Accordingly, the patients of the present study with long-standing SPECT perfusion defects showed lower CFR than the control group. An abnormal CFR had been already described immediately after balloon angioplasty [5,27,28], probably because of a slow recovery of autoregulation in the microvascular bed [29]. This reduction is primarily due to an increased flow velocity at rest [5,27,28], in relation to the failure of microvessel bed to vasoconstrict appropriately and/or to epicardial vasoconstriction mediated by a myogenic response and/or neural mechanism [30]. In contrast to previous studies showing normalization of CFR after three [31], five [32] or six months [5], CFR was persistently reduced in our patients with SPECT-derived perfusion defects. A suboptimal Doppler flow wire derived CFR had been observed six months after PTCA without restenosis by DEBATE investigators [33]. In the suboptimal CFR group the reduction of CFR was mainly due to a long-standing elevation in resting peak velocities while in our patients with perfusion defects it was due to a blunted maximal vasodilator response to dypiridamole. It is conceivable that this alteration could depend on endothelial damage of coronary microcirculation [34] preceding the procedure and persisting long time after PTCA. Coronary microcirculatory vasoconstriction induced by endothelial dysfunction has been described as effect of spontaneous myocardial ischemia [35] as well as in conditions other than epicardial coronary artery stenosis, as diabetes mellitus [36,37] arterial hypertension [26,38] and LV hypertrophy [39,40], which can alter microvascular function. However, an alternative interpretation of our findings include the possibility that a residual coronary stenosis might be anatomically insignificant but hemodynamically important, thus explaining a discrepancy between the percentage of lumen reduction and the amount of regional flow reserve.
Myocardial systolic function and perfusion defect after PTCA
The reduction of myocardial systolic function expressed by the decrease of low and high-dose dobutamine Sm in middle septum, i.e. in the territory supplied by LAD, is not surprising in patients with perfusion defects after PTCA. Of note, myocardial systolic performance of middle septum was not significantly different between the 2 groups at rest. These findings are consistent with an altered myocardial systolic velocity response to exercise already described by C-DT in patients with coronary artery disease [13,41]. Also WMSI, not different at rest, became significantly higher at low and high-dose dobutamine in patients with perfusion defects. This increase (involving LV segments of LAD territory) during stress occurred only in 5/11 patients who had lower Sm peak velocities at peak dobutamine stress and lower CFR than patients without inducible wall motion abnormalities. Myocardial reperfusion injury may include LV regional systolic dysfunction as irreversible manifestation, it depending by a reduction of myocardial blood flow reserve [5]. Inducible wall motion abnormalities in the presence of a successful coronary revascularization might indicate a very severe microvascular damage.
Association between CFR and myocardial systolic function
It is recognized that the extent of stress dobutamine-induced dissinergy is associated to the degree of CFR reduction in patients with significant coronary artery stenosis, an invasive myocardial fractional flow reserve ≤0.75 having a sensitivity of 76% and a specificity of 97% [42]. In agreement with these findings, we found a positive association between the functional degree of vasodilator microvascular coronary circulation and the magnitude of regional myocardial systolic function at peak dobutamine stress, i.e. Sm peak velocity of the wall (middle septum) supplied by LAD. Accordingly, we also found a lower but significant negative relation between CFR and stress peak WMSI. Since patients undergoing successful reperfusion procedures generally present a good stress-echo LV functional response [6], our data suggest the ability of C-DT to detect even minor forms of LV regional myocardial dysfunction occurring under these circumstances.
Study Limitations
The main limitation of the present study include the fact that the negativity of coronary angiography can not exclude definitely the presence of coronary restenosis while an invasive measurement of CFR by Doppler flow wire or an intra-coronary ultrasound could have been crucial to clarify this issue. Unfortunately, these evaluations were not included into our study protocol. In addition, it has also to be underscored that patients with perfusion defects of the present study had greater LV mass, a factor which can itself induce a reduction of CFR [39,40].
Clinical implications
The results of the present study suggest that CFR impairment may be detectable after PTCA even in the absence of coronary restenosis, it depending by an altered coronary microvascular function. In this clinical scenario, SPECT stress perfusion defects have to be interpreted as false positive results for PCTA restenosis while they reflect a true physiologic impairment in regional CFR with some associated degree of systolic impairment detectable by C-TD. Quantitative parameters of CFR and C-TD can provide an additive value over conventional stress echocardiographic assessment.
Competing interests
Financial competing interests
• In the past five years we didn't receive reimbursements, fees, funding, or salary from an organization that may in any way gain or lose financially from the publication of this manuscript, either now or in the future.
• No organization financied this manuscript.
• We didn't hold any stocks or shares in an organization that may in any way gain or lose financially from the publication of this manuscript, either now or in the future.
• We didn't hold or are currently applying for any patents relating to the content of the manuscript We didn't receive reimbursements, fees, funding, or salary from an organization that holds or has applied for patents relating to the content of the manuscript.
• We don't have any other financial competing interests.
Non-financial competing interests
There are not any non-financial competing interests (political, personal, religious, academic, intellectual, commercial or any other) to declare in relation to this manuscript.
We have not a competing interest, please discuss it with the editorial office.
==== Refs
RITA-2 trial participants Coronary angioplasty versus medical therapy for angina: the second Randomised Intervention Treatment of Angina (RITA-2) trial Lancet 1997 350 461 468 9274581 10.1016/S0140-6736(97)07298-X
Erbel R Haude M Hopp HW Franzen D Rupprecght HJ Heunlein B Coronary-artery stenting compared with balloon angioplasty in patients with coronary artery disease: Benestent Study Group N Engl J Med 1994 331 489 495 8041413 10.1056/NEJM199408253310801
Topol EJ Coronary artery stents: gauging, gorging and gouging N Engl J Med 1998 339 1702 1704 9834311 10.1056/NEJM199812033392311
Hansen PR Myocardial reperfusion injury: experimental evidence and clinical relevance Eur Heart J 1995 16 734 740 7588916
Van Liebergen RAM Pick JJ Koch KT de Winter RJ Lie KI Immediate and long-term effect of balloon angioplasty or stenting implantation on the absolute and relative coronary blood flow velocity reserve Circulation 1998 98 2133 2140 9815867
Bax JJ Wijns W Cornel JH Visser FC Boersma E Fioretti PM Accuracy of the currently available techniques for prediction of functional recovery after revascularization in patients with left ventricular dysfunction due to chronic coronary artery disease: comparison of pooled data J Am Coll Cardiol 1997 30 1451 1460 9362401 10.1016/S0735-1097(97)00352-5
Takeuchi M Miura Y Toyokawa T Araki M Nakashima Y Kuroiwa A The comparative diagnostic value of dobutamine stress echocardiography and thallium stress tomography for detecting restenosis after coronary angioplasty J Am Soc Echocardiogr 1995 8 696 702 9417213
Heinle SK Lieberman EB Ancukiewicz M Waugh RA Bashore TM Kisslo J Usefulness of dobutamine echocardiography for detecting restenosis after percutaneous transluminal coronary angioplasty Am J Cardiol 1993 72 1220 1225 8256695 10.1016/0002-9149(93)90287-M
Beygui F Le Feuvre C Maunoury C Helft G Antonietti T Metzger JP Vacheron A Detection of coronary restenosis by exercise electrocardiography thallium-201 perfusion imaging and coronary angiography in asymptomatic patients after percutaneous transluminal coronary angioplasty Am J Cardiol 2000 86 35 40 10867089 10.1016/S0002-9149(00)00825-0
Picano E Bedetti G Varga A Cseh E The comparable diagnostic accuracies of dobutamine-stress and dipyridamole-stress echocardiographies : a meta-analysis Coron Artery Dis 2000 11 151 159 10758817 10.1097/00019501-200003000-00010
Caiati C Montaldo C Zedda N Bina A Iliceto S A new non-invasive method for coronary flow reserve assessment: contrast-enhanced transthoracic second harmonic echo Doppler Circulation 1999 99 771 778 9989962
Hozumi T Yoshida K Ogata Y Akasaka T Asami Y Morioka S Noninvasive assessment of significant left anterior descending coronary artery coronary stenosis by coronary flow velocity reserve with transthoracic color Doppler echocardiography Circulation 1998 97 1557 1562 9593560
Pasquet A Armstrong G Rimmerman C Marwick TH Correlation of myocardial Doppler velocity response to exercise with independent evidence of myocardial ischemia by dual-isotope single-photon emission computed tomography Am J Cardiol 2000 85 536 542 11078263 10.1016/S0002-9149(99)00807-3
Galderisi M Cicala S De Simone L Caso P Petrocelli A Pietropaolo L Celentano A Mininni N de Divitiis O Impact of myocardial diastolic dysfunction on coronary flow reserve in hypertensive patients with left ventricular hypertrophy Ital Heart J 2001 2 677 684 11666096
Wilson RF Johnson MR Marcus ML Aylward PE Skorton DJ Collins S White CW The effect of coronary angioplasty on coronary flow reserve Circulation 1988 77 873 885 2964950
De Puey EG Parmett S Ghesani M Rozanski A Nichols K Salensky H Comparison of Tc-99m sestamibi and Tl-201 gated perfusion SPECT J Nucl Cardiol 1999 6 278 285 10385183 10.1016/S1071-3581(99)90040-5
Bavelaar-Croon CD Pauwels EK van der Wass EE Gated single-photon emission computed tomographic myocardial imaging: a new tool in clinical cardiology Am Heart J 2001 141 383 390 11231435 10.1067/mhj.2001.112780
Sahn DJ De Maria A Kisslo J Weyman A for the Committee of M-Mode standardization of the American Society of Echocardiography Recommendations regarding quantitation in M-Mode echocardiography: results of a survey of echocardiographic measurements Circulation 1978 58 1072 1083 709763
de Simone G Daniels SR Devereux RB Meyer RA Roman MJ de Divitiis O Alderman MH Left ventricular mass and body size in normotensive children and adults: assessment of allometric relations and impact of overweight J Am Coll Cardiol 1992 20 1251 1260 1401629
Schiller NB Shah PM Crawdford M De Maria A Devereux RB Feigenbaum H for the American Society of Echocardiography Committee on Standards, Subcommittee on Quantitation of Two-dimensional Echocardiograms Recommendations for quantitation of the left ventricle by two-dimensional echocardiography J Am Soc Echocardiogr 1989 2 358 367 2698218
Armstrong WF Pellikka PA Ryan T Crouse L Zoghbi WA Stress echocardiography: recommendations for performance and interpretation of stress echocardiography. Stress Echocardiography Task Force of the Nomenclature and Standards Committee of the American Society of Echocardiography J Am Soc Echocardiogr 1998 11 97 104 9487482
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
Cannon RO Restenosis after angioplasty N Engl J Med 2002 346 1182 1183 11961145 10.1056/NEJM200204183461602
Chilian WM Layne SM Coronary microvascular responses to reductions in perfusion pressure: evidence for persistent arteriolar vasomotor tone during coronary hypoperfusion Circ Res 1990 66 1227 1238 2335023
Fragasso G Lu C Dabrowski P Pagnotta P Sheiban I Chierchia SI Comparison of stress/rest myocardial perfusion tomography, dipyridamole and dobutamine stress echocardiography for the detection of coronary disease in hypertensive patients with chest pain and positive exercise test J Am Coll Cardiol 1999 34 441 447 10440157 10.1016/S0735-1097(99)00231-4
Hasdai D Gibbons RJ Holmes DR JrHigano ST Lerman A Coronary endothelial dysfunction in humans is associated with myocardial perfusion defects Circulation 1997 96 3390 95 9396432
Nanto S Kodama K Hori M Mishima A Inoue M Kamada T Temporal increase in resting coronary blood flow causes an impairment of coronary flow reserve after balloon angioplasty Am Heart J 1992 123 28 36 1729849 10.1016/0002-8703(92)90743-F
Kern MJ Deligonul U Vandormael M Labovitz A Gudipati CV Gabliani G Bodet J Shah Y Kennedy HL Impaired coronary vasodilator reserve in the immediate post-coronary angioplasty period: analysis of coronary artery flow velocity indexes and regional cardiac venous efflux J Am Coll Cardiol 1989 13 860 872 2522463
Miller DD Verani MS Current status of myocardial perfusion imaging after percutaneous transluminal coronary angioplasty J Am Coll Cardiol 1994 24 260 266 8006276
Fischell TA Bausbach KN McDonald TV Evidence of altered epicardial coronary artery autoregulation as a cause of distal coronary vasoconstriction after successful percutaneous transluminal coronary angioplasty J Clin Investig 1990 86 575 584 2384604
Uren NG Delayed recovery of coronary resistive vessel function after coronary angioplasty J Am Coll Cardiol 1993 21 612 621 8436742
Zulstra F Reiber JC Julliere Y Serruys PW Normalization of coronary flow reserve by percutaneous transluminal coronary angioplasty Am J Cardiol 1988 61 55 60 2962486 10.1016/0002-9149(88)91304-5
Albertal M Regar E Van Langenhove G Carlier SG Serrano P Boersma E Bruyne B Di Mario C Piek J Serruys PW on behalf of the DEBATE investigators Flow velocity and predictors of suboptimal coronary flow velocity reserve after coronary balloon angioplasty Eur Heart J 2002 23 133 138 11785995 10.1053/euhj.2001.2708
Smith TP Canty JM Modulation of coronary autoregolatory responses by nitric oxide Circ Res 1993 73 232 240 8330372
Marzilli M Sambuceti G Fedele S L'Abbate A Coronary microcirculatory vasoconstriction during ischemia in patients with unstable angina J Am Coll Cardiol 2000 35 327 334 10676677 10.1016/S0735-1097(99)00554-9
Pitkanen OP Nuutila P Raitakari OT Ronnemaa T Koskinen PJ Iida H Lehtimaki TJ Laine HK Takala T Viikari JS Knuuti J Coronary flow reserve is reduced in young men with IDDM Diabetes 1998 47 248 254 9519721
Yokoyama I Ohtake T Momomura S Yonekura K Woo-Soo S Nishikawa J Sasaki Y Omata M Hyperglycemia rather than insulin resistance is related to reduced coronary flow reserve in NIDDM Diabetes 1998 47 119 124 9421384
Astarita C Palinkas A Nicolini E Maresca FS Varga A Picano E Dypiridamole-atropine stress echocardiography versus exercise SPECT scintigraphy for detection of coronay artery disease hypertensive with positive exercise test J Hypertens 2001 19 679 682 11330869 10.1097/00004872-200103000-00018
Gimelli A Schneider-Eicke J Neglia D Sambuceti G Giorgetti A Bigalli G Parodi G Pedrinelli R Parodi O Homogeneously reduced versus regionally impaired myocardial blood flow in hypertensive patients: two different patterns of myocardial perfusion associated with degree of hypertrophy J Am Coll Cardiol 1998 31 366 373 9462581 10.1016/S0735-1097(97)00503-2
Hamasaki S Al Suwaidi J Higano ST Miyauchi K Holmes DR jrLerman A Attenuated coronary flow reserve and vascular remodeling in patients with hypertension and left ventricular hypertrophy J Am Coll Cardiol 2000 35 1654 60 10807473 10.1016/S0735-1097(00)00594-5
Cain P Baglin T Case C Spicer D Short L Marwick TH Application of tissue Doppler to interpretation of dobutamine echocardiography and comparison with quantitative coronary angiography Am J Cardiol 2001 87 525 531 11230833 10.1016/S0002-9149(00)01425-9
Bartunek J Marwick TH Rodrigues AC Vincent M Van Schuerbeeck E Sys SU de Bruyne B Dobutamine-induced wall motion abnormalities: correlation with myocardial fractional flow reserve and quantitative coronary angiography J Am Coll Cardiol 1996 27 1429 1436 8626954 10.1016/0735-1097(96)00022-8
| 15581428 | PMC539289 | CC BY | 2021-01-04 16:38:29 | no | Cardiovasc Ultrasound. 2004 Dec 6; 2:26 | utf-8 | Cardiovasc Ultrasound | 2,004 | 10.1186/1476-7120-2-26 | oa_comm |
==== Front
J Immune Based Ther VaccinesJournal of Immune Based Therapies and Vaccines1476-8518BioMed Central London 1476-8518-2-91554117510.1186/1476-8518-2-9Original ResearchAntigenized antibodies expressing Vβ8.2 TCR peptides immunize against rat experimental allergic encephalomyelitis Musselli Cristina [email protected] Svetlana [email protected] Maurizio [email protected] The Department of Medicine and Cancer Center, University of California, San Diego, La Jolla CA USA2004 12 11 2004 2 9 9 24 6 2004 12 11 2004 Copyright © 2004 Musselli et al; licensee BioMed Central Ltd.2004Musselli 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
Immunity against the T cell receptor (TCR) is considered to play a central role in the regulation of experimental allergic encephalomyelitis (EAE), a model system of autoimmune disease characterized by a restricted usage of TCR genes. Methods of specific vaccination against the TCR of pathogenetic T cells have included attenuated T cells and synthetic peptides from the sequence of the TCR. These approaches have led to the concept that anti-idiotypic immunity against antigenic sites of the TCR, which are a key regulatory element in this disease.
Methods
The present study in the Lewis rat used a conventional idiotypic immunization based on antigenized antibodies expressing selected peptide sequences of the Vβ8.2 TCR (93ASSDSSNTE101 and 39DMGHGLRLIHYSYDVNSTEKG59).
Results
The study demonstrates that vaccination with antigenized antibodies markedly attenuates, and in some instances, prevents clinical EAE induced with the encephalitogenic peptide 68GSLPQKSQRSQDENPVVHF88 in complete Freunds' adjuvant (CFA). Antigenized antibodies induced an anti-idiotypic response against the Vβ8.2 TCR, which was detected by ELISA and flowcytometry. No evidence was obtained of a T cell response against the corresponding Vβ8.2 TCR peptides.
Conclusions
The results indicate that antigenized antibodies expressing conformationally-constrained TCR peptides are a simple means to induce humoral anti-idiotypic immunity against the TCR and to vaccinate against EAE. The study also suggests the possibility to target idiotypic determinants of TCR borne on pathogenetic T cells to vaccinate against disease.
EAETCRIdiotypeRegulation
==== Body
Introduction
Experimental allergic encephalomyelitis (EAE) is an experimentally induced autoimmune disease mediated by T cells. It can be induced in susceptible animals either by immunization with myelin basic protein (MBP) or proteolipid protein PLP, or by immunization with synthetic peptides from the MBP sequence [1]. EAE can also be initiated by the passive transfer of encephalitogenic, MBP-specific T cell lines or clones [2,3]. In the Lewis rat, EAE is characterized by a self limiting, ascending, hind limb paralysis. Histologically, EAE is hallmarked by perivascular and submeningeal infiltration of inflammatory cells within the brain and spinal cord [4]. After recovery, animals become refractory to further induction of paralysis by immunization with MBP. Owing to similarities in clinical expression and histopathology, EAE has long been recognized as an animal model for multiple sclerosis, a demyelinating chronic inflammatory disease in humans of unknown origin. For this reason, studies on EAE are thought to elucidate aspects of the pathogenesis and indicate possible ways of immune intervention.
EAE is mediated by MHC class II -restricted, MBP-specific CD4+ T lymphocytes bearing an antigen receptor (TCR) variable (V) regions belonging to a limited set of TCR V region gene families [5,6] and restricted Vα-Vβ gene combinations [7]. Several rational approaches have been used to prevent EAE, including passive transfer of monoclonal antibodies that interfere with the recognition of the MHC, TCR and MBP peptide complex [8,9], antibodies against CD4 [10] and T regulatory cells [11-14]. Active immunity against attenuated encephalitogenic T cells was shown to prevent the induction of disease [15,16] and vaccination with synthetic peptides of the complementarity-determining regions (CDR) of the TCR of ecephalitogenic T cells, confer resistance to EAE in the rat [17-20]. Together these facts indicated that T cells are crucial to the pathogenesis of EAE and, in converse, immunity to idiotypic determinants of the TCR of encephalitogenic T cells may be protective.
Approaches to directly target the TCR of pathogenetic T cells are an attractive direction for therapy and immunointervention as well as an opportunity to further understand the immunological events involved in protection in vivo. However, limitations exist to methods available for TCR vaccination. Vaccination using attenuated encephalitogenic T cells requires that these are specifically expanded in vitro and can only be used in an autologous system. Synthetic peptides, albeit successful in several instances [17-20], offer no tri-dimensional conformation and may even yield to opposite effect, e.g., worsening of disease [21,22]. Similarly, vaccination with single chain TCR was shown to either prevent or exacerbate EAE in mice [23].
In previous work from this laboratory we demonstrated the induction of anti-receptor immunity using immunoglobulins (Ig) expressing discrete peptide portions of human CD4 [24]. We refer to such Ig as antigenized antibodies, i.e., Ig molecules in which foreign peptide sequences are conformationally-constrained and expressed in the complementority-determining region (CDR) loops [25]. Immunization with antigenized antibodies is an efficient method to focus the immune response against defined epitopes of foreign antigens. If CDR sequences of TCRs are functionally comparable to Ig idiotypes, antigenized antibodies provide a tool to induce anti-idiotypic responses against TCR. Here, we used antibodies antigenized with TCR sequences as vaccines to control disease. We engineered two antibodies encompassing in the CDR3 of the heavy (H) chain two synthetic peptides from the sequence of rat Vβ8.2 gene product, 39DMGHGLRLIHYSYDVNSTEKG59 (CDR2) and 93ASSDSSNTE101 (CDR3, VDJ junction), both reported to confer protection against EAE in the Lewis rat [17-20] when used as vaccines. The results show that vaccination with antigenized antibodies expressing sequences of encephalitogenic T cells induces anti-idiotypic immunity against the TCR and high level resistance against EAE.
Material and Methods
Animals
Eight week old, weight-matched female Lewis rats were purchased from Charles River Laboratories (Wilmington, MA). Animals were housed (three rats per cage) in the animal facility of the Universitiy of California, San Diego. Food and water were provided at libitum.
Antigenized antibodies
The peptide sequences 93ASSDSSNTE101 and 39DMGHGLRLIHYSYDVNSTEKG59 were engineered into the CDR3 loop of the murine VH62 gene [26] according to our published methods [27]. The antigenized VH was then ligated in plasmid vector containing a human γ1 constant (C) region gene. Transfection of the plasmid DNA was performed on murine J558L cells, a H-chain defective variant of myeloma J558, carrying the rearrangement for a λ1 light (L) chain [28]. The resulting antigenized antibodies were termed γ1TCR-I and γ1TCR-II, respectively (Figure 1). Wild-type transfectoma antibodies γ1WT and γ2bWT [26] engineered to have the same C and V regions, but lacking the TCR peptides in the CDR3 of the H chain, served as controls. Transfected cells were incubated without selection for 24 hours and then selected in the presence of 1.2 mg/ml G418 (GIBCO). G418-resistant clones secreting high level of Ig were identified by enzyme-linked immunosorbent assay (ELISA) using horseradish peroxidase (HRP)-conjugated goat antibody to human Ig (Sigma) [29]. Cultures secreting 10–20 μg/ml were selected, expanded, and their supernatants precipitated by (NH4)2SO4. Antibodies were purified by affinity chromatography on a Protein A-Sepharose column (Pharmacia-LKB, Alameda, CA) equilibrated with 3 M NaCl/1M glycine, pH 8.9. Elution was performed using glycine 0.1 M- HCl/0.5 M NaCl pH 2.8. The eluted fractions were neutralized using 1 M Tris-HCl, pH 8.0, and dialyzed against 0.15 M phosphate-buffered saline (PBS) pH 7.3. The purity of the antibodies was assessed by electrophoresis on a 10% Sodium Dodecyl Sulfate (SDS)-Polyacrylamide Gel (PAGE).
Figure 1 Schematic representation of the two V regions antigenized with TCR sequences. In each case the H chain of the antigenized antibody is formed of a murine VH62 region in which the CDR3 has been engineered to express either 93ASSDSSNTE101 or 39DMGHGLRLIHYSYDVNSTEKG59 sequence between two Val-Pro (VP) doublets of the unique cloning site in the CDR3 loop of VH62. The complete H chain is the product of the fusion of the antigenized VH region with a human γ1C region. The light (L) chain (not shown) is the murine λ1 which is provided by the J558L host cell. (H chain not to scale).
Synthetic peptides
Synthetic peptide GSLPQKSQRSQDENPVVHF corresponding to amino acid residues 68–88 of guinea-pig MBP [30], DMGHGLRLIHYSYDVNSTEKG corresponding to amino acid residues 39–59 of rat Vβ8.2 (CDR2 peptide), ASSDSSNTE corresponding to amino acid residues 93–101 of rat Vβ8.2 (CDR3 peptide) rat [17,18], and the (NANP)3 peptide of Plasmodium falciparum parasite [31] were all synthesized in the Peptide Synthesis Facility of the Universitiy of California, San Diego. After synthesis peptides were analyzed by HPLC for purity. Peptide KKSIQFHWKNSNQIKILGNQGSFLTKGPS corresponding to residues 21–49 of the extracellular domain of human CD4 was described previously [32].
Enzyme-linked immunosorbent assay (ELISA)
Serum antibodies against antigenized antibodies and their control were determined by ELISA on 96-well polystyrene microtiter plates (Costar, Cambridge, MA) coated (5 μg/ml – 50 μl/well) with γ1TCR-I, γ1TCR-II, γ2bTCR-I proteins in 0.9% NaCl by drying at 37°C. The wells were blocked with a 1% bovine serum albumin (BSA) in phosphate-buffered saline (PBS), and then incubated overnight at +4°C with individual rat sera diluted in PBS containing 1% BSA and 0.05% Tween 20 (PBSA). After washing, the bound antibodies were detected by adding peroxidase-conjugated goat antibodies to rat IgG (γ specific) (Biomeda, CA) at 1:500 dilution in PBSA for 1 hour at room temperature. After washing, the bound peroxidase was measured by adding o-phenylenendiamine (100 μl/well) and H2O2. After 30 minutes, the plates were read in a micro-plate reader (Vmax, Molecular Devices) at 492 nm. Tests were done in duplicate. Antibodies to TCR peptides were detected in ELISA on 96-well polystyrene microtiter plates coated (10 μg/ml) with the Vβ8.2 synthetic peptides 39DMGHGLRLIHYSYDVNSTEKG59 and 93ASSDSSNTE101 in 0.1M carbonate buffer, pH 9.6, by overnight incubation at +4 C. After blocking unreactive sites, sera (1:25 dilution in PBSA) were added to plates and incubated overnight at +4°C. The bound antibodies and reactive peroxidase were detected as detailed above. Ig reactive with synthetic peptide 21KKSIQFHWKNSNQIKILGNQGSFLTKGPS49 of human CD4 were determined on 96-well polystyrene microtiter plates coated (2.5 μg/ml) with peptide 21–49 in 0.9% NaCl by drying at 37 C as previously established [32]. Briefly, sera (1:400 dilution in PBSA) were incubated overnight at +4°C. After washing, the test was continued as specified above. Plates were read in a micro-plate reader (Vmax, Molecular Devices) at 492 nm.
FACS analysis
Autoantibodies reactive with the Vβ8.2+ TCR were sought by flowcytometry on the S23B1E11 T cell hybridoma [33], derived from the fusion of Vβ8.2+ CD4 T lymphoblasts specific for MBP with the murine TCR α/β- BW1100.129.237 thymoma cell line [33]. For FACS analysis the following procedure was utilized. 106 hybridoma T cells in 100 μl of RPMI-1640 containing 1% egg albumin, 0.01% NaN3 and 10 mM Hepes, were incubated with rat sera (1:10 dilution) for 90 minutes at +4°C. Cells were washed three times with cold RPMI-1640 and subsequently incubated with a fluorescein-isotyocianate (FITC)-conjugated goat antibody (0.5 μg/106 cells) to rat Ig (H+L) (Caltag, So. San Francisco, CA) for 20 minutes at +4°C. After incubation, the cells were washed twice, resuspended in 1% paraformaldehyde, and analyzed in a FACS Scan (BD Biosciences). To stain for dead cells, 20 μl of propidium iodide in PBS were added to unfixed cells before FACS analysis. R-phycoerythrin conjugated mouse monoclonal antibody R78 (IgG1, k) specific for the rat Vβ8.2, the kind gift of Pharmingen (San Diego, CA), was used to control for the expression of the Vβ8.2 TCR on S23B1E11 hybridoma cells.
In vitro proliferative response
Poplyteal, inguinal and paraortic lymph nodes were removed from immunized animals at different times, dissociated and washed in RPMI-1640. Lymph node cells were plated in round-bottom 96-well plates at 2.5 × 105 cells/well in the presence of various (10–100 956;g/ml) amounts of antigen in 200 μl of RPMI containing 10% FCS, 100 U/ml penicillin, 100 μg/ml streptomycin, 4 mM glutamine, 0.1 mM non-essential aminoacids, 1 mM sodium pyruvate and 0.5 μM 2-β mercaptoethanol. Cultures were incubated for 72 hours in a 10% CO2 atmosphere. The evening before harvest 1 μCi/well of [3H]-thymidine was added to each well. Cells were harvested onto glass fiber filters and counted on an automatic Beckman LS 6000IC β-counter.
Vaccinations and immunizations schedule
Animals were vaccinated with antigenized antibodies (100 μg/rat) in complete Freunds' adjuvant (CFA) divided equally between the posterior paws (25 μl each) and two points in the back subcutaneously. A booster injection (50 μg/rat) in incomplete Freunds' adjuvant (IFA) was given subcutaneously on day 21. EAE was induced on day 28 by immunization with MBP peptide 68GSLPQKSQRSQDENPVVHF88 (30 μg/rat) in the anterior paws (25 μl each) in CFA (H37RA 10 mg/ml). Control rats were similarly injected with transfectoma antibody γ1WT or γ2bWT. Rats inoculated with Freunds' adjuvant only served as additional control. Serum samples were collected from the retro-orbital sinus on day 0 before vaccination, day 21 before booster injection, day 28 before EAE induction, and day 50 after recovery from disease. Sera were stored at -20 C until use.
Clinical evaluation of EAE
EAE was monitored daily by two operators for clinical signs using the following scale: grade 0 = no appreciable symptoms; grade 1 = tail atony; grade 2 = paraparesis; grade 3 = paraplegia; grade 4 = paraplegia with forelimb weakness, moribund state. Typically symptoms of disease began to appear on day 11–13 from the injection of the encephalitogenic peptide. The Disease Index was calculated according to the formula: [(Maximum Score) × (Duration of Disease) × (Incidence)].
Statistical Methods
Statistical analyses was performed using the Fisher's test.
Results
Vaccination with antigenized antibodies and effect on EAE
Two antigenized antibodies were engineered to express the CDR3 93ASSDSSNTE101 and CDR2 39DMGHGLRLIHYSYDVNSTEKG59 sequences, and were termed γ1TCR-I and γ1TCR-II, respectively (Figure 1). Rats were vaccinated with an individual antigenized antibody and received a booster injection 21 days later. EAE was induced on day 28 by immunization with the encephalitogenic MBP peptide 68GSLPQKSQRSQDENPVVHF88. As shown in Table 1, vaccination with both γ1TCR-I and γ1TCR-II reduced disease severity. Rats immunized with γ1TCR-I (group I) had a disease index of 1.8. Within this group, two out of six rats (33%) did not develop disease, one had grade 1 and three had grade 2. None proceeded through grade 3 or 4. Rats immunized with γ1TCR-II (group II) had a disease index of 4.9. Within this group two out of ten rats (20%) did not develop the disease, two had grade 1, four had grade 2 and two had grade 3. In contrast, all fifteen control rats vaccinated with γ1WT or given CFA only (groups III and IV) developed EAE with a disease index ranging between 11.3 and 22.4. Unmanipulated rats immunized with the MBP peptide (group V) developed EAE with a disease index of 25.2. There was a direct correlation between the severity of the disease and its duration. In rats immunized with γ1TCR-I, the disease lasted on average for 2.5 days and in rats immunized with γ1TCR-II 3.8 days. In contrast, in all the other groups (groups III-V) the duration of the disease was significantly longer (6–7 days). Of note, although group III rats had an overall lower score than unmanipulated rats, they differed from rats in group I or group II by the above mentioned parameters and these difference were statistically significant (Table 1). CFA did not confer protection. Taken together, these data indicate that active immunity elicited with antigenized antibodies expressing rat Vβ8.2 TCR peptides was effective in markedly reducing the severity of EAE in the Lewis rat.
Table 1 Vaccination against antigenized antibodies expressing TCR peptides protects from EAE
Severity of Disease*
Group No. Rats Immunogen Incidence Max Score (mean ± SD) Duration (mean ± SD) Disease Index
I 6 γ1TCR-I 4/6 1.1 ± 0.9a 2.5 ± 2.2b 1.8
II 10 γ1TCR-II 8/10 1.6 ± 1.0c 3.8 ± 2.3d 4.9
III 10 γ1WT 10/10 2.2 ± 0.9 5.1 ± 1.0 11.3
IV 5 CFA 5/5 3.4 ± 0.9 6.6 ± 1.3 22.4
V 6 - 6/6 3.5 ± 0.5 7.2 ± 1.3 25.2
* EAE was scored according to incidence, severity and duration. Disease index was calculated as follows: Mean Maximum Score × Mean Duration Disease × Incidence.
Significance: (a) Group I vs Group III p = 0.04 and Group I vs Group V p = 0.0002; (b) Group I vs Group III p = 0.009 and Group I vs Group V p = 0.0001; (c) Group II vs Group III p = 0.16 and Group II vs Group V p = 0.0005; (d) Group II vs Group III p = 0.12 and Group II vs Group V p = 0.001.
Antibody responses after vaccination
Antibodies in response to the immunogen were assessed by solid-phase ELISA at various times after immunization. As shown in Table 2, antibody titers against the immunogen developed in each group (group I-III) irrespective of which antibody was used to detect the antibody response in sera. This suggests that the human constant region of the antigenized antibodies is immunogenic in the rat. Antibody titers increased after the booster immunization and after challenge with the encephalitogenic MBP peptide. Control rats (group IV-V) did not mount any antibody response. No reactivity was found on the 19mer MBP peptide (GSLPQKSQRSQDENPVVHF) used as a control. Anti-TCR (anti-idiotypic) antibodies were tested using two approaches. In the first case, sera of immunized animals were tested on Vβ8.2 synthetic peptides by ELISA. A weak but distinct response was detected in both instances starting on day 21 or 28 (Figure 2). Sera from control animals did not react with TCR peptides. Together with the fact that these were tested at a 1:25 dilution it appears that the anti-idiotypic response is weak. In the second case, we tested anti-idiotypic antibodies for their reactivity with the TCR in its native configuration. This was done by flowcytometry using the Vβ8.2+ T cell hybridoma S23B1E11 as the cell substrate. Two out of six rats in group I had a bright cellular staining (Figure 3). Reactive antibodies were detectable on day 21, 28 and day 50. Rats immunized with γ1TCR-II (group II) as well control rats (group III-V) were negative. Interestingly, the two rats whose sera reacted with TCR by flowcytometry did not develop symptoms of EAE.
Table 2 Detection of antibodies against γ1TCR-I and γ1TCR-II in vaccinated Lewis rats
Days After Vaccination
a Immunogen Rats (No.) Responders (No.) 0 21 28 50
γ1TCR-I 6 6/6 ≤2.3* 3.9 ± .2 4.2 ± .2 4.5 ± 0.2
γ1TCR-II 10 10/10 ≤2.3 3.7 ± 0.2 4.1 ± 0.1 4.5 ± 0.2
γ1WT 10 10/10 ≤2.3 3.2 ± 0.4 3.9 ± 0.2 4 ± 0.2
CFA 5 0/5 ≤2.3 ≤2.3 2.6 ≤2.3
- 6 0/6 - - ≤2.3 ≤2.3
b
γ1TCR-I 6 6/6 ≤2.3 4 ± 0.2 4 ± 0.2 4.6 ± 0.3
γ1TCR-II 10 10/10 ≤2.3 4.1 ± 0.3 4.4 ± 0.3 4.7 ± 0.5
γ1WT 10 10/10 ≤2.3 3.3 ± 0.3 4 ± 0.2 4.2 ± 0.2
CFA 5 0/5 ≤2.3 ≤2.3 ≤2.3 ≤2.3
- 6 0/6 - - ≤2.3 ≤2.3
* Antibody titers are expressed in log10. Sera were tested on microtiter plates coated with each of the TCR antigenized antibody γ1TCR-I (panel a) or γ1TCR-II (pane b). End point dilutions were determined as the last serum dilution binding with an OD ≥ 0.200.
Figure 2 Antibody response to TCR peptides following vaccination with antigenized antibody γ1TCR-I or γ1TCR-II 39DMGHGLRLIHYSYDVNSTEKG59 tested on the ASSDSSNTE (panel a) or (panel b). The number of rats in each group is that indicated in Table 1. Results are expressed as Log2 ± SD.
Figure 3 Sera from rats vaccinated with γ1TCR-I bind Vβ8.2+ T cells by flowcytometry. Vβ8.2+ S23B1E11 T cell hybridoma were used as substrate. Sera were tested at 1:25 dilution. Bound antibodies were revealed using a FITC-conjugated goat antibody to rat Ig.
Vaccination with a murine antigenized antibody
To explore the importance of foreigness of the constant region on the immunogenicity of the Vβ8.2 peptides we engineered an antigenized antibody with a murine γ2b constant region. Homology search using the BLAST program of the NCBI gene bank indicated that the murine γ2b C region is 56.7% identical to the rat γ2b C region, with a homology of 71% between residues 106 and 333. Because significant protection was found in rats vaccinated with the antibody expressing the 93ASSDSSNTE101 peptide (γ1TCR-I), we engineered an antibody with the same V region (γ2bTCR-I). Rats vaccinated with γ2bTCR-I and subsequently immunized with MBP peptide, were protected only partially compared to rats vaccinated with γ1TCR-I (10.2 vs. 1.9) (Table 3). Notably, within the six rats immunized with γ2bTCR-I, two were grade ≤ 2 and four developed a grade 3 for an average of two days. On the other hand, three out of six rats immunized with control antibody γ2bWT proceeded through a grade 4 disease. Similarly, all five control rats (group III and IV) developed a grade 4 disease. Of note, although the severity of the disease in group I rats was less than in control group II, the difference was not statistically significant (Table 3). All rats developed antibodies to the respective immunogen. However, when compared with the total antibody titer of rats immunized with γ1TCR-I and γ1TCR-II the titers were on average lower at single time points (Table 4). All sera reacted with the synthetic peptide 93ASSDSSNTE101 starting from day 21 with a progressive increase over time (Figure 3).
Table 3 Protection against EAE by vaccination with antigenized antibodies with a murine γ2b constant region
Severity of Disease*
Group No. Rats Immunogen Incidence Max Score (mean ± SD) Duration (mean ± SD) Disease Index
I 6 γ2bTCR-I 6/6 2.5 ± 0.8a 4.2 ± 0.4b 10.5
II 6 γ2bWT 6/6 3.0 ± 1.3 5.7 ± 1.7 17.1
III 2 CFA 2/2 4 6 24
IV 3 - 3/3 4 7.3 ± 0.6 29.2
* EAE was scored according to incidence, severity and duration. Disease index was calculated as follows: Mean Maximum Score × Mean Duration of Disease × Incidence.
Significance: (a) Group I vs Group II p = 0.394 and Group I vs Group IV p = 0.051; (b) Group I vs Group III p = 0.05 and Group I vs Group IV p = 0.026.
Table 4 Detection of antibodies against γ2bTCR-I in vaccinated Lewis rats
Days After Vaccination
Immunogen Rats (No.) Responders (No.) 0 21 28 50
γ2bTCR-I 6 6/6 ≤2.3 3.2 ± 0,2 3.7 ± 0.1 4 ± 0.3
γ2bWT 6 6/6 ≤2.3 3 ± 0.3 3.4 ± 0.3 3.6 ± 0.3
CFA 2 0/2 ≤2.3 ≤2.3 ≤2.3 ≤2.3
- 3 0/3 - - ≤2.3 ≤2.3
* Antibody titers are expressed in log10. Sera were tested on microtiter plates coated with γ2bTCR-I. End point dilutions were determined as the last serum dilution binding with an OD ≥ 0.200.
Serum antibodies of vaccinated rats bind a synthetic peptide of human CD4
In the attempt to correlate the antibody response after vaccination with protection, the sera of vaccinated rats and their controls were tested on a synthetic peptide corresponding to amino acid residues 21–49 of the first extra-cellular domain of human CD4. This peptide binds Ig irrespective of antigen specificity and heavy chain isotype with an affinity of 10-5 M (26). It also binds antigen:antibody complexes formed at molar equivalence with an affinity about 100 fold higher [31]. When the sera of vaccinated rats were assayed on plates coated with the synthetic peptide of human CD4, strong binding was observed by sera from all rats immunized with γ1TCR-I whereas sera from rats immunized with γ1TCR-II or γ1WT bound much less (Figure 5a). Control sera of groups IV and V did not bind. Binding could be attributed either to a differential property of the two antigenized V regions or to differences in the immune response triggered by the V regions themselves. To distinguish between the two possibilities two experiments were performed. First, we assessed binding of γ1TCR-I and γ1TCR-II on the CD4 peptide. Both bound equally at saturating and non-saturating concentrations (data not shown). Second, we tested sera of rats immunized with γ2bTCR-I considering that, if the effect was due to the immune response against 93ASSDSSNTE101, we would have found similar results. As shown (Figure 5b), the sera of γ2bTCR-I vaccinated rats all bound to the CD4 peptide comparably to rats vaccinated with γ1TCR-I. This suggests that binding to the CD4 peptide may reflect differences in the type of V regions utilized by the antibodies generated in vivo in response to immunization with the TCR peptide 93ASSDSSNTE101 as compared with the TCR peptide 39DMGHGLRLIHYSYDVNSTEKG59 or the wild type V region. Further studies will be needed to clarify this issue.
Figure 5 Sera from rats vaccinated with γ1TCR-I or γ2bTCR-I bind synthetic peptide 21–49 of human CD4.
Proliferative response
Spleen cells and draining lymph nodes of rats tested 15 or 30 days after the initial immunization were tested in a proliferative assay against the Vβ8.2 peptides. No proliferative response was detected (data not shown).
Discussion
In this report we demonstrate that the severity of EAE in the Lewis rat can be greatly attenuated, and in some instances completely prevented, by active immunization with antigenized antibodies expressing amino acid sequences of the rat Vβ8.2 gene product. Immunity against synthetic peptides of the TCR has been shown to be effective in preventing or reducing the severity of EAE in the rat [17-20], suggesting that autoimmunity against the TCR reacting with encephalitogenic sequences of MBP is key to immunoregulatory events. The control of pathogenetic T cells may involve both T cells and antibodies. Autoregulation via T cells in EAE is well established. Thus, spontaneous recovery from EAE is impaired by splenectomy or thymectomy [34] and EAE can be prevented by vaccination with "attenuated" pathogenic T cells [15]. Autoregulation in EAE may involve both CD4+ and CD8+ T cells [35-38] as well as suppression by cytolytic T-T interactions [39]. A prevailing idea has been that in the rat [40] and in the mouse [41] idiotypic determinants of the TCR may be autoimmunogenic and contribute to mechanisms of immune regulation leading to protection. On the other hand, at least in a few instances, monoclonal antibodies against these TcR Vβ region [9,42,43] or against TCR idiotype [44] have been shown to block or attenuate disease.
Here we show that immunity against idiotopes of antibodies engineered to express TCR peptides is effective in generating anti-idiotypic immunity directed against rat Vβ8.2 TCR gene product. Importantly, this type of immunity protected from EAE. The new approach used herein to induce anti-TCR immunity is based on conventional idiotypic immunization in which antigenized antibodies mimic the immunogenic properties of soluble TCR functioning as a surrogate internal image [45] in much the same way as previously demonstrated for a non-self antigen [31]. The present approach is reminiscent of experiments in which induction of anti-idiotypic immunity against TCR with specificity for MHC was obtained by immunization with soluble alloantibodies of relevant specificity [46,47], or by immunization with autologous idiotype positive molecules that are shed from the cell surface in the serum [48]. Thus, antibodies purposely modified to express selected loops of the TCR obviate the necessity to purify the receptor, isolate idiotypic TCR molecules from the serum, or use antigen-specific T cell blasts.
Antibodies reacting with TCR peptides were detected in every vaccinated rat indicating that immunization with antigenized antibodies is an efficient method to induce an anti-idiotypic response specific for a somatic receptor. The fact that only two out of sixteen vaccinated rats developed antibodies against the native receptor detectable by flowcytometry on Vβ8.2+ T cells suggests that cross-reactive anti-idiotypic antibodies may be very low titer. Alternatively, they may be adsorbed on T cells in vivo precluding their detection in the serum. The first possibility is consistent with the self nature of TCR peptides and a predicted paucity of self reactive clonotypes within the natural B cell repertoire. Interestingly, we noted that the anti-idiotypic response against a non-self peptide expressed in an antigenized antibody [31] is much greater than the one observed here against a self peptide. That only rats vaccinated with the antigenized antibody expressing the 93ASSDSSNTE101 sequence developed flowcytometry-reactive autoantibodies could reflect difference in conformation once the two peptides are embedded in the CDR3 loop of an antigenized antibody. For instance, 93ASSDSSNTE101 could be better surface exposed and more stably expressed 39DMGHGLRLIHYSYDVNSTEKG59. A computer-assisted comparison of hydrophilicity profiles [49] of the 93ASSDSSNTE101 peptide in the parental. TCR Vβ8.2 gene product and in the antibody V region shows that in both instances the peptide is highly hydrophilic (Figure 5). On the other hand, the Vβ8.2 CDR2 region shows a highly hydrophilic profile alternating with large hydrophobic regions of poorly exposed amino acid residues, both in the parental TCR and in the antibody CDR (data not shown).
Our data show that although the process of antibody antigenization allows one to conformationally-constrain and express discrete peptide sequences of somatic receptors, the induction of anti-receptor antibodies is not directly predictable. Previously, we demonstrated flowcytometry-reactive antibodies to human CD4 in a high proportion (75 %) of cases [24]. We conclude that the physical characteristics of a given receptor peptide (e.g., length, hydrophilicity, etc.) likely determine its ability to induce antibodies cross-reactive with the native receptor.
Interestingly, rats immunized with the antigenized antibody expressing 93ASSDSSNTE101 but not 39DMGHGLRLIHYSYDVNSTEKG59 reacted immunologically with a synthetic peptide of human CD4 previously described to bind Ig [32]. Because the two antigenized antibodies reacted equally with the CD4 peptide and only differ by the composition of their CDR3, we suggest that binding to CD4 by anti- 93ASSDSSNTE101 serum antibodies underscores qualitative differences of the immune response between rats immunized with γ1TCR-I and 1TCR-II, respectively. Thus, it appears as if 93ASSDSSNTE101 induced a different immune response than 39DMGHGLRLIHYSYDVNSTEKG59. Furthermore, since vaccination with γ1TCR-I also promoted greater protection from EAE, it is tempting to speculate that a component of the anti-idiotypic response against γ1TCR-I is associated with protection.
In conclusion, three points have emerged from this study. First, antigenized antibodies expressing conformationally-constrained loops of the Vβ8.2 TCR can be used as vaccines in the prevention of EAE in the Lewis rat. Our new approach to generate anti-TCR immunity, confirms the relevance of anti-idiotypic regulation in controlling rat EAE [17,18,20]. Second, since a weak antibody anti-idiotypic response in the apparent lack of a cell proliferative response was associated with protection, it appears as if a humoral anti-TCR response is relevant to protection from disease. Although this contrasts the relevance of T cell immunity in the regulation of EAE in the rat, reports exist to support the idea that humoral immunity is also important [20,50,51]. EAE was shown to be prevented or attenuated by passive transfer of serum from rats recovering from EAE [52], or by passive transfer of monoclonal antibodies against these TCR Vβ region and its idiotypes [9,42-44]. However, whether anti-idiotypic antibodies against the TCR predispose to anergy, apoptosis or killing of pathogenetic T cells remains to be determined. Finally, our study indicates that antigenized antibodies can be used as vaccines in conditions where immunopathology and disease involve receptors on somatic cells, and anti-receptor immunity alone could prevent or mitigate a pathological condition.
Competing interests
The authors declare that they have no competing interests
Figure 4 Antibody response to TCR peptide 93ASSDSSNTE101 following vaccination with antigenized antibody γ2bTCR-I. The number of rats in each group are not indicated in Table 3. Results are expressed as means of Log2 ± SD.
Figure 6 Hydrophilicity profiles of TCR peptides-containing V regions. Hydrophilic profile of the rat Vβ8.2 TCR, amino acid residues 80–130, inclusive of the CDR3 sequence 93ASSDSSNTE101.
Figure 7 Hydrophilic profile of the mouse VH62, amino acid residues 80–125, engineered with the 93ASSDSSNTE101 peptide of the rat Vβ8.2 TCR-CDR3.
Acknowledgments
This work was supported by NIH grant PO1AI33204.
==== Refs
Swanborg RH Experimental autoimmune encephalomyelitis in the rat: lessons in T-cell immunology and autoreactivity Immunol Rev 2001 184 129 135 12086308 10.1034/j.1600-065x.2001.1840112.x
Hoffman PM Gaston DD Spitler LE Comparison of experimental allergic encephalomyelitis induced with spinal cord, basic protein, and synthetic encephalitogenic peptide Clin Immunol Immunopathol 1973 1 364 371 4127172 10.1016/0090-1229(73)90053-6
Richert JR Kies MW Alvord EJ Enhanced transfer of experimental allergic encephalomyelitis with Lewis rat lymph node cells J Neuroimmunol 1981 1 195 203 6175659 10.1016/0165-5728(81)90044-8
Polman CH Dijkstra CD Sminia T Koetsier JC Immunohistological analysis of macrophages in the central nervous system of Lewis rats with acute experimental allergic encephalomyelitis J Neuroimmunol 1986 11 215 222 3485656 10.1016/0165-5728(86)90005-6
Acha-Orbea H Mitchell DJ Timmerman L Wraith DC Taich GS Waldor MK Zamvil S McDevitt H Steinman L Limited heterogeneity of TcRs from lymphocytes mediating autoimmune encephalomyelitis aloows specific immune intervention Cell 1988 54 263 273 2455603 10.1016/0092-8674(88)90558-2
Urban J Kumar V Kono DH Gomez C Horvath SJ Clayton J Ando DG Sercarz EE Hood L Restricted use of T cell receptor V genes in murine autoimmune encephalomyelitis raises possibilities for antibody therapy. Cell 1988 54 577 592 2456857 10.1016/0092-8674(88)90079-7
Herber-Katz E A new hierarchy of TCR specificity: Autoimmune diseases are defined by particular VaVb combinations and not by antigen specificity Cold Spring Harbor Laboratory Press 1989 54 875 878
Steinman L Solomon D Lim M Zamvil S Sriram S Prevention of experimental allergic encephalitis with in vivo administration of anti I-A antibody. Decreased accumulation of radiolabelled lymph node cells in the central nervous system J Neuroimmunol 1983 5 91 97 6603474 10.1016/0165-5728(83)90029-2
Owhashi M Herber-Katz E Protection from experimental allergic encephalomyelitis conferred by a monoclonal antibody directed against a shared idiotype on rat T cell receptors specific for myelin basic protein J Exp Med 1988 168 2153 2164 2462007 10.1084/jem.168.6.2153
Brostoff SW Mason DW Experimental allergic encephalomyelitis: successful treatment in vivo with a monoclonal antibody that recognizes T helper cells J Immunol 1984 133 1938 1942 6236260
Welch AM Swanborg RH Characterization of suppressor cells involved in regulation of experimental allergic encephalomyelitis. Eur J Immunol 1976 6 910 912
Adda DH Beraud E Depieds R Evidence for suppressor cells in Lewis rats experimental allergic encefalomyelitis Eur J Immunol 1977 7 620 625 923633
Beraud E Varriale S farnarier C bernard D Suppressor cells in Lewis rats with experimental allergic encephalomyelitis: prevention of the disease and inhibition of lymphocytes proliferation by suppressor cells or their products. Eur J Immunol 1982 12 926 930 6217977
Miller A Lider O Roberts AB Sporn MB Weiner HL Suppressor T cells generated by oral tolerization to myelin basic protein suppress both in vitro and in vivo immune responses by the release of transforming growth factor beta after antigen-specific triggering Proc Natl Acad Sci U S A 1992 89 421 425 1370356
Ben-Nun A Wekerle H Cohen IR Vaccination against autoimmune encephalomyelitis with T-lymphocyte line cells reactive against myelin basic protein Nature 1981 292 60 61? 6974307 10.1038/292060a0
Rennie J The body against itself Scientific American 1990 263 106 115 2148647
Howell MD Winters ST Olee T Powell HC Carlo DJ Brostoff SW Vaccination against experimental allergic encephalomyelitis with T cell receptor peptides Science 1989 246 668 670 2814489
Vandenbark AA Hashim G Offern H Immunization with a synthetic T-cell receptor V-region peptide protects against experimental autoimmune encephalomyelitis Nature 1989 341 541 544 2477708 10.1038/341541a0
Offner H Hashim GA Vandenbark AA T cell receptor peptide therapy triggers autoregulation of experimental encephalomyelitis Science 1991 251 430 432 1989076
Stevens DB Karpus WJ Gould KE Swanborg RH Studies of V beta 8 T cell receptor peptide treatment in experimental autoimmune encephalomyelitis J Neuroimmunol 1992 37 123 129 1372325 10.1016/0165-5728(92)90163-F
Desquenne-Clark L Esch TR Otvos L Jr Heber-Katz E T-cell receptor peptide immunisation leads to enhanced and chronic experimental allergic encephalomyelitis Proc Natl Acad Sci USA 1991 88 7219 7223 1714594
Matsumoto Y Tsuchida M Hanawa H Abo T T cell receptor peptide therapy for autoimmune encephalomyelitis: stronger immunization is necessary for effective vaccination Cell Immunol 1994 153 468 478 8118876 10.1006/cimm.1994.1043
Kumar V Coulsell E Ober B Hubbard G Sercarz E Ward ES Recombinant T cell receptor molecules can prevent and reverse experimental autoimmune encephalomyelitis: dose effects and involvement of both CD4 and CD8 T cells J Immunol 1997 159 5150 5156 9366445
Lanza P Billetta R Antonenko S Zanetti M Active immunity against the CD4 receptor using an antibody antigenized with residues 41-55 of the first extracellular domain Proc Natl Acad Sci USA 1993 90 11683 11687 8265609
Zanetti M Antigenized antibodies Nature 1992 355 466 477 10.1038/355476a0
Sollazzo M Hasemann CA Meek KD Glotz D Capra JD Zanetti M Molecular characterization of the VH region of murine autoantibodies from neonatal and adult BALB/c mice Eur J Immunol 1989 19 453 457 2495969
Sollazzo M Billetta R Zanetti M Expression of an exogenous peptide epitope genetically engineereed in the variable domain of an immunoglobulin: implications for antibody and peptide folding. Protein Eng 1990 4 215 220 1706095
Morrison S Transfectomas provide novel chimeric antibodies Science 1985 229 1202 1207 3929380
Billetta R Zanetti M Ligand expression using antigenization of antibody: Principle and methods ImmunoMethods 1992 1 41 51
Hashim GA Carvalho EF Sharpe RD Definition and synthesis of the essential amino acid sequence for experimental allergic encephalomyelitis in Lewis rats J Immunol 1978 121 665 670 79612
Billetta R Hollingdale RM Zanetti M Immunogenicity of an engineered internal image antibody Proc Natl Acad Sci USA 1991 88 4713 4717 1905014
Lenert P Kroon D Spiegelberg H Golub ES Zanetti M Human CD4 binds immunoglobulins Science 1990 248 1639 1643 2363051
Gold DP Offner H Sun D Wiley S Vandenbark AA Wilson DB Analysis of T cell receptor beta chains in Lewis rats with experimental allergic encephalomyelitis: conserved complementarity determining region 3 J Exp Med 1991 174 1467 1476 1836012 10.1084/jem.174.6.1467
Ben-Nun A Ron Y Cohen IR Spontaneous remission of autoimmune encephalomyelitis is inhibited by splenectomy, thymectomy or ageing. Nature 1980 288 389 390 6159555 10.1038/288389a0
Karpus WJ Gould KE Swanborg RH CD4+ suppressor cells of autoimmune encephalomyelitis respond to T cell receptor-associated determinants on effector cells by interleukin-4 secretion Eur J Immunol 1992 22 1757 1763 1378016
Koh DR Ho A Rahemtulla A Penninger J Mak TW Experimental allergic encephalomyelitis (EAE) in mice lacking CD4+ T cells Eur J Immunol 1994 24 2250 2253 7916298
Jiang H Zhang SI Pernis B Role of CD8+ T cells in murine experimental allergic encephalomyelitis Science 1992 256 1213 1215 1375398
Koh DR Fung LW Ho A Gray D Acha OH Mak TW Less mortality but more relapses in experimental allergic encephalomyelitis in CD8-/- mice Science 1992 256 1210 1213 1589800
Sun D Qin Y Chluba J Epplen JT Wekerle H Suppression of experimentally induced autoimmune encephalomyelitis by cytolytic T-T cell interactions Nature 1988 332 843 845 2965794 10.1038/332843a0
Heber-Katz E Acha-Orbea H The V-region disease hypothesis: evidence from autoimmune encephalomyelitis Immunol Today 1989 10 164 169 2663017 10.1016/0167-5699(89)90174-6
Kumar V Sercarz E An integrative model of regulation centered on recognition of TCR peptide/MHC complexes Immunol Rev 2001 182 113 121 11722628 10.1034/j.1600-065X.2001.1820109.x
Hashim GA Vandenbark AA Galang AB Diamanduros T Carvalho E Srinivasa J Jones R Vainiene M Morrison WJ Offner H Antibodies specific for VB8 receptor peptide suppress experimental autoimmune encephalomyelitis J Immunol 1990 144 4621 4627 1693637
Matsumoto Y Tsuchida M Hanawa H Abo T Successful prevention and treatment of autoimmune encephalomyelitis by short-term administration of anti-T-cell receptor alpha beta antibody Immunology 1994 81 1 7 7510661
Zhou SR Whitaker JN Specific modulation of T cells and murine experimental allergic encephalomyelitis by monoclonal anti-idiotypic antibodies J Immunol 1993 150 1629 1642 7679432
Jerne NK Roland J Cazenave PA Recurrent idiotypes and internal images EMBO J 1982 1 243 247 6985352
McKearn TJ Hamada Y Stuart FP Fitch FW Anti-receptor antibody and resistance to graft-versus-host disease Nature 1974 251 648 650 4153575
Binz H Wigzell H Successful induction of specific tolerance to transplantation antigens using autoimmunisation against the recipient's own natural antibodies Nature 1976 262 294 295 958374
Binz H Wigzell H Specific transplantation tolerance induced by autoimmunization against the individual's own, naturally occurring idiotypic, antigen-binding receptors J Exp Med 1976 144 1438 1457 12245 10.1084/jem.144.6.1438
Kyte J Doolittle RF A simple method for displaying the hydropathic character of a protein J Mol Biol 1982 157 105 132 7108955
Sun D Synthetic peptides representing sequence 39 to 59 of rat V beta 8 TCR fail to elicit regulatory T cells reactive with V beta 8 TCR on rat encephalitogenic T cells Cell Immunol 1992 141 200 210 1372843
Kawano Y Sasamoto Y Vacchio MS Hodes RJ Gery I Immune responses against self-TCR peptides Cell Immunol 1994 159 235 245 7527731 10.1006/cimm.1994.1310
Paterson PY Harwin SM Suppression of allergic encephalomyelitis in rats by means of antibody serum J Exp Med 1963 117 755 780 13941827 10.1084/jem.117.5.755
| 15541175 | PMC539290 | CC BY | 2021-01-04 16:37:44 | no | J Immune Based Ther Vaccines. 2004 Nov 12; 2:9 | utf-8 | J Immune Based Ther Vaccines | 2,004 | 10.1186/1476-8518-2-9 | oa_comm |
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-2-801558506810.1186/1477-7827-2-80ResearchFunctional effects of 17alpha-hydroxyprogesterone caproate (17P) on human myometrial contractility in vitro Sexton Donal J [email protected]'Reilly Michael W [email protected] Anne M [email protected] John J [email protected] Department of Obstetrics and Gynaecology, National University of Ireland Galway, Clinical Science Institute, University College Hospital, Newcastle Road, Galway, Ireland2004 7 12 2004 2 80 80 18 10 2004 7 12 2004 Copyright © 2004 Sexton et al; licensee BioMed Central Ltd.2004Sexton 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
17alpha-hydroxyprogesterone caproate (17P) administration reportedly improves outcome for women with a previous spontaneous preterm delivery. This study, using in vitro strips of human uterine smooth muscle, aimed to investigate the direct non-genomic effects of 17P on spontaneous and induced contractions in tissues obtained during pregnancy, and in the non-pregnant state.
Methods
Biopsies of human myometrium were obtained at elective cesarean section, and from hysterectomy specimens, and dissected strips suspended for isometric recordings. The effects of 17P (1 nmol/L -10 micro mol/L) on spontaneous and agonist-induced (oxytocin 0.5 nmol/L for pregnant, phenylephrine 10 μmol/L for non-pregnant) contractions were measured. Integrals of contractile activity, including the mean maximal inhibition values (MMI) observed at the maximal concentration, were compared with those from simultaneously run control strips.
Results
There was no significant direct effect exerted by 17P on pregnant or non-pregnant human myometrial contractility. The MMI ± SEM for spontaneous contractions in pregnant myometrium was 4.9% ± 7.2 (n = 6; P = 0.309) and for oxytocin-induced contractions was 2.2% ± 1.3 (n = 6; P = 0.128). For non-pregnant myometrium, the MMI ± SEM for spontaneous contractions was 8.8% ± 11.0 (n = 6; P = 0.121) and for phenylephrine induced contractions was -7.9% ± 6.5 (n = 6; P = 0.966).
Conclusions
The putative benefits of 17P for preterm labor prevention are not achieved, even partially, by a direct utero-relaxant effect. These findings outline the possibility that genomic effects of 17P, achieved over long periods of administration, are required for its reported therapeutic benefits.
==== Body
Background
Preterm delivery constitutes a major problem in obstetric practice because of the large associated contribution to perinatal mortality and morbidity [1,2]. A significant proportion of all preterm deliveries are due to spontaneous preterm labor [2]. Despite much research effort, until recently, no effective method of preventing or treating preterm labor, and improving neonatal outcome, has been available. Meta-analysis of various tocolytic compounds in clinical practice has revealed that while they resulted in a delay in the interval to delivery, of time periods up to a week, they did not reduce the incidence of preterm delivery at different gestational ages [3]. More importantly, their use, in comparison to placebo, was not associated with any benefit in terms of objective measures of neonatal wellbeing or morbidity.
It has however been recently reported that weekly injections of 17-alpha-hydroxyprogesterone caproate (17P), in women who have had a previous spontaneous preterm delivery, significantly reduces the risk of preterm delivery before 37, 35 and 32 week's gestation [4]. In addition, infants of women treated with 17P had significantly lower rates of necrotising enterocolitis, intraventricular haemorrhage and the need for supplemental oxygen. While evidence for the use of progestational compounds for prevention of preterm delivery, and recurrent miscarriage, has been conflicting [5-7], meta-analysis restricted to trials of 17P has suggested a significant reduction in the preterm delivery rate. This reported benefit of 17P, while being a welcome development in therapeutic strategies for preterm labor, has raised many questions in relation to whether the same benefit would apply to low risk groups, and the potential effects, if any, from the castor-oil injection of placebo used [8,9]. One of the most important questions relates to the mechanism by which the drug works and there are currently no data to answer this. Progestins have the potential to exert both genomic and non-genomic effects. The aims of this study were focused specifically on the latter mechanism i.e. to investigate the direct effects of 17P on contractions of isolated human myometrium, both spontaneous and agonist-induced, in tissue obtained during pregnancy and in the non-pregnant state.
Methods
Patient Recruitment and Tissue collection
Patient recruitment took place in the Department of Obstetrics and Gynaecology, University College Hospital Galway. Ethical committee approval for tissue collection was obtained from the Research Ethics Committee at University College Hospital Galway and recruitment was by written informed consent. The biopsies were excised from the upper lip of the lower uterine segment incision in the midline i.e. upper portion of lower uterine segment. Women undergoing induction of labor were excluded from the study. For hysterectomy specimens, all women were pre-menopausal and undergoing abdominal hysterectomy without evidence of malignant uterine disease. Biopsies of myometrial tissue from hysterectomy specimens were obtained from the fundus. The biopsies were immediately placed in Krebs-Henseleit physiological salt solution (PSS), pH 7.4, containing: 4.7 mmol/L KCl, 118 mmol/L NaCl, 1.2 mmol/L Mg2SO4, 1.2 mmol/L CaCl2, 1.2 mmol/L KPO4, 25 mmol/L NaHCO3, and 11 mmol/L glucose. Tissues were stored at 4° C and used within 12 hours of collection.
Tissue Bath Experiments
Longitudinal myometrial strips were dissected measuring approximately 2 × 2 × 10 mm. The strips were mounted for isometric recording under 2 grams of tension in organ tissue baths, as previously described [10-12]. The tissue baths contained 10 ml of Krebs-Henseleit physiologic salt solution maintained at 37°C, pH 7.4 and gassed continuously with 95%O2/5%CO2. Myometrial strips were allowed to equilibrate for at least 60 minutes, during which time the Krebs-Henseleit physiologic salt solution was changed every 15 minutes. After the equilibration period, regular spontaneous myometrial contractions were allowed to develop. In separate experiments contractions were induced by the addition of oxytocin (0.5 nmol/L) to pregnant myometrial strips, and phenylephrine (10 μmol/L) to strips obtained from non-pregnant myometrium.
Once regular phasic contractions had developed, the integrated tension for the first 20 minutes was calculated, and this value served as a control since no significant spontaneous reduction in myometrial contractility was observed over the duration of experiments in control strips (without addition of vehicle or 17P). The mechanical response of tissues was then measured by calculation of the integral of selected areas for 20 minute periods, corresponding to the cumulative exposure to 17P or vehicle, using the PowerLab hardware data acquisition system and Chart v3.6 software (AD Instruments, Hastings, UK). At 20 minute intervals 17P was added in a cumulative manner at concentrations of 1 nmol/L, 10 nmol/L, 100 nmol/L, 1 μmol/L, and 10 μmol/L (i.e., 1 × 10-9 - 1 × 10-5M). Control strips (i.e. without exposure to 17P) were run simultaneously and for a similar duration, consisting of two separate groups of experiments as follows: 1. exposure of strips to PSS only; 2. exposure of strips to PSS and the vehicle for 17P. The overall duration of experiments was therefore 3 hours which is in accordance with standard in vitro myometrial experiments, allowing for an accurate drug exposure period of 20 minutes. The effects of 17P were assessed by subtracting the integrals of contractility measured after each bath exposure of 17P, from those obtained in control experiments (vehicle only). This allowed for calculation of the net effect of 17P on myometrial contractility. Potential effects of the vehicle were obtained by subtraction of the mean integrals measured after vehicle exposure, from those obtained without addition of vehicle i.e. spontaneous or agonist-induced contractions in PSS only. Percentage contractility was calculated by expressing the net integral measured after each 17P concentration addition, as a percentage of the integral calculated in the 20 minute period prior to any 17P addition. The mean maximal inhibition (MMI) refers to the mean percentage relaxation [i.e. 100% - mean contractility measured for each separate n sample of non-pregnant and pregnant myometrium] observed at the maximal bath concentration of 17P (i.e. 10 μmol/L), or corresponding concentration of vehicle for control strips.
Drugs and Solutions
Oxytocin, phenylephrine and 17P were purchased from Sigma-Aldrich, Dublin, Ireland. Because 17P is a lipid soluble compound it was necessary to dissolve it in a lipophilic vehicle. The vehicle used consisted of 75% Dimethyl sulfoxide (DMSO): 25% ethanol, to make a 1 mmol/L (10-3M) stock solution of 17P. The resultant final tissue bath concentrations of DMSO and ethanol, at the maximum concentration of 17P investigated (i.e. 10 μmol/L or 10-5M), were 0.83% and 0.28% respectively, allowing for the cumulative effect at each bath addition. DMSO and ethanol were purchased from Sigma-Aldrich, Dublin, Ireland. Fresh Krebs-Henseleit physiological salt solution was made daily. Fresh 17P solutions were prepared on the day of experimentation and were maintained at room temperature for the duration of the experiment.
Statistical Analysis
The integrals of contractile activity measured were compared to control values, and expressed as a percentage of that measured before drug addition. The effects of 17P on myometrial contractility were calculated for each 20-minute period of exposure from 1 nmol/L-10 μmol/L, and compared with the measurements obtained from control strips (ie. spontaneous, oxytocin-induced or phenylephrine-induced contractility, in the presence and absence of vehicle). The integrals of contractile activity were compared using a one-way ANOVA, which was followed by a Tukey HSD post-hoc. A value of P < 0.05 was accepted as statistical significance.
Results
There were 6 women recruited for the study at the time of elective cesarean section. The reasons for cesarean section included previous cesarean section (n = 2), breech presentation (n = 2), unstable lie (n = 1) and postmaturity with poor cervical Bishop score (n = 1). The mean age, ± standard error of the mean (SEM), was 36.3 ± 2.8 years (range 26–46). The median gestation at the time of cesarean section was 39 weeks (range 38–41). In relation to parity, 2 of the women were para 0, 1 was para 1, and 3 were of parity greater than 1, at the time of recruitment. There were 6 women recruited at the time of abdominal hysterectomy. The reasons for hysterectomy included menorrhagia (n = 4), irregular vaginal bleeding (n = 1), and fibroids with an ovarian cyst (n = 1). The mean age ± SEM was 42.2 ± 2.4 years (range 36–50).
For non-pregnant myometrial tissue representative recordings are shown in Figure 1. In Figure 1A a recording of spontaneous myometrial contractions is shown. In Figure 1B, the effects of vehicle for 17P on spontaneous contractions is demonstrated, and in Figure 1C the effects of addition of 17P are shown. The results of the calculated integrals are provided in Table 1. Bath exposure of the strips from hysterectomy specimens to 17P did not result in any alteration in contractile activity, at any of the bath concentrations studied, in comparison to vehicle only experiments (spontaneous contractions: P = 0.121; phenylephrine-induced contractions: P = 0.966).
Figure 1 Effects of 17P on myometrial contractility in non-pregnant tissue. Representative recordings of spontaneous contractions in PSS only (A), the effects of cumulative additions of 17P vehicle (B), and the effects of cumulative additions of 17P (C) are shown.
Table 1 Mean Maximal Inhibition Values for Vehicle and 17P.
Myometrial Contractility Non-Pregnant Pregnant
Net Vehicle Relaxation ± SEM Net 17P Relaxation ± SEM Net Vehicle Relaxation ± SEM Net 17P Relaxation ± SEM
Spontaneous -0.6% ± 12.1 (n = 6) 8.8% ± 11.0 (n = 6) 49.7% ± 11.9 (n = 6)♦ 4.9% ± 7.2 (n = 6)
*Agonist-Induced 31.8% ± 5.0 (n = 6) -7.9% ± 6.5 (n = 6) 56.2% ± 2.3 (n = 6)¶ 2.2% ± 1.3 (n = 6)
*Phenylephrine-induced in non-pregnant myometrium and oxytocin-induced in pregnant myometrium.
♦ P = 0.022 in comparison to controls with PSS only
¶ P < 0.001 in comparison to controls with PSS only
The results obtained from myometrium during pregnancy are similarly shown in Figure 2, and in Table 1. Figures 2A, 2B and 2C demonstrate recordings of oxytocin-induced contractions, the effects of vehicle, and the effects of 17P respectively. There was no significant net relaxant or uterotonic effect exerted by 17P on pregnant myometrial contractility, at any of the bath concentrations studied experiments (spontaneous contractions P = 0.309; oxytocin-induced contractions: P = 0.128). No significant difference was observed between the effects of 17P on contractility in either pregnant or non-pregnant myometrium.
Figure 2 Effects of 17P on myometrial contractility in pregnant tissue. Representative recordings of oxytocin-induced contractions in PSS only (A), the effects of cumulative additions of 17P vehicle (B), and the effects of cumulative additions of 17P (C) are shown.
The vehicle for 17P independently exerted a uterorelaxant effect on spontaneous and agonist-induced contractions in pregnant myometrial tissue only (spontaneous contractions P = 0.022; oxytocin-induced contractions: P = 0.000), and this was not observed in non-pregnant myometrial tissue (spontaneous contractions P = 0.241; phenylephrine-induced contractions: P = 0.068) (Table 1).
Discussion
These results demonstrate that 17P does not appear to exert a direct relaxant effect on human myometrial contractions in vitro, in tissue obtained during the third trimester of pregnancy, or in the non-pregnant state. These findings are in contrast to the inhibitory effect of progesterone derivatives on spontaneous contractions in animal uterine tissues [13,14], and therefore suggest that the reported benefit of 17P in preventing preterm delivery in women who have had a previous preterm delivery, involves other mechanisms of action, presumably via its genomic effects. It seems likely that prolonged exposure of uterine smooth muscle to 17P, with resultant activation of the progesterone receptor isoforms, has the potential to modify gene transcription in order to maintain physiological uterine quiescence. However no direct effect on contractile activity was observed in our study over a period of hours of exposure.
Previous studies, using various progestins, have yielded conflicting results in terms of the direct effects of these metabolites on human myometrial contractility in vitro. Progesterone metabolites, in some studies, have been reported to decrease both the frequency and amplitude of contractions [15-17], while other reports have outlined that progesterone stimulates the frequency and tonus of contractions in term human myometrium [18,19]. It has been hypothesized that progesterone addition to myometrial strips only enhances contractility if the tissue specimen was never deprived of progesterone i.e. placed immediately in a medium containing progesterone [19]. The studies to date have included various progesterone metabolites but we are unaware of any previous studies evaluating the effects of 17P on human myometrial contractions in vitro. The focus on 17P in this study has arisen from the recently reported randomized clinical trial outlining its benefits clinically for recurrent preterm labor.
17P is a naturally occurring progesterone which has been isolated from corpus luteum and adrenal gland. The synthetic caproate ester, like the naturally occurring compound, is a steroid, is highly lipophilic and is inactive when administered orally. To achieve solubility for clinical studies, the vehicle used for injection was castor-oil [4]. This led to significant controversy in relation to the potential effect of castor-oil on myometrial contractions in the placebo arm of the recently reported study [8]. For laboratory, or in vitro studies, achieving the required solubility of progestins can also be a difficult process. Efforts to achieve solubility in previous reports have included the use of Hepes buffer and ultrasound baths [19,20] with a resultant maximum 50% solubility. After numerous attempts at solubility for 17P in our studies, the most appropriate vehicle was a combination of DMSO and ethanol. While both of these compounds may exert an effect on uterine contractility, the maximal bath concentrations of both solvents was 0.83% and 0.28% respectively, which is acceptable for in vitro studies of this nature. In addition, control experiments, with PSS only, and PSS plus vehicle, were simultaneously run, to clearly evaluate any potential effect of vehicle. It is however apparent from our results, that addition of vehicle only did exert a significant relaxant effect on myometrium obtained during pregnancy, but no effect of vehicle was observed in myometrium obtained in the non-pregnant state. For both tissues, addition of 17P did not alter contractile activity. While the vehicle effects observed were sizeable, the experiment design, the numbers of patients recruited, and the reproducibility of the results, clearly indicate that 17P does not exert a direct effect on human myometrial contractility. In addition, while the experiments described here investigated the effects of 17P added cumulatively, separate experiments (data not shown) using a single dose (the highest dose) revealed similar results i.e. there was no evidence of tachyphylaxis over the duration of the experiments.
There are some limitations to the methodology used in our study. The tissue biopsies from the non-pregnant uterus were obtained from the body of the uterus, while those obtained at the time of caesarean section were excised from the lower uterine segment. The results were similar for both tissue types, and there is reasonable evidence to suggest that the functional characteristics of lower and upper uterine segment myometrium are similar [21]. At present there are no data pertaining to differential progesterone receptor expression levels between upper and lower uterine segments. There are also obvious ethical constraints in obtaining biopsies from the upper segment of the uterus at cesarean section, and it is not feasible to dissect strips with certainty from the lower segment of the non-pregnant uterus. Secondly, the tissue samples obtained during pregnancy were all recruited from women at term. It could be that preterm myometrium displays a different responsiveness. Finally, in vivo administration of 17P could potentially result in the formation of a metabolite with a more efficient uterorelaxant effect.
Conclusions
Whether 17P injections become incorporated for routine use in the management of preterm labor remains to be seen. Questions surrounding its true benefits, the associated difficulties in terms of vehicle solubility and placebo, and the mechanism of action, remain. Our findings highlight the solubility problems for scientific evaluation as occurred in clinical studies. Whatever the possible mechanism of action of 17P in reducing the incidence and adverse sequelae of preterm labor, this study demonstrates that it does not exert a direct relaxant effect, unlike other conventional methods of tocolysis investigated, and raised the likeliehood of a genomic effect secondary to long term administration during pregnancy.
Authors' contributions
DJS performed the experiments and wrote the manuscript. MWO'R performed the experiments. AMF analysed the data and wrote the manuscript. JJM designed, supervised the study and wrote the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We are grateful to the Medical and Midwifery Staff at University College Hospital Galway for their assistance in patient recruitment and obtaining biopsy specimens.
==== Refs
Goldenberg RL Rouse DJ Prevention of premature birth N Engl J Med 1998 339 313 320 9682045 10.1056/NEJM199807303390506
Slattery MM Morrison JJ Preterm delivery Lancet 2002 360 1489 1497 12433531 10.1016/S0140-6736(02)11476-0
Gyetvai K Hannah ME Hodnett ED Ohlsson A Tocolytics for preterm labor: a systematic review Obstet Gynecol 1999 94 869 877 10546776 10.1016/S0029-7844(99)00329-4
Meis PJ Klebanoff M Thom E Dombrowski MP Sibai B Moawad AH Spong CY Hauth JC Miodovnik M Varner MW Leveno KJ Caritis SN Iams JD Wapner RJ Conway D O'Sullivan MJ Carpenter M Mercer B Ramin SM Thorp JM Peaceman AM Gabbe S Prevention of preterm delivery by 17 alpha-hydroxyprogesterone caproate N Engl J Med 2003 348 2379 2385 12802023 10.1056/NEJMoa035140
Johnson JWC Austin KL Jones GS Davis GH King TM Efficacy of 17α-hydroxyprogesterone caproate in the prevention of premature labor N Engl J Med 1975 293 675 680 1099445
Hauth JC Gilstrap LC IIIBrekken AL Hauth JM The effect of 17 alpha-hydroxyprogesterone caproate on pregnancy outcome in an active-duty military population Am J Obstet Gynecol 1983 146 187 190 6682631
Goldstein P Berrier J Rosen S Sacks HS Chalmers TC A meta-analysis of randomized control trials of progestational agents in pregnancy Br J Obstet Gynaecol 1989 96 265 274 2653414
Brancazio LR Murtha AP Heine RP Prevention of recurrent preterm delivery by 17 alpha-hydroxyprogesterone caproate N Engl J Med 2003 349 1087 1088 12968095 10.1056/NEJM200309113491115
Greene MF Progesterone and preterm delivery – déjà vu all over again N Engl J Med 2003 348 2453 2455 12802032 10.1056/NEJMe030081
Moran CJ Friel AM Smith TJ Cairns M Morrison JJ Expression and modulation of Rho kinase in human pregnant myometrium Mol Hum Reprod 2002 8 196 200 11818523 10.1093/molehr/8.2.196
Doheny HC Houlihan DD Ravikumar N Smith TJ Morrison JJ Human chorionic gonadotrophin relaxation of human pregnant myometrium and activation of the BKCa Channel J Clin Endocrinol Metab 2003 88 4310 4315 12970303 10.1210/jc.2003-030221
Potter SM Astbury K Morrison JJ Effects of Thyroptropin Releasing Hormone (TRH) on human myometrium and umbilical vasculature in vitro Am J Obstet Gynecol 2004 190 246 251 14749667 10.1016/S0002-9378(03)00853-6
Kubli-Garfias C Hoyo-Vadillo C Lopez-Nieto E Ponce-Monter H Inhibition of spontaneous contractions of the rat pregnant uterus by progesterone metabolites Proc West Pharmacol Soc 1983 26 115 118 6889324
Putman CD Brann DW Kolbeck RC Mahesh VB Inhibition of uterine contractility by progesterone and progesterone metabolites: Mediation by progesterone and gamma amino butyric acidA receptor systems Biol Reprod 1991 45 266 272 1664743
Barnafi L Larraguibel R The in vitro effect of progesterone and oestrogens on the spontaneous and oxytocin-induced activity of the human myometrium Acta Endocrinol 1974 76 172 177 4406581
Lofgren M Holst J Backstrom T Effects in vitro of progesterone and two 5alpha-reduced progestins, 5alpha-pregnane-3, 20-dione and 5alpha-pregnane-3alpha-ol-20-one, on contracting human myometrium at term Acta Obstet Gynecol Scand 1992 71 28 33 1315093
Thornton S Terzidou V Clark A Blanks A Progesterone metabolite and spontaneous myometrial contractions in vitro Lancet 1999 353 1327 1329 10218536 10.1016/S0140-6736(99)00147-6
Fu X Rezapour M Lofgren M Ulmsten U Backstrom T Antitachyphylactic effects of progesterone and oxytocin on term human myometrial contractile activity in vitro Obstet Gynecol 1993 82 532 538 8377978
Fu X Rezapour M Lofgren M Ulmsten U Backstrom T Unexpected stimulatory effect of progesterone on human myometrial contractile activity in vitro Obstet Gynecol 1993 82 23 28 8515921
Fu X Ulmsten U Backstrom T Interaction of sex steroids and oxytocin on term human myometrial contractile activity in vitro Obstet Gynecol 1994 84 272 277 8041546
Luckas MJ Wray S A comparison of the contractile properties of human myometrium obtained from the upper and lower uterine segments BJOG 2000 107 1309 1311 11028587
| 15585068 | PMC539291 | CC BY | 2021-01-04 16:36:43 | no | Reprod Biol Endocrinol. 2004 Dec 7; 2:80 | utf-8 | Reprod Biol Endocrinol | 2,004 | 10.1186/1477-7827-2-80 | oa_comm |
==== Front
J NeuroinflammationJournal of Neuroinflammation1742-2094BioMed Central London 1742-2094-1-241558828710.1186/1742-2094-1-24ResearchPassive immunotherapy against Aβ in aged APP-transgenic mice reverses cognitive deficits and depletes parenchymal amyloid deposits in spite of increased vascular amyloid and microhemorrhage Wilcock Donna M [email protected] Amyn [email protected] Arnon [email protected] Sangeetha [email protected] Melissa J [email protected] Marcia N [email protected] Dave [email protected] Alzheimer's Research Laboratory, University of South Florida, Department of Pharmacology, 12901 Bruce B Downs Blvd, Tampa, Florida 33612, USA2 Alzheimer's Research Laboratory, University of South Florida, Department of Interdisciplinary Oncology, 12901 Bruce B Downs Blvd, Tampa, Florida 33612, USA3 Rinat Neuroscience Corp., 3155 Porter Drive, Palo Alto, California 94304, USA2004 8 12 2004 1 24 24 10 11 2004 8 12 2004 Copyright © 2004 Wilcock et al; licensee BioMed Central Ltd.2004Wilcock 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
Anti-Aβ immunotherapy in transgenic mice reduces both diffuse and compact amyloid deposits, improves memory function and clears early-stage phospho-tau aggregates. As most Alzheimer disease cases occur well past midlife, the current study examined adoptive transfer of anti-Aβ antibodies to 19- and 23-month old APP-transgenic mice.
Methods
We investigated the effects of weekly anti-Aβ antibody treatment on radial-arm water-maze performance, parenchymal and vascular amyloid loads, and the presence of microhemorrhage in the brain. 19-month-old mice were treated for 1, 2 or 3 months while 23-month-old mice were treated for 5 months. Only the 23-month-old mice were subject to radial-arm water-maze testing.
Results
After 3 months of weekly injections, this passive immunization protocol completely reversed learning and memory deficits in these mice, a benefit that was undiminished after 5 months of treatment. Dramatic reductions of diffuse Aβ immunostaining and parenchymal Congophilic amyloid deposits were observed after five months, indicating that even well-established amyloid deposits are susceptible to immunotherapy. However, cerebral amyloid angiopathy increased substantially with immunotherapy, and some deposits were associated with microhemorrhage. Reanalysis of results collected from an earlier time-course study demonstrated that these increases in vascular deposits were dependent on the duration of immunotherapy.
Conclusions
The cognitive benefits of passive immunotherapy persist in spite of the presence of vascular amyloid and small hemorrhages. These data suggest that clinical trials evaluating such treatments will require precautions to minimize potential adverse events associated with microhemorrhage.
==== Body
Background
Alzheimer's disease is characterized not only by the presence of parenchymal amyloid deposits and intracellular tangles but also by the presence of amyloid deposits in the vasculature, a condition referred to as cerebral amyloid angiopathy (CAA). The CAA observed in both Alzheimer's disease patients [1] and some of the transgenic mouse models [2] is primarily composed of the shorter form of amyloid beta (Aβ), Aβ1–40, while the majority of amyloid deposits in the parenchyma are composed of Aβ1–42, although the compact amyloid deposits also contain Aβ1–40.
Anti-Aβ immunotherapy has been considered as a potential treatment for Alzheimer's disease for some time [3,4]. Active immunization with a vaccine including Aβ1–42 fibrils progressed to human clinical trials where its administration was suspended due to meningoencephalitits in a subset of patients [5]. To date there have been pathology reports on two patients who participated in the trial and subsequently died [6,7]. Both reports note that while the numbers of parenchymal amyloid deposits appeared lower than expected in these cases, the CAA in these patients did not appear outside the normal range for Alzheimer's disease. In addition, one report mentioned multiple cortical hemorrhages and the presence of hemosiderin around the CAA vessels [7].
Given the adverse reactions to the active immunization, the irreversibility of such procedures and the variable antibody response to vaccines in older individuals [8], passive immunization against the Aβ peptide emerged as an alternative immunotherapeutic strategy. Studies in young and middle aged APP-transgenic mice have reported significant amyloid reductions with passive immunization [9-11]. Such treatments also demonstrate rapid improvements of memory function in APP-transgenic mice, sometimes without detectable reductions in amyloid [12-14]. Most recently, intracranial administration of anti-Aβ antibodies has been shown to not only remove Aβ but also clear, early-stage, hyperphosphorylated-tau aggregates [15]. Importantly, in the only prior study evaluating adoptive antibody transfer in older APP-transgenic mice, Pfeifer et al. [16] reported a doubling of cerebral microhemorrhages associated with significant reductions in amyloid burden after administration of an N-terminal specific anti-Aβ antibody.
Materials and Methods
Experiment design
Mice derived from APP Tg2576 mice were obtained from our breeding program at University of South Florida started in 1996 [17]. For the 5-month treatment study, 13 APP-transgenic mice, aged 23 months, were assigned to one of two groups. The first group received weekly intraperitoneal anti-Aβ antibody injections (antibody 2286; mouse-monoclonal anti-human Aβ28–40 IgG1; Rinat Neurosciences, Palo Alto, CA) for a period of 5 months (n = 6). The second group received weekly intraperitoneal anti-AMN antibody (2906; mouse-monoclonal anti-Drosophila amnesiac protein IgG1; Rinat Neurosciences, Palo Alto, CA) injections for a period of 5 months (n = 7). Seven nontransgenic mice were also assigned to one of two groups. The first group received weekly intraperitoneal anti-Aβ antibody injections for a period of 5 months (n = 4). The second group received weekly intraperitoneal anti-AMN antibody injections for a period of 5 months (n = 3).
For the time course study of 1-, 2- or 3-month treatment, 22 APP-transgenic mice aged 19 months were assigned to one of four experimental groups, as described previously [14]. The first three groups received weekly intraperitoneal anti-Aβ antibody injections for 3 months, 2 months or 1 month, ending when all mice were 22 months of age. The fourth group received weekly intraperitoneal anti-AMN antibody injections for 3 months.
Behavioral analysis
Following 3 and 5 months of treatment, the mice from the 5-month study were subjected to a two-day radial-arm water-maze paradigm. The apparatus was a 6-arm maze as described previously [18]. On day one, 15 trials were run in three blocks of 5. A cohort of 4 mice were run sequentially for each block (i.e., each of 4 mice get trial one, then the same mice get trial two, etc.). After each 5-trial block, a second cohort of mice was run permitting an extended rest period before mice were exposed to the second block of 5 trials. The goal arm was different for each mouse in a cohort to minimize odor cues. The start arm was varied for each trial, with the goal arm remaining constant for a given individual for both days. For the first 11 trials, the platform was alternately visible then hidden (hidden for the last 4 trials). On day two, the mice were run in exactly the same manner as day one except that the platform was hidden forall trials. The number of errors (incorrect arm entries) was measured in a one-minute time frame. As standard practice, mice failing to make an arm choice in 20 seconds are assigned one error, but no mice in this study had to be assigned an error in this manner. The same individual administered the antibody treatments and placed mice in the radial-arm water maze. Due to the numbers of mice in the study the researcher was unaware of treatment group identity of each mouse. Also, the dependent measures in the radial-arm water-maze task are quantitative, not evaluative, so the potential for tester bias is reduced. In order to minimize the influence of individual trial variability, each mouse's errors for 3 consecutive trials were averaged producing 5 data points for each day, which were analyzed statistically by ANOVA using StatView (SAS Institute Inc., NC).
Tissue preparation and histology
On the day of sacrifice mice were weighed, overdosed with 100 mg/kg Nembutal (Abbott laboratories, North Chicago, IL), and then intracardially perfused with 25 mL of 0.9% sodium chloride. Brains were rapidly removed, and the left half of the brain was immersion fixed for 24 h in freshly prepared 4% paraformaldehyde in 100 mM KPO4 (pH 7.2) for histopathology. The hemi-brains were then incubated for 24 h in 10%, 20% and 30% sucrose sequentially for cyroprotection. Horizontal sections of 25 μ thickness were collected using a sliding microtome and stored at 4°C in Dulbecco's phosphate-buffered saline with sodium azide (pH 7.2) to prevent microbial growth. A series of 8 equally spaced tissue sections 600 μ apart were randomly selected spanning the entire brain and stained using free-floating immunohistochemistry for total Aβ (rabbit polyclonal anti-pan Aβ; Biosource, Camarillo, CA, 1:10,000) as previously described [2,14]. A second series of tissue sections 600 μm apart were stained using 0.2% Congo red in NaCl-saturated 80% ethanol. Another set of sections were also mounted and stained for hemosiderin using 2% potassium ferrocyanide in 2% hydrochloric acid for 15 min, followed by a counterstain in a 1% neutral red solution for 10 min. Quantification of Congo red staining and Aβ immunohistochemistry was performed using the Image-Pro Plus (Media Cybernetics, Silver Spring, MD) to analyze the percent area occupied by positive stain. One region of the frontal cortex and three regions of the hippocampus were analyzed (to ensure that there was no regional bias in the hippocampal values). The initial analysis of Congo red was performed to give a total value. A second analysis was performed after manually editing out all of the parenchymal amyloid deposits to yield a percent area restricted to vascular Congo red staining. To estimate the parenchymal area of Congo red, we subtracted the vascular amyloid values from the total percentage. For the hemosiderin stain the numbers of Prussian blue-positive sites were counted on all sections and the average number of sites per section calculated. Looking at the sections at a low magnification we were able to observe a qualitative differences between animals; however, the percent area was so low that many fields contained no positive stain. Eight equally spaced sections were examined and the number of positive profiles was determined and averaged to a per-section value. To assess possible treatment-related differences, the values for each treatment group were analyzed by one-way ANOVA followed by Fisher's LSD means comparisons.
Results
Reversal of cognitive deficits by passive amyloid immunotherapy
The radial-arm water-maze task detects spatial learning and memory deficits in transgenic mouse models [18,19]. We treated 23-month-old mice for 5 months with anti-Aβ antibody 2286 or control antibody 2906 (against a Drosophila-specific protein) and tested them for spatial navigation learning in a two-day version of the radial-arm water maze after 3 months of treatment and, using a new platform location, again after 5 months of treatment. At both testing times we found that APP-transgenic mice treated with the control antibody failed to learn platform location over two days of testing and were significantly impaired compared to the nontransgenic mice treated with either antibody (Fig. 1). However, APP-transgenic mice administered the anti-Aβ antibodies demonstrated a complete reversal of the impairment observed in the control-treated APP-transgenic mice, ending day two with a mean performance near 0.5 errors per trial (Fig. 1). Although learning at the later time point, when the mice were 28 months of age, may have been slightly slower for all groups, there was no impairment of the anti-Aβ antibody-treated APP.
Figure 1 Spatial learning deficits in APP-transgenic mice were reversed following 3 and 5 months of immunization. Mice were tested in a two-day version of the radial-arm water maze. Solid lines represent APP-transgenic mice while dashed lines represent nontransgenic mice. Open symbols indicate anti-AMN, control-antibody treatment (○: APP-transgenic, control antibody; △: nontransgenic, control antibody) while closed symbols indicate anti-Aβ antibody treatment (●: APP-transgenic, Aβ antibody; ▲: nontransgenic, Aβ antibody). Panel A shows mean number of errors made over the two-day trial period following 3 months of immunization. Each data point is the average of 3 trials. Panel B shows the mean number of errors made over the 2-day trial period following 5 months of immunization. For both graphs * indicates p < 0.05, ** indicates p < 0.001 when the APP-transgenic mice receiving control antibody are compared with the remaining groups.
Passive amyloid immunotherapy clears parenchymal Aβ deposits, but increases vascular amyloid
In a prior experiment examining the effects of passive anti-Aβ immunotherapy for 1, 2 or 3 months in APP-transgenic mice killed at 21 months of age [14], we found a time-dependent reduction of both Aβ immunostaining of diffuse and fibrillar deposits and Congo-red staining of fibrillar amyloid deposits. In the current study we found a similar reduction in both Aβ immunostaining (Table 1) and total Congo-red staining (Fig. 2A, left panel; p < 0.001 frontal cortex and p < 0.01 hippocampus) after 5 months of immunotherapy. We noted that the bulk of what remained was vascular amyloid. We then separately analyzed vascular and parenchymal deposits which revealed a near 90% reduction in parenchymal deposits (p < 0.001) but a 3–4 fold elevation of vascular Congo-red staining (p < 0.0001; Fig. 2A, center and right panels, respectively). We also separately analyzed vascular and parenchymal Congo-red staining on mice from our earlier study [14], treated passively for 1, 2 or 3 months with anti-Aβ or control antibody, and found a similar result. There was a graded reduction in overall Congo-red staining nearing 75% as duration of antibody exposure increased (as reported previously; Fig. 2B). However, when separated into vascular Congo-red deposits and parenchymal deposits, there was an antibody-exposure-time-dependent increase in vascular deposition in both hippocampus and frontal cortex (Fig. 2C; p < 0.05 frontal cortex and hippocampus) and a corresponding nearly 90% decrease in parenchymal deposits (Fig. 2D; p < 0.001 in frontal cortex and hippocampus).
Table 1 Total Aβ load is significantly reduced following 5 months of anti-Aβ antibody treatment. Percent area occupied by positive immunohistochemical stain for Aβ is shown ± standard error of the mean for both the frontal cortex and hippocampus. Also shown is the percent reduction of Aβ observed following anti-Aβ antibody treatment
Region % area positive for Aβ: control treated % area positive for Aβ: anti-Aβ treated % reduction following anti-Aβ antibody treatment
Frontal Cortex 34.855 ± 2.265 9.681 ± 0.754 72
Hippocampus 23.994 ± 0.985 8.212 ± 0.596 66
Figure 2 Passive immunization with anti-Aβ antibodies decreases total and parenchymal amyloid loads while increasing vascular amyloid in frontal cortex and hippocampus of APP-transgenic mice. Panel A shows total amyloid load measured with Congo red, vascular amyloid load and parenchymal amyloid load from APP-transgenic mice administered control IgG (C) or anti-Aβ IgG (Aβ) for a period of 5 months. Panels B-D show total amyloid load (Panel B), vascular amyloid load (Panel C) and parenchymal amyloid load (Panel D) from APP-transgenic mice administered control IgG for 3 months (Cont IgG) or anti-Aβ IgG for a period of 1, 2, or 3 months (Anti-Aβ IgG). For all panels, the solid bar and solid line represent values from the frontal cortex, while the open bar and dashed line represent values from the hippocampus. ** p < 0.01.
These differences were readily observed examining micrographs of sections from these mice. Mice treated with control antibodies revealed occasional cortical vascular amyloid deposits (22 months, Fig. 3A, 28 months, Fig. 3C), while mice administered anti-Aβ antibodies had increased amounts of vascular amyloid staining (3-month treatment, Fig 3B; 5-month treatment, Fig 3D). Those vessels containing amyloid following treatment with anti-Aβ antibody also exhibited apparent increases in microglial activation as measured by CD45 expression (Fig. 3F) compared to mice treated with control antibody (Fig. 3E). Unfortunately, the shifting numbers and sizes of vascular and parenchymal deposits caused by the antibody therapy greatly complicated measurement of microglial activation per vascular deposit area so that this apparent increase in staining intensity could not be quantified accurately.
Figure 3 Increased Congo red staining of blood vessels following anti-Aβ antibody administration is associated with activated microglia. Panels A and B are from the frontal cortex of 22-month-old APP-transgenic mice immunized for 3 months with either control antibody (3A) or anti-Aβ antibody (3B). Panels C and D are from the frontal cortex of 28-month-old APP-transgenic mice immunized for 5 months with either control antibody (3C) or anti-Aβ antibody (3D). Panels E and F show a high-magnification image of CD45 immunohistochemistry (black) counterstained with Congo red (red) from 28-month-old APP-transgenic mice immunized for 5 months with either control antibody (Panel E) or anti-Aβ antibody (Panel F). Panels A-D, magnification = 100X. Scale bar in Panel B = 50 μ for panels A-D. Panels E-F, magnification = 200X. Scale bar in Panel E = 25 μm for panels E-F.
Passive amyloid immunotherapy causes increased microhemorrhage
We used the Prussian blue histological stain to label hemosiderin, a ferric oxide material produced in the breakdown of hemoglobin. Extravenous blood in the brain leads to microglial phagocytosis of the erythrocytes and breakdown of the hemoglobin within them. These ferric oxide-containing microglia are thus markers of past hemorrhage. In untreated, aged APP-transgenic mice we observed very few profiles positive for Prussian-blue staining in the frontal cortex (section counterstained with neutral red; Fig. 4A). However, following anti-Aβ antibody treatment for 5 months we observed an increase in the number of Prussian-blue profiles in the frontal cortex, which were readily detectable at a low magnification in the microscope (Fig. 4B). In the absence of anti-Aβ treatment, or even when treated with antibody for one month, most vessels did not stain with Prussian blue, and could be identified only using the red counterstain (Fig. 4C). However, even with 3 months of anti-Aβ antibody treatment we observed frequent vessels with associated Prussian-blue staining (Fig 4D). Using adjacent sections stained for Congo red, we confirmed that all vessels showing microhemorrhage contained amyloid (Figs. 4E and 4F; we were unable to double-label Prussian blue-stained sections with either Congo red or thioflavine-S). However, only a minority of vessels containing amyloid demonstrated hemorrhage.
Figure 4 Microhemorrhage associated with CAA following systemic administration of anti-Aβ antibodies. Panels A and B are low magnification images of the frontal cortex of APP-transgenic mice receiving either control antibodies (Panel A) or anti-Aβ antibodies (Panel B) for a period of 5 months. Panels C and D show representative images of amyloid containing vessels stained for Prussian blue (blue), counterstained with neutral red (red), from APP-transgenic mice receiving either control antibodies (Panel C) or anti-Aβ antibodies (Panel D) for a period of 3 months. Panel E shows a blood vessel in the frontal cortex stained for Prussian blue (blue), counterstained with neutral red, from an APP transgenic mouse administered anti-Aβ antibodies for 5 months. Panel F shows the same blood vessel on an adjacent section stained for Congo red, indicating that the blood vessel does in fact contain amyloid. Scale bar panel A = 120 μm for panels A-B. Scale bar panel C = 25 μm for panels C-D. Scale bar in panel F = 25 μm for panels E-F.
When we counted the number of Prussian blue-positive profiles in those animals receiving control antibody there was an average of one profile per every two sections (Fig. 5) and this number remained the same in both control groups (aged 22 or 28 months). Following treatment with anti-Aβ antibody for a period of two months we observed a striking increase in Prussian-blue staining, approximately five times that observed in either the control group or the mice immunized for one month (Fig. 5, p < 0.001). Following this initial increase in Prussian-blue staining, we observed a linear increase in staining associated with increasing duration of anti-Aβ antibody treatment (Fig 5). Five months of anti-Aβ antibody treatment demonstrated a six-fold increase in Prussian-blue staining when compared the control groups (Fig. 5).
Figure 5 Number of Prussian blue-positive profiles increases with duration of anti-Aβ antibody exposure. The graph shows quantification of the average number of Prussian blue-positive profiles per section from mice administered control IgG for 3 or 5 months (Cont) or anti-Aβ IgG for 1, 2, 3 or 5 months (anti-Aβ). ** p < 0.01.
Discussion
Earlier studies with vaccines against the Aβ peptide demonstrated protection from the learning and memory deficits associated with amyloid accumulation in APP-transgenic mice [14,19]. Passive immunization protocols with anti-Aβ antibodies also produced cognitive benefits, in some cases even in the absence of significant reduction in amyloid burden [12,13]. Our recent work found that 3 months of anti-Aβ treatment of 18-month-old APP-transgenic mice improved spontaneous alternation performance on the Y-maze [14]. In the present work we confirmed that passive anti-amyloid immunotherapy can reverse spatial learning deficits in APP-transgenic mice and that this benefit of immunotherapy is retained, even in aged mice (26 and 28 months old at testing) with long-established amyloid pathology.
Additionally, we describe a more rapid means of testing spatial reference memory to reveal learning and memory deficits in APP-transgenic mice. This two-day version of the radial arm water maze included greater spacing of individual trials (mice spent time in their home cage after every trial), combined with less spacing of aggregate trials (fifteen trials per day rather than four or five) to facilitate learning of platform location in the nontransgenic mice, with a clear absence of learning in the age-matched transgenic mice.
A substantial reduction in total Congophilic amyloid deposits was observed in old APP-transgenic mice treated with anti-Aβ antibodies for 2 or more months. This measurement of total Congo-red staining included both parenchymal and vascular amyloid staining. When we analyzed the sections for only vascular amyloid (CAA) we found that this measure was significantly increased following 2, 3 and 5 months of anti-Aβ antibody treatment. The remaining parenchymal amyloid load was almost completely eliminated with this antibody approach. Clearly, because total amyloid load was significantly reduced not all amyloid was shifted into the vessels; but, it appears that at least some of the Congophilic material was redistributed to the vasculature. At the present time the mechanism for this redistribution is unclear. However, one possibility is that the microglia associated with the antibody-opsonized amyloid, either by phagocytosis or surface binding, and transported the material to the vasculature, possibly in an attempt to expel it. We and others have shown evidence for microglial involvement in the removal of amyloid using both intracranial anti-Aβ antibody injections [11,21] and systemically administered anti-Aβ antibody treatment [14], as well as ex vivo studies [10,22]. Here we also report our impression that microglia surrounding CAA vessels in immunized mice expressed more CD45 than control transgenic mice. This increased expression could be due to either increased expression in the same number of microglial cells or an increased number of microglial cells in these animals. It is feasible that this microglial activation was simply in reaction to the presence of increased amyloid in the blood vessels. However, it is equally likely that microglia activated by the opsonized material migrated to the vessels for disposal of the amyloid.
Cerebral amyloid angiopathy (CAA) is defined as the deposition of congophilic material in meningeal and cerebral arteries and arterioles (capillaries and veins can also show CAA but less frequently), and it occurs to some extent in nearly all Alzheimer's disease patients [23]. Severe CAA, affecting about 15% of cases, can be associated with both infarction and hemorrhagic injury [24,25]. It has also been shown that the severity of CAA can be directly linked to the severity of dementia in Alzheimer's disease patients [26].
In the current study we found a significantly increased number of microhemorrhages in the brain as detected by Prussian-blue staining, associated with the increase in CAA following passive immunization. Another transgenic mouse model of amyloid deposition, the APP23 mice, have been shown to deposit amyloid in both brain parenchyma and blood vessels and show a CAA associated increase in spontaneous cerebral hemorrhages [27]. Moreover, Pfeifer et al. [16] showed that these spontaneous hemorrhages were significantly increased following 5 months of passive immunization of 21-month-old APP23 mice using an anti-Aβ antibody with an N-terminal epitope, similar to those typically developed in active immunization with vaccines [4,28,29]. When young mice (6 months of age) were immunized following the same protocol, no hemorrhages were observed. More recently, DeMattos et al. [30] showed that passive immunization with an N-terminal antibody (3D6: directed against amino acids 1–5 of Aβ) of PDAPP transgenic mice also resulted in significantly increased microhemorrhage. They were unable to detect increased microhemorrhage with a mid-domain antibody (266: directed against amino acids 13–28 of Aβ). Notably, antibody 266 fails to bind Aβ deposited in CAA vessels or amyloid plaques [31]. Importantly, Ferrer et al. [7] noted the presence of CAA and microhemorrhage in the brain of one patient that participated in the Aβ-vaccine trial, even though the parenchymal amyloid appeared lower than expected. Also, Nicoll et al. [6] noted that CAA appeared unaffected in the brain of another patient that participated in the Aβ-vaccine trial.
It remains to be determined whether these observations regarding increased CAA and microhemorrhage in transgenic mice are relevant to trials of passive immunotherapy in humans. It should be noted that, in spite of extending the period of immunotherapy to 5 months, there was no discernable loss of the cognitive benefits of immunotherapy in the transgenic mice, all of whom showed increased microhemorrhage. While the observation that antibody 266 does not result in vascular leakage encourages testing of this idiotype, data from the Zurich cohort of the Aβ vaccine trial argue that brain-reactive antibodies may be important for cognitive benefits [32].
Conclusions
Our opinion is that these results suggest that passive immunotherapy against Aβ should proceed with appropriate precautions taken to minimize the risk of hemorrhage (e.g., by excluding patients taking anticoagulants) and instituting measures to detect such hemorrhages if they do occur, irrespective of the antibody specificity or proclivity for microhemorrhage in aged APP-transgenic mice.
List of abbreviations
Aβ : Amyloid-beta.
APP: Amyloid precursor protein
CAA: Cerebral amyloid angiopathy.
IgG1: Immunoglobulin G type 1.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
DMW treated the mice, performed the behavioral analysis, processed the tissue and performed pathological analyses, and drafted the manuscript. ARojiani evaluated slides and provided expert opinion regarding CAA and microhemorrhage. ARosenthal and SS developed, produced and purified the antibodies used in the studies. MJF performed DNA extraction and PCR for genotyping of the mice. MNG oversees the breeding colony generating mice for the studies, collected samples from the mice and assisted in editing the manuscript. DM conceived the design of the study, guided data interpretation and assisted in editing the manuscript.
Acknowledgements
This work was supported by National Institutes of Aging / NIH grants AG15490 (MNG) and AG18478 (DM). DMW is the Benjamin Scholar in Alzheimer's disease research. We would like to thank Keisha Symmonds who aided in histological processing of the tissue and Nedda Wilson who was responsible for animal husbandry during the study. We would also like to thank Lori Lutz for assisting in editing the manuscript.
==== Refs
Iwatsubo T Odaka A Suzuki N Mizusawa H Nukina N Ihara Y Visualization of A beta 42(43) and A beta 40 in senile plaques with end-specific A beta monoclonals: evidence that an initially deposited species is A beta 42(43) Neuron 1994 13 45 53 8043280 10.1016/0896-6273(94)90458-8
Gordon MN Holcomb LA Jantzen PT DiCarlo G Wilcock D Boyett KW Connor K Melachrino J O'Callaghan JP Morgan D Time course of the development of Alzheimer-like pathology in the doubly transgenic PS1+APP mouse Exp Neurol 2002 173 183 195 11822882 10.1006/exnr.2001.7754
Solomon B Koppel R Hanan E Katzav T Monoclonal antibodies inhibit in vitro fibrillar aggregation of the Alzheimer beta-amyloid peptide Proc Natl Acad Sci USA 1996 93 452 5 8552659 10.1073/pnas.93.1.452
Schenk D Barbour R Dunn W Gordon G Grajeda H Guido T Hu K Huang J Johnson-Wood K Khan K Kholodenko D Lee M Liao Z Lieberburg I Motter R Mutter L Soriano F Shopp G Vasquez N Vandevert C Walker S Wogulis M Yednock T Games D Seubert P Immunization with amyloid-beta attenuates Alzheimer-disease-like pathology in the PDAPP mouse Nature 1999 400 173 177 10408445 10.1038/22124
Orgogozo JM Gilman S Dartigues JF Laurent B Puel M Kirby LC Jouanny P Dubois B Eisner L Flitman S Michel BF Boada M Frank A Hock C Subacute meningoencephalitis in a subset of patients with AD after Abeta42 immunization Neurology 2003 61 46 54 12847155
Nicoll JA Wilkinson D Holmes C Steart P Markham H Weller RO Neuropathology of human Alzheimer disease after immunization with amyloid-beta peptide: a case report Nat Med 2003 9 448 452 12640446 10.1038/nm840
Ferrer I Boada Rovira M Sanchez Guerra ML Rey MJ Costa-Jussa F Neuropathology and pathogenesis of encephalitis following amyloid-beta immunization in Alzheimer's disease Brain Pathol 2004 14 11 20 14997933
Weksler ME Immunology and the elderly: an historical perspective for future international action Mech Ageing Dev 1997 93 1 6 9089565 10.1016/S0047-6374(96)01822-2
DeMattos RB Bales KR Cummins DJ Dodart JC Paul SM Holtzman DM Peripheral anti-Aβ antibody alters CNS and plasma Aβ clearance and decreases brain Aβ burden in a mouse model of Alzheimer's disease Proc Natl Acad Sci USA 2001 98 8850 8855 11438712 10.1073/pnas.151261398
Bard F Cannon C Barbour R Burke RL Games D Grajeda H Guido T Hu K Huang J Johnson-Wood K Khan K Kholodenko D Lee M Lieberberg I Motter R Nguyen M Soriano F Vasquez N Weiss K Welch B Seubert P Schenk D Yednock T Peripherally administered antibodies against amyloid β-peptide enter the central nervous system and reduce pathology in a mouse model of Alzheimer's disease Nat Med 2000 6 916 919 10932230 10.1038/78682
Wilcock DM DiCarlo G Henderson D Jackson J Clarke K Ugen KE Gordon MN Morgan Intracranially administered anti-Aβ antibodies reduce β-amyloid deposition by mechanisms independent of and associated with microglial activation J Neurosci 2003 213 3745 3751 12736345
Dodart JC Bales KR Gannon KS Greene SJ DeMattos RB Mathis C DeLong CA Wu S Wu X Holtzman DM Paul SM Immunization reverses memory deficits without reducing brain Abeta burden in Alzheimer's disease model Nat Neurosci 2002 5 452 457 11941374
Kotilinek LA Bacskai B Westerman M Kawarabayashi T Younkin L Hyman BT Younkin S Ashe KH Reversible memory loss in a mouse transgenic model of Alzheimer's disease J Neurosci 2002 22 6331 6335 12151510
Wilcock DM Rojiani A Rosenthal A Levkowitz G Subbarao S Alamed J Wilson D Wilson N Freeman MJ Gordon MN Morgan D Passive amyloid immunotherapy clears amyloid and transiently activates microglia in a transgenic mouse model of amyloid deposition J Neurosci 2004 24 6144 6151 15240806 10.1523/JNEUROSCI.1090-04.2004
Oddo S Billings L Kesslak JP Cribbs DH LaFerla FM Abeta Immunotherapy Leads to Clearance of Early, but Not Late, Hyperphosphorylated Tau Aggregates via the Proteasome Neuron 2004 43 321 333 15294141 10.1016/j.neuron.2004.07.003
Pfeifer M Boncristiano S Bondolfi L Stalder A Deller T Staufenbiel M Mathews PM Jucker M Cerebral hemorrhage after passive anti-Aβ immunotherapy Science 2002 298 1379 12434053 10.1126/science.1078259
Holcomb L Gordon MN McGowan E Yu X Benkovic S Jantzen P Wright K Saad I Mueller R Morgan D Sanders S Zehr C O'Campo K Hardy J Prada CM Eckman C Younkin S Hsiao K Duff K Accelerated Alzheimer-type phenotype in transgenic mice carrying both mutant amyloid precursor protein and presenilin 1 transgenes Nat Med 1998 4 97 100 9427614 10.1038/nm0198-097
Gordon MN King DL Diamond DM Jantzen PT Boyett KV Hope CE Hatcher JM DiCarlo G Gottschall WP Morgan D Arendash GW Correlation between cognitive deficits and Abeta deposits in transgenic APP+PS1 mice Neurobiol Aging 2001 22 377 385 11378242 10.1016/S0197-4580(00)00249-9
Morgan D Diamond DM Gottschall PE Ugen KE Dickey C Hardy J Duff K Jantzen P DiCarlo G Wilcock D Connor K Hatcher J Hope C Gordon M Arendash GW A beta peptide vaccination prevents memory loss in an animal model of Alzheimer's disease Nature 2000 408 982 985 11140686 10.1038/35050116
Janus C Pearson J McLaurin J Mathews PM Jiang Y Schmidt SD Chishti MA Horne P Heslin D French J Mount HT Nixon RA Mercken M Bergeron C Fraser PE George-Hyslop P Westaway D A beta peptide immunization reduces behavioral impairment and plaques in a model of Alzheimer's disease Nature 2000 408 979 982 11140685 10.1038/35050110
Wilcock DM Muniredddy SK Rosenthal A Ugen KE Gordon MN Morgan D Microglial Activation Facilitates Aβ Plaque Removal Following Intracranial Anti-Aβ Antibody Administration Neurobiol Dis 2004 15 11 20 14751766 10.1016/j.nbd.2003.09.015
Bard F Barbour R Cannon C Carretto R Fox M Games D Guido T Hoenow K Hu K Johnson-Wood K Khan K Kholodenko D Lee C Lee M Motter R Nguyen M Reed A Schenk D Tang P Vasquez N Seubert P Yednock T Epitope and isotype specificities of antibodies to beta -amyloid peptide for protection against Alzheimer's disease-like neuropathology Proc Natl Acad Sci USA 2003 100 2023 2028 12566568 10.1073/pnas.0436286100
Jellinger KA Alzheimer disease and cerebrovascular pathology: an update J Neural Transm 2002 109 813 836 12111471 10.1007/s007020200068
Olichney JM Ellis RJ Katzman R Sabbagh MN Hansen L Types of cerebrovascular lesions associated with severe cerebral amyloid angiopathy Ann NY Acad Sci 1997 826 493 497 9329731
Maurino J Saposnik G Lepera S Rey RC Sica RE Multiple simultaneous intracerbral hemorrhages: clinical features and outcome Arch Neurol 2001 58 629 632 11295994 10.1001/archneur.58.4.629
Thal DR Ghebremedhin E Orantes M Wiestler OD Vascular pathology in Alzheimer disease: correlation of cerebral amyloid angiopathy and arteriosclerosis/lipohyalinosis with cognitive decline J Neuropathol Exp Neurol 2003 62 1287 1301 14692704
Winkler DT Bondolfi L Herzig MC Jann L Calhoun ME Wiederhold KH Tolnay M Staufenbiel M Jucker M Spontaneous hemorrhagic stroke in a mouse model of cerebral amyloid angiopathy J Neurosci 2001 21 1619 1627 11222652
Dickey CA Morgan DG Kudchodkar S Weiner DB Bai Y Cao C Gordon MN Ugen KE Duration and specificity of humoral immune responses in mice vaccinated with the Alzheimer's disease-associated beta-amyloid 1–42 peptide DNA Cell Biol 2001 20 723 729 11788050 10.1089/10445490152717587
McLaurin J Cecal R Kierstead ME Tian X Phinney AL Manea M French JE Lambermon MH Darabie AA Brown ME Janus C Chishti MA Horne P Westaway D Fraser PE Mount HT Przybylski M St George-Hyslop P Therapeutically effective antibodies against amyloid-beta peptide target amyloid-beta residues 4–10 and inhibit cytotoxicity and fibrillogenesis Nat Med 2002 8 1263 1269 12379850 10.1038/nm790
DeMattos RB Boone LI Hepburn DL Parsadanian M Bryan MT Ness DK Piroozi KS Holtzman DM Bales KR Gitter BD Paul SM Racke M In vitro and in vivo characterization of beta-amyloid antibodies binding to cerebral amyloid angiopathy (CAA) and the selective exacerbation of CAA-associated microhemorrhage Neurobiol Aging 2004 25 577
Racke M Bryan MT DeMattos RB Binding differences between Abeta antibodies in CAA containing vessels from PDAPP-transgenic mice Neurobiol Aging 2004 25 588
Hock C Konietzko U Streffer JR Tracy J Signorell A Muller-Tillmanns B Lemke U Henke K Moritz E Garcia E Wollmer MA Umbricht D de Quervain DJ Hofmann M Maddalena A Papassotiropoulos A Nitsch RM Antibodies against beta-amyloid slow cognitive decline in Alzheimer's disease Neuron 2003 38 547 554 12765607 10.1016/S0896-6273(03)00294-0
| 15588287 | PMC539292 | CC BY | 2021-01-04 16:38:19 | no | J Neuroinflammation. 2004 Dec 8; 1:24 | utf-8 | J Neuroinflammation | 2,004 | 10.1186/1742-2094-1-24 | oa_comm |
==== Front
Theor Biol Med ModelTheoretical Biology & Medical Modelling1742-4682BioMed Central London 1742-4682-1-131554649210.1186/1742-4682-1-13ReviewMetabolic scaling: consensus or controversy? Agutter Paul S [email protected] Denys N [email protected] Theoretical and Cell Biology Consultancy, 26 Castle Hill, Glossop, Derbyshire, SK13 7RR, UK2 BioMedES, Hilton Campus MG7, Aberdeen AB24 4FA, UK2004 16 11 2004 1 13 13 10 7 2004 16 11 2004 Copyright © 2004 Agutter and Wheatley; licensee BioMed Central Ltd.2004Agutter and Wheatley; 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 body mass (M) and standard metabolic rate (B) among living organisms remains controversial, though it is widely accepted that in many cases B is approximately proportional to the three-quarters power of M.
Results
The biological significance of the straight-line plots obtained over wide ranges of species when B is plotted against log M remains a matter of debate. In this article we review the values ascribed to the gradients of such graphs (typically 0.75, according to the majority view), and we assess various attempts to explain the allometric power-law phenomenon, placing emphasis on the most recent publications.
Conclusion
Although many of the models that have been advanced have significant attractions, none can be accepted without serious reservations, and the possibility that no one model can fit all cases has to be more seriously entertained.
metabolic rateallometric scalingpower lawssupply networksfluid flow
==== Body
Introduction: Kleiber and metabolic scaling
In 1932, Kleiber published a paper in an obscure journal [1] showing that standard metabolic rates among mammals varied with the three-quarters power of body mass: the so-called "elephant to mouse curve", termed "Kleiber's law" in this review. Since that date, this and similar allometric scaling phenomena have been widely and often intensively investigated. These investigations have generated continuing debates. At least three broad issues remain contentious, each compounded on the one hand by the problem of obtaining valid data (in particular, finding procedures by which reliable and reproducible measures of standard metabolic rate can be obtained, especially in poikilotherms) and on the other by statistical considerations (in particular, the validity of fitting scattered points to a straight line on a semi-logarithmic plot).
The first issue is disagreement as to whether any consistent relationship obtains between standard metabolic rate and body mass. Moreover, those who acknowledge such a relationship hold divergent opinions about its range of application. Is it valid only for limited numbers of taxa, or is it universal? Since the 1960s there has been a measure of consensus: a consistent allometric scaling relationship does exist, at least among homoiotherms. Nevertheless, not all biologists agree, and scepticism is widespread, particularly about the alleged universality of Kleiber's law.
Second, assuming that some version of Kleiber's law (a consistent metabolic scaling relationship) applies to at least some taxa, there are disagreements about the gradient of the semi-log plot. That is, if B = aMb, where B = standard metabolic rate, M = body mass, and a and b are constants, what is the value of b? Kleiber [1] and many subsequent investigators claimed that b = 0.75, and on this matter too a measure of consensus has obtained since the 1960s. Once again, however, not all biologists agree. A significant minority of investigators hold that b = 0.67; and other values have been suggested, at least for some organisms.
Third, assuming a consistent scaling relationship and an agreed value of b, how is Kleiber's law to be interpreted mechanistically? What is its physical or biological basis? For those who claim that b = 0.67, this issue is simple: standard metabolic rate depends on the organism's surface to volume ratio. But for proponents of the majority view, that b = 0.75, the issue is not simple at all. Many interpretations have been proposed, and since several of these are of recent coinage and seem to be mutually incompatible, a critical comparative review seems timely.
Kleiber's initial paper [1] found support within a decade. The allometric scaling relationship B = aMb (B = standard metabolic rate, M = body mass, a and b are constants and b is taken to be approximately 0.75), was inferred by other investigators during the 1930s [2,3]. Relevant data have been reviewed periodically since then (e.g. [4-15]) and recent developments have rekindled interest in the field.
Many biological variables other than standard metabolic rate also reportedly fit quarter-power scalings (relationships of the kind V = kMb, where V is the variable in question, k is a constant and b = n/4; n = 3 for metabolic rate). Examples include lifespans, growth rates, densities of trees in forests, and numbers of species in ecosystems (see e.g. [9]). Some commentators infer that Kleiber's law is, or points to, a universal biological principle, which they have sought to uncover. Others doubt this, not least because it is unclear how (for example) tree densities can be consequences of metabolic scaling or can have the same mechanistic basis. This article focuses on the metabolic rate literature, mentioning other variables only in passing, because most debates in the field have arisen from metabolic rate measurements.
Variations in the value of b
Most debates about the value of b assume some version of Kleiber's law: i.e. that a single allometric scaling relationship fits metabolic rates over a wide range of organisms. However, as noted in the introduction, there are dissenters. Everyone acknowledges considerable variation both within and among taxa, no matter whether b = 0.75, 0.67, or some other number. The question is whether these variations are deviations from a general law, or whether there is no such law. Conflicting opinions on this fundamental point recall the traditional philosophical difference between physicists and biologists: the former are inclined to see abstract mathematical generalities in any set of numerical data, the latter to see concrete particulars. All recent attempts to explain Kleiber's law by "universal" models have involved physicists and mathematicians; the sceptics are predominantly biologists.
Dodds et al. [16] re-examined published scaling data from Kleiber's original paper onwards and concluded that the consensus (b = 0.75) was not statistically supported. Feldman [17] found no evidence for any wide-ranging allometric power law in biology and dismissed all attempts to explain scaling relationships by physical or mathematical principles. Atanasov and Dimitrov [18] found evidence that b ranges from around 0.67 to more than 0.9 over all major animal groups, the values perhaps reflecting complexity of organisation; single values such as 0.75 emerge only as averages over each group. Other investigators have been less sceptical; publications by Enquist and Niklas [19,20] give particularly impressive support to the generality of Kleiber's law because Niklas was previously among the doubters.
Whatever one's position, it is indisputable that the Kleiber relationship has many exceptions, even among mammals. Bartels [21] showed that some mammals, such as shrews, have B values well above those expected from the Kleiber curve. Andersen [22] discussed the high B values for whales and seals and attributed them to the cold-water habitat. Nevertheless, Kleiber's law has been extended beyond placental mammals to birds and marsupials. Birds have generally higher a values than placental mammals and marsupials have lower ones, but the 0.75-power relationship is still inferred by many investigators (e.g. [4]). McNab [13] accepted Kleiber's law as a general approximation but emphasized species variations, which he attributed to differences in diet, habitat and physiological adaptation. Elgar and Harvey [23] also found variability among groups of species but reasoned that standard metabolic rates vary taxonomically rather than with temperature regulation, food intake or activity. Economos [24] was also critical of McNab, at least in respect of mammals.
It is difficult to define "standard metabolic rate" in poikilotherms; ambient temperature, time since last meal and other variables markedly affect measurements [9,13,25]. A heterogeneous array of poikilotherm data [5] revealed an "average" b value of roughly 0.75. There were wide divergences in some taxa; notwithstanding these, Hemmingsen [4,5] argued that over all animals, plants and protists, metabolic rate scales as the 0.75-power of body mass. More recently published data [26,27] support this conclusion for a wide range of organisms and body masses. However, a careful re-evaluation of Hemmingsen's data by Prothero [28] cast further doubt on the applicability of Kleiber's law to unicellular organisms. Scepticism persists, mostly on the grounds of the intrinsic variability of the data, which is too often underestimated because it is disguised in the customary logarithmic plots and is seldom subjected to adequate statistical analysis [11,29]. However, this too has been debated; a suitable choice of procedures for estimating parameters might eliminate inconsistencies and discrepancies from the data, giving more credence to the belief that b = 0.75 over a wide range of taxa [30]. In the following section we shall examine some of the more divergent data in more detail.
In short, there is a clear but by no means total consensus that (i) Kleiber's law is widely (even universally) applicable in biology, (ii) b is approximately 0.75. Variability in the data is generally admitted, so the consensus – and the claim that Kleiber's law manifests a general biological principle – can legitimately be doubted.
The mass transfer model [31]
Some of the doubts about the consensus are powerfully supported by studies on small aquatic organisms. Reviewing a large literature on metabolic rates in aquatic invertebrates and algae, Patterson [31] deployed chemical engineering principles to explain why the b values ranged from about 0.3 to 1.2 in these taxa (his Table 1 provides an excellent summary). Assuming that the delivery of nutrients to each organism entails diffusion through a boundary layer, Patterson showed how water movements and organism size might affect such delivery and hence determine metabolic rate. Using simple geometrical models of organisms (plates, cylinders and spheres), he derived b values ranging from 0.31 to 1.25, more or less consistent with the experimental values.
Patterson plotted two dimensionless numbers against each other, viz. Sherwood number, Sh = hmW/D, where hm = mass transfer coefficient, W = characteristic dimension of organism and D = diffusivity; and Reynolds number (a function of organism size), Re = ρUW/μ, where ρ = density, U = water flow speed and μ = coefficient of viscosity. The graphs, which had the form Sh = c.Red, where d = 0.5 for ideal laminar flow and 0.8 for turbulent flow (c is a constant of proportionality), revealed the relative importance of diffusion and mass transfer (convective movement) in the supply of materials. Patterson was able to derive an expression for hm, and was thus able to relate the supply of materials to body mass.
The two main attractions of this model are (1) good agreement with a wide range of data and (2) derivation from basic physical principles without ad hoc biological or other assumptions. Patterson's approach has implicit support in the literature: Coulson [32] used chemical engineering principles to argue that mammalian metabolic rates are supply-limited, but he did not develop the argument in mathematical detail. However, Patterson's model has drawbacks. First, it is hard to see how his reasoning can be generalised to other taxa, notwithstanding Coulson's proposal (discussed in a later section). Second, by focusing on diffusion and convective mass transfer, he ignored active processes in the uptake of materials, which are likely to dominate in many organisms. Third, he assumed that metabolism in general is supply-limited; in homoiotherms at least, it is more nearly demand-limited under resting conditions, though even this is an oversimplification.
The Patterson model has not been given much attention by other investigators in the field and perhaps it deserves more consideration. Despite its inherent limitations (it is exclusively concerned with small aquatic eukaryotes) it is a potentially fruitful contribution to biophysics.
Scaling of metabolic rate with surface-to-mass ratio
Several workers accept the reality of allometric scaling but question the value b = 0.75, which a consensus of physiologists has accepted since the 1960s. Many of these sceptics claim that the "true" value of b is 0.66 or 0.67 because the principal determinant of metabolic scaling is the surface-to-volume ratio of the organism; hence, assuming constant body density, the surface-to-mass ratio. The first study to suggest this explanation for the mass dependence of B is attributed to Rubner [33], who studied metabolic rates in various breeds of dog. Heusner [34] reported that b is approximately 0.67 for any single mammalian species and suggested that the interspecies value of 0.75 is a statistical artefact. Feldman and McMahon [35] disagreed, but Heusner sustained his position in subsequent articles. For instance, reviewing a substantial body of published data [36], he argued that metabolic rate data for small and large mammals lie on parallel regression lines, each with a gradient of approximately 0.67 but with different intercepts (i.e. values of a, termed the "specific mass coefficients"). Hayssen and Lacy [37] found b = 0.65 for small mammals and b = 0.86 for large ones, again suggesting that b = 0.75 is a cross-species "average" with no biological significance; but it is questionable whether their data were measurements of standard metabolic rate in all cases. McNab [13] reported lower values: 0.60 and 0.75, respectively. Heusner [36] reasoned that if a few large mammals are added to a sample of predominantly small ones, a single regression line for all the data might have a gradient around 0.75. This, however, is misleading, as the following paragraphs will argue.
According to Heusner, the ratio B/M0.67 is a mass-independent measure of standard metabolism. Variations indicate the effects of factors other than body mass. Other workers broadly share Heusner's opinion (see e.g. [12] for review and [38] for a good recent exemplar). Bartels [21] found a value of 0.66 for mammals; Bennett and Harvey [39] reported 0.67 for birds. Of course, if B varies as M0.67, the interesting problem is not the index (b) in the Kleiber equation but the allegedly constant relationship between specific mass coefficient (a) and body size. This point was developed by Wieser [40], who distinguished the ontogeny of metabolism, which comprises several phases but follows the surface rule (M0.67) overall, from the phylogeny of metabolism, which concerns the mass coefficients (a). Following Heusner's argument, Wieser [40] wrote the allometric power law in the form B = anM0.66 and deduced that the specific mass coefficient an = aM0.09. Here, a is an interspecific mass coefficient (3.34 w in mammals if M is in kg). Another difficulty with this type of explanation lies in the calculation of body surface area; the Meeh coefficient, k, where surface area = kM0.67, is difficult to measure unequivocally but is generally taken as ~10 (see [3]). Yet another possible difficulty was identified by Butler et al. [41], who questioned Heusner's dimensional analysis argument and concluded that no version of Kleiber's law (i.e. no value of b that is constant over a range of species) could be substantiated by his approach.
The claim that b = 0.67 remains a minority view. Those who accept it are faced with the twin difficulties of (i) establishing that their estimates of surface area are correct and (ii) explaining why, in Wieser's notation, an = aM0.09. Moreover, even if such arguments as Heusner's are valid for homoiotherms, it is hard to justify their extrapolation to poikilothermic animals, plants and unicellular organisms, all of which are held by consensus to fit Kleiber's law (but see the two preceding sections). Why should temperature fluxes across the body surface be the main determinants of metabolic rate in poikilotherms, particularly microorganisms? Even in mammals, maintenance of body temperature might not be the main contributor to energy turnover at rest (see later). Contrary to the view of Dodds et al. [16], therefore, b = 0.67 cannot be treated as a "null hypothesis".
Throughout the remainder of this article, the consensus position will be assumed: Kleiber's law is valid for a wide range of organisms, and b = 0.75. This assumption is made tacitly and provisionally and does not imply dismissal of the foregoing sceptical arguments; but a field can only be reviewed coherently from the consensus point of view.
McMahon's model [42]
A vertical column displaced by a sufficiently large lateral force buckles elastically. The critical length of column, lcr, = k(E/ρ)1/3d2/3, where d = column diameter, E = Young's modulus and ρ = density. If E and ρ are constant then lcr3 = cd2, where c is a constant of proportionality. McMahon [42] applied this reasoning to bone dimensions for stationary quadrupeds. In a running quadruped the limbs support bending rather than buckling loads but the vertebral column receives an end thrust that generates a buckling load. It follows that all bone proportions change in the same way with animal size. The mass of a limb, wl, = αld2, where α is a constant. If wl is proportional to M, as it generally must be, then M = βld2, where β is another constant. Hence (given the above relationship between l and d) M is proportional to l4, implying that l is proportional to wl1/4; hence d is proportional to wl3/8, or M3/8. Empirical support for this relationship appeared in [43].
McMahon [42] also applied this argument to muscles. The work done by a contracting muscle, W, is proportional to σAΔl, where σ is tensile strength, A is the cross-sectional area and Δl is the length change during contraction. The power developed, W/t (t = time), is therefore σAΔl/Δt. Since σ and Δl/Δt are roughly constant and independent of species, W/t varies with A; and since A is proportional to d2, W/t it is proportional to d2, and therefore to (M3/8)2 = M3/4. If this deduction applies to any skeletal muscle (as seems plausible), then it applies to the entire set of metabolic variables supplying the muscular system with nutrients and oxygen. Hence, B varies as M3/4. A broadly comparable but simpler argument was advanced by Nevill [44]; large mammals have proportionately more muscle mass than smaller ones. If the contribution of the muscle to B (which Nevill assumes is proportional to M) is partialled out, then the residual B is proportional to M2/3. Nevill's paper is seldom cited.
One difficulty with McMahon's model is that little of the energy turnover under conditions of standard metabolic rate measurement entails muscle contraction. The model might still be valid if maximum metabolic rate followed the same allometric scaling law as B; this has been widely believed, and Taylor et al. [45] adduced evidence for it. However, recent detailed studies [46-48] indicate that maximum metabolic rate in birds and mammals scales as M0.88, not M0.75, although there are disagreements about whether aerobic capacity determines the allometry of maximum metabolic rate [48,49]. Weibel [50] presented a large set of data to this effect. (On the other hand, there are reports that in birds the index decreases rather than increases with increasing metabolic output, e.g. [58].) Another drawback of the McMahon model is that it cannot apply to organisms without muscles, such as protists. This perhaps explains why McMahon's elegant deduction has been largely ignored in recent debates about Kleiber's law.
The Economos model [51]
An increased gravitational field increases energy metabolism in animals [52,53]. Work against gravity is proportional to M1.0. If maintenance metabolism were related to surface area (proportional to M0.67) then a combination of the two effects, surface-to-mass ratio and work against gravity, might explain the observed M3/4 relationship. This model [51] is difficult to assess: it is not clear why the two proposed factors, surface area dependence and gravitational loading, should combine for all animals (and other taxa) in just the right proportions to generate a 0.75-power dependence on body mass. To take just one example, aquatic microbes are more affected by Brownian motion than by gravity, so why should they show the same balance between surface-to-mass ratio and gravitational effects as mice or elephants? Pace et al. [54] suggested that the Economos model could be critically tested under conditions of weightlessness in space. No corroboration (or refutation) by studies on astronauts has been reported.
Allometric scaling in cells and tissues
Before more recent models purporting to explain Kleiber's law are discussed, some comments are needed on scaling of metabolism at the organ, tissue and cell levels. Belief that the Kleiber relationship can be explained in terms of the inherent properties of the cells dates from the 1930s [3,55] and persists (e.g. [56,57].
Standard metabolic rate (B) is usually measured as oxygen consumption rate, which correlates with nutrient utilization [9,15] and rates of excretion of nitrogenous and other wastes [2]; so research in the field has been dominated by respiratory studies. Lung volume, trachaeal volume, vital capacity and tidal volume all scale as M but respiratory frequency varies as M-0.31, ventilation rate as M0.77 and oxygen consumption rate as M0.72 [58-60]. All mammals extract a similar percentage of oxygen (~3%) from respired air [9]. The significance of "pulmonary diffusion capacity" has been debated; it scales as M1.0 so it is disproportionate in bigger animals [17,61-65].
Stahl [60] described the scaling of cardiovascular and haematological data. Blood haemoglobin concentration is the same for all mammals except those adapted to high altitudes. Blood volume is ~6–7% of body volume for all mammals except aquatic ones. Erythrocyte volume varies with species but bears no obvious relationship to M. The oxygen affinity of haemoglobin varies with body size, being lower in smaller mammals, which unload oxygen to their tissues more rapidly. Capillary density is more or less constant in mammals with bodies larger than a rat's, though it is greater in the smallest mammals [65]. The heart accounts for ~0.6% of body mass in all mammals [66]. Heart rate scales as M-0.25, cardiac output as M0.81 (60) and circulation time as M0.25. The energy cost of supplying the body with 1 ml of oxygen is similar for all mammals [15].
Standard metabolic rate has two main components: service functions, e.g. the operation of heart and lungs; and cellular maintenance functions, e.g. protein and nucleic acid turnover (e.g. [67]). Krebs [68] elucidated this second component by studying tissue slices; his investigation has since been extended. Oxygen consumption per kg decreases with increasing M in all tissues, but tissues do not all scale identically. Horse brain and kidney have half the oxygen consumption rates of mouse brain and kidney but the difference between these species in respect of liver, lung and spleen is 4-fold [68-70]. Metabolic rate in liver scales as M0.63; for some organs the exponent is closer to 1.0; the sum of oxygen consumption rates over all tissues gives – approximately – the expected 0.75 index [71]. The difficulty of recalculating B from tissue-slice data is considerable, so the Martin and Fuhrman calculation [71] has wide confidence limits. Spaargen [72] suggested that tissues that use little oxygen constitute different percentages of body mass in large and small mammals, leading to a distortion of the surface law (B = M2/3), which would otherwise be valid. More recently, however, Wang et al. [73] repeated the Martin and Fuhrman calculation using improved data, and found impressive support for the consensus B = M3/4.
Cells of any one histological type are size-invariant among mammals but allometric scaling is reported at the cellular level; e.g. the metabolic rate of isolated hepatocytes scales as M-0.18 [74]. Numbers of mitochondria per gram of liver (or per hepatocyte), however, scale as M-0.1 [75,76]. The apparent discrepancy between these values might be illusory (cf. [77]), or it might indicate a greater proton leak in mitochondria from livers of smaller animals [78] or allometry in redox slip [79]. Also, larger animals have smaller inner mitochondrial membrane surface areas (the scaling is M-0.1) and different fatty acid compositions [71]. The discrepancy between the scalings of hepatocyte and whole-body metabolism is probably explained by the decrease in liver mass, which scales as M0.82 [75,80]. Combining liver mass with hepatocyte oxygen consumption, the derived scaling for liver metabolism is M0.82.M-0.18 = M0.64, consistent with the experimental tissue-slice data (M0.63; see above). Combining liver mass with mitochondrial number per hepatocyte gives a similar value [77]. Cytochrome c and cytochrome oxidase contents scale roughly as M0.75 [81-85]. The allometric scaling of mitochondrial inner membrane area, and the body-size-related differences in unsaturated fatty acid content, remain unexplained.
Isolated mammalian cells reportedly attain the same mitochondrial numbers and activities after several generations in culture, irrespective of the tissue of origin or the organism's body mass [86-88]. If allometric scaling is lost at the cellular level after several generations in vitro, then presumably mitochondrial densities, inner membrane areas and cytochrome levels somehow become "normalized". This is a readily testable prediction [see [89]], but it does not appear to have been subjected to critical experiments. If it is corroborated there will be interesting mechanisms to investigate.
The main conclusions from this section are: (a) different organs make different contributions to the scaling of whole-organism metabolic rates; (b) differences at the cellular level make relatively small contributions to scaling at the organ level; (c) these differences at cellular level might disappear altogether after several generations in culture. The most striking conclusion is (b). It implies that allometric scaling of metabolic rate does not after all, for the most part, reside in cellular function but at higher levels of physiological organisation. If this is the case, then the alleged applicability of Kleiber's law to unicellular organisms is called into question.
Resource-flow models
Coulson's flow model [42] was mentioned earlier. It relates tissue or organ oxygen consumption rates to circulation times, i.e. to the rate of supply of oxygen and nutrients, and these scale as M0.25 (see previous section). Coulson's approach contrasts with traditional biochemical measurements: the principal variable is not the concentration of a resource but the supply rate; metabolic activity depends on encounter frequency not concentration. This perspective merits further development, particularly by extension to the cell internum [89-93]. Obviously, it is within the cell that the reactant molecules are passed over the catalysts; and the flow rate increases with the cell's metabolic activity, as Hochachka [93] cogently described.
However, flow theories advanced to explain Kleiber's law have not followed this line of argument. Banavar et al. [94,95] and Dreyer and co-workers [27,96] have shown that the Kleiber relationship can be deduced from the geometries of transport networks, without reference to fluid dynamics. Broadly, these authors argue that as a supply network with local connectivity branches from a single source (in a mammalian circulatory system, the heart is the source), the number of sites supplied by the network increases. Natural selection has optimized the efficiency of supply. A general relationship can be derived between body size and flow rate in the network: delivery rates per unit mass of tissue vary with the quarter-power of body size (M), implying the validity of Kleiber's law.
The most detailed account of this argument [95] begins with the reasonable assumption that M scales with LD, where L is the physical length of the organism and D is its dimensionality. It proceeds with a theorem: the sum of flows through all parts of the network, F, is proportional to the (dimensionless) length multiplied by the metabolic rate. A quantity measuring the total flow of metabolites per unit mass of organism is then defined: r1 = F/M. r1 (which has units of inverse time) measures the dependence of the network's geometry on body mass, so it indicates the energy cost of metabolite delivery. Another parameter, r2, measures the metabolite demand by the tissues: r2 = the dimensionless length of the "service volume" (the amount of tissue that consumes one unit of metabolite per unit time). It is then deduced that B is proportional to (Mr1/r2)D/(D+1). Provided that r1 and r2 change proportionately – i.e. supply always matches demand – then for a three-dimensional organism, Kleiber's law follows. According to Banavar et al. [94], deviations from Kleiber's law indicate inefficiency or some physiological compensation process.
This model has been criticized [97] because the assumed network does not resemble (e.g.) the mammalian circulatory system, where only terminal nodes (capillaries), not all nodes (as the model implies), are metabolite exchange sites. Also, the model seems to predict that r1/r2 will decrease as B rises from standard to maximal; but the best data suggest the opposite trend (see earlier discussion: [46-48]). Banavar et al. do not explicitly allow for differences among tissue types, which are considerable (see above), except perhaps in terms of rather implausible variations among r1/r2 ratios. On the other hand, the model is simple and flexible and it reflects recent developments in the physics of networks. If it could be applied to flow at the cellular level, it might accord with the requirements discussed at the beginning of this section; though it is difficult to see how this can be achieved.
Rau [98] also advanced a fluid-flow model, but his conception is physical not geometrical. Assuming Pouseille flow through an array of similar tubes, such as capillaries, and a roughly constant flow speed, Rau used scaling arguments to derive the relationship t = kM1/4, where t is the transport time and k is a constant. If the fluid transport rate (essentially the reciprocal of t) is proportional to B/M, Kleiber's law follows. However, Rau's model appears to assume that because metabolic rate is energy per unit time, it can be equated with the product of fluid volume flow rate and pressure (since energy is equal to pressure times volume). This assumption, which appears to be based exclusively on dimensional analysis, is fallacious.
Four-dimensional models
Blum [99] observed that the "volume" of an n-dimensional sphere of radius r is V = πn/2rn/Γ(n/2 + 1), and that A = dV/dr = nπn/2rn-1/Γ(n/2 + 1). Here, Γ(n) is the gamma-function such that Γ(n + 1) = nn, Γ(2) = 1 and Γ(3/2) = π1/2/2. Suppose two objects have "volumes" V1 and V2 and "areas" A1 and A2. From the foregoing, A1/A2 = (V1/V2)(n-1)/n; so if n = 4, a 3/4-power relationship between "volumes" (hence, masses?) emerges from a familiar mathematical principle. Might Kleiber's law therefore follow from a four-dimensional description of organisms? Speakman [100] pointed out that if n = 4, then A is volume (it has three dimensions) and V is hypervolume, the biological significance of which is obscure. However, West et al. [88,101,102] have indeed proposed a four-dimensional model to explain the Kleiber relationship, and considerable claims have been made for their account.
This model addresses the supply of materials (particularly oxygen) through space-filling fractal networks of branching tubes. It assumes that as a result of natural selection, organisms maximize their use of resources. The initial account [101] assumed that energy dissipation is minimised at all branch-points in the network and that the terminal branches are size-invariant (for instance, blood capillaries are the same lengths and diameters in mice and elephants). Kleiber's law and analogous scalings were deduced from these assumptions. In particular, the three-quarters-power exponent was shown to be inherent in the geometry of a branching network that preserves total cross-sectional area at each branch point. The circulatory systems of large animals such as mammals are not exactly area-preserving, but West et al. [101] reasoned that this objection could be circumvented by considering the pulsatile flow generated in the larger arteries by the action of the heart.
A second, simpler account [102] developed the model from a geometrical basis. The crucial feature of the branching network is the size-invariance of the terminal units. The effective exchange area, a, is a function of the element lengths at each level of the hierarchy, but one of these, the terminal one (l0), is invariant. Writing Φ as a dimensionless function of the (dimensionless) ratio l1/l2 leads to
a (l0, l1, l2,...) = l12Φ(l0/l1, l2/l1...)
Introducing a scaling factor, λ, leads to
a (l0, l1, l2,...) = λ2l12 Φ(l0/λ l1, l2/l1...)
which is not proportional to λ2 because l0 is fixed. The dependence of Φ on λ is not known a priori, but it can be parameterized as Φ(l0/λ l1, l2/l1...) = λεΦ(l0/l1, l2/l1...), where ε is between 0 and 1. This power law reflects the fractal character of the network's hierarchical organization. Similar reasoning is applied to body volume, hence body mass, and the following expression for the exchange surface area is derived:-
a = kMr, r = (2 + ε)/(3 + ε + ζ),
where k is a constant and ζ (0 < ζ < 1) is an arbitrary exponent of length, just as ε is an arbitrary exponent of area. If natural selection has acted to maximize the scaling of a, then ε must tend to 1 and ζ to 0. This gives r = 0.75. If a limits the supply of oxygen and nutrients, and hence determines standard metabolic rate, then B is proportional to a and Kleiber's law follows.
The model has several attractions: it derives from well-established physical principles, invokes natural selection and is mathematically impeccable. It implies that cells and organelles transport materials internally along space-filling fractal networks rather than by "diffusion", which seems correct [83,85,86,103]. The self-similarity of these transport networks is emphasized particularly in [88]. The dimensionalities of effective exchange surfaces, a, are predicted to be closer to 3 than 2; empirically, the microscopic convolutions of surfaces such as the mammalian intestinal mucosa are well known. The mass of the smallest possible mammal is deduced and shown to be close to the mass of the shrew. Other approaches to exchange networks, assuming minimum energy expenditure and scale-invariance, have led to similar models [104]. The model can be adapted, with no loss of rigour, to new data: Gillooly et al. [105] showed that the fractal supply network principle can be combined with simple Boltzmann kinetics to explain the effects of both body mass and temperature on metabolic rates. Since mass and temperature are the primary determinants of many physiological and ecological parameters, this work suggests that the model [88] could revolutionize biology.
This is an impressive range of successes. However, West and his co-workers make claims that are less compelling. The observation that cytochrome oxidase catalytic rates fit the same allometric curve as whole-organism metabolic rates is claimed as corroboration. However, cytochrome oxidase is not an organism, or a cell: it does not have a metabolic rate. It is also debatable whether mitochondria can be said to have "metabolic rates". (In contrast, Hochachka and Somero [106] noted that oxygen turnover in the whole biosphere can be fitted to the same curve; but they recognized this as "a contingent fact with no biological significance".) Also, the explanation derived by West and his colleagues for the alleged body-mass-invariance of the metabolic rates of cultured cells (see earlier) is mathematically neat, but it leads to no experimentally testable predictions, and the heterogeneous data sources cited in this context make the explicandum itself unconvincing. Finally, the model is said to explain the quarter-power scalings of a wide range of biological variables other than metabolic rate, including population densities of trees [19] and carnivorous animals [107], plant growth rates, vascular network structure and maturation times [18,108], and life-spans [88]. It is not clear why any of these variables should depend on the fractal geometries of space-filling supply networks, still less on metabolic rates; though there is widespread interest in the application of scaling laws in ecology, for instance in modelling biodiversity [109] and food webs [110].
Moreover, there are definite flaws in the model:-
(1) If West et al. were correct, maximal and standard metabolic rates should both scale as M0.75. The weight of evidence suggests that maximum rate in homoiotherms scales as M0.88 (see earlier discussion [46-49] and following section).
(2) During maximal energy output by an organism, the supply of material is likely to be limiting. For example, in mammals, muscle contraction is responsible for most of the energy turnover at maximum output and it is generally believed that the rate is limited by oxygen supply (if anaerobic capacity is ignored). However, under standard metabolic rate conditions, energy demand is generally more significant, i.e. for the service and cellular maintenance functions mentioned previously. Therefore, it is not clear why the geometry and physics of the supply system should predict the allometric scaling of standard rather than maximal metabolic rate. ("Supply" and "demand" under conditions of maximal aerobic metabolism are complex terms because many physiological steps are involved. The extent to which each step limits the maximum metabolic rate might be quantifiable by a suitable extension of metabolic control analysis [111]; this remains an active research area to which West et al. scarcely refer.)
(3) The mathematical derivations given in [101] are idealisations, but they do not seem to allow for large deviations from b = 0.75. However, there are often wide differences among empirical b values, as discussed earlier; these were addressed in, for example, [18] and [31]. Also, the model does not account, or allow, for the differences in allometric scaling among mammalian tissues and organs [66,73,80].
(4) West et al. accept that some of their proposed hierarchical supply networks might be "virtual" (as in mitochondria) rather than explicit (as in mammalian blood circulation), but it is not clear why such networks must always have the same geometry. For instance, why should the intracellular network discussed by Hochachka [93] show area-preserving branching? There is no evidence that it does. Moreover, the "flow" of reductants through mitochondria presumably takes place in the plane of the inner membrane, which has one dimension fewer than (say) the mammalian circulatory system, so even if mitochondria can be said to have "metabolic rates", the 0.75-power law cannot apply here; yet, allegedly, it does apply.
These difficulties show that the West et al. model, despite its impressive economy, elegance, consistency and range, cannot be accepted unreservedly in its present form. The very generality, or "universality", of this model has made it suspect for some biologists [25]. The implication that it reveals a long-suspected universal biological principle implicit in Kleiber's law has ensured its attraction for others [14].
The model of Darveau and co-workers [112]
This group elaborated a multi-cause rather than a single-cause account of allometric scaling. Their "allometric cascade" model holds that each step in the physiological and biochemical pathways involved in ATP biosynthesis and utilization has its own scaling behaviour and makes its own contribution (defined by a control coefficient between 0 and 1) to the whole-organism metabolic rate. Thus, many linked steps rather than a single overarching principle account for Kleiber's law.
This idea is inherently plausible, and the model is attractive because it draws upon recent advances in metabolic control analysis in biochemistry [111] and physiology [113]. It emphasises that standard metabolic rate is determined by energy demand, not supply; and it predicts an exponent for maximal metabolic rate in mammals between 0.8 and 0.9, rather than 0.75, which agrees with experimental findings [46-49] and the data cited by Weibel [50]. Implicitly – though the authors do not emphasize this – it seems capable of explaining b values that are far from 0.75 (cf [31]). It is hardly surprising, therefore, that many responses to the Darveau et al. model have been positive [e.g. [114]].
However, Darveau et al. made no attempt to explain why the values of b are typically around 0.75, as West et al. and others have done. The model is phenomenological, not physical and mathematical; their equations are not derived from any fundamental principle(s). Moreover, their data cover only some three orders of magnitude of body mass, whereas many studies have involved much wider ranges. This might make their overall b values misleading [103] or, alternatively, more credible [18]. When their equations are applied to a mass range of eight orders of magnitude, different b values are obtained, not necessarily consistent with published data; but on the other hand, the published data might not be correct.
In the first published account of this model [112] the mathematical argument was flawed. The basic equation was given in the form B = aΣciMb(i), where a is a constant coefficient, cI is the control coefficient of the ith step in the cascade and b(i) is the exponent of the ith step. By definition, the sum of all the cI values is unity. Darveau et al. did not derive this equation; they stated it. They also stated that the overall exponent, the b term in the Kleiber equation, is a weighted average of all the individual b(i) values, the weighting being determined by the relevant control coefficients. It has been suggested that this leads to untenable inferences. For example, since the units of B and a are fixed, the units of cI must depend on those of b(i); but by definition, both b(i) and cI must be dimensionless. Also, according to the basic equation, the contribution made by each step to the overall metabolic rate depends on the units in which body mass is measured. If this criticism is valid then it is impossible to evaluate the model as it stands, because any attempt to align its predictions with experimental data would be meaningless. Another reservation about this model is that it does not purport to apply to all taxa, as the West et al. model does; it relates only to metazoa, and in particular to homoiotherms. However, most of the relevant data in the literature concern homoiotherms.
A subsequent publication from this group [115] re-stated the basic equation in the form B = aΣcI(M/m)b(i), where the constant a is described as the "characteristic metabolic rate" of an animal with characteristic body mass m. This eliminates the problem of mass units, because the mass term has been rendered dimensionless; and it is mathematically simple to express control coefficients in dimensionless form. The revised equation might therefore be immune to some of the criticisms levelled at its predecessor. However, some of the earlier reservations remain: the equation remains phenomenological, not physical or geometrical; and the restriction in its range of application is explicit. Nevertheless, these considerations by no means invalidate the model. Indeed, it is supported by data from experiments in exercise physiology [116].
The models of Darveau et al. [112,115], Banavar et al. [94,95] and West et al. [88,102] all have attractive features; but they all have flaws, and they cannot be reconciled with one another. If the positive contributions to biology that these models represent could be further developed, and their defects eliminated, could they be harmonized? If so, the advancement of our understanding would be considerable.
Conclusions
Several explanatory or quasi-explanatory models have been proposed for the allometric scaling of metabolic rate with body mass. Most of them have significant attractions, particularly the most recent ones, but none of them can be unreservedly accepted. The variability of experimental data leaves room for doubt that Kleiber's law is universally or even widely applicable in biology [17,117], yet most workers in the field presume that it is. Even if such doubts are set aside, no model has yet addressed every relevant issue. For example, the biochemical reasons for the allometric scalings of mitochondrial inner membrane areas and unsaturated fatty acid contents, and the direct proportionality of "pulmonary diffusion capacity" to body mass, remain unexplained. Despite the continuing controversy in the field, the consensus remains, and practical use has been made of Kleiber's law, for example in making numerical predictions of anatomical and physiological parameters for veterinary applications [118]. Perhaps the last word should be given to Bokma [119], whose most recent paper explores the power-scaling of metabolic rate to body mass (b) on an intra-specific basis from a total of 113 species. He came to the conclusion that there was no single universal value of b. This evidence alone must make us more sceptical of there being some unifying law involved that demands that b holds close to 0.75. There is clearly no consensus otherwise Nature, Science and the Proceedings of the National Academy of Sciences USA would cease to publish so regularly many of the articles to which we have referred. The subject is not only unresolved, but remains very much within the general interest of biologists.
Kleiber's law remains a fascinating mystery; possibly a delusion, possibly a widespread or even ubiquitous biological phenomenon for which no entirely satisfactory account has yet been offered. Recent developments, though mutually conflicting as they stand, have the potential to lead to new insights and to uncover one or more general biological principles that will have a profound impact on our understanding of the living world.
Acknowledgements
We are indebted to Raul Suarez, Jim Clegg, John Porteous and George Somero for their critical comments, helpful discussions and encouragement.
==== Refs
Kleiber M Body size and metabolism Hilgardia 1932 6 315 353
Brody S Procter RC Ashworth US Basal metabolism, endogenous nitrogen, creatinine and neutral sulfur excretion as functions of body weight Univ Missouri Agric Exp Sta Res Bull 1934 220 1 40
Benedict FG Vital Energetics: a Study in Comparative Basal Metabolism 1938 Washington DC: Carnegie Institute of Washington
Hemmingsen AM The relationship of standard (basal) energy metabolism to total fresh weight of living organisms Rep Steno Mom Hosp Copenhagen 1950 4 1 58
Hemmingsen AM Energy metabolism as related to body size and respiratory surfaces, and its evolution Rep Steno Mom Hosp Copenhagen 1960 9 1 110
Zeuthen E Oxygen uptake as related to body size in organisms Quart Rev Biol 1953 28 1 12 13047555 10.1086/399308
Kleiber M The Fire of Life. An Introduction to Animal Energetics 1961 New York: Wiley
Schmidt-Nielsen K Scaling in biology: the consequences of size J Exp Zool 1975 194 287 308 811757
Schmidt-Nielsen K Scaling: Why is Animal Size So Important? 1984 Cambridge: Cambridge University Press
Peters RH The Ecological Implications of Body Size 1983 Cambridge: Cambridge University Press
McMahon TA Bonner JT On Size and Life 1983 New York: Scientific American Books
Calder WA Size, Function, and Life History 1984 Cambridge, Ma: Harvard University Press
McNab BK Complications inherent in scaling the basal rate of metabolism in mammals Quart Rev Biol 1988 63 25 54 3287424 10.1086/415715
Niklas KJ Plant Allometry: the Scaling of Form and Process 1994 Chicago: University of Chicago Press
Brown JH West GB eds Scaling in Biology 2000 New York: Oxford University Press
Dodds PS Rothman DH Weitz JS Re-examination of the '3/4-law' of metabolism J Theoret Biol 2001 209 9 27 11237567 10.1006/jtbi.2000.2238
Feldman HA On the allometric mass exponent, when it exists J Theoret Biol 1995 172 187 197 7891456 10.1006/jtbi.1995.0015
Atanasov AT Dimitrov BD Changes of the power coefficient in the 'metabolism-mass' relationship in the evolutionary process of animals Biosystems 2002 66 65 71 12204443 10.1016/S0303-2647(02)00034-5
Niklas KJ Enquist BJ Invariant scaling relationships for interspecific plant biomass production rates and body size Proc Natl Acad Sci USA 2001 98 2922 2927 11226342 10.1073/pnas.041590298
Enquist BJ Niklas KJ Invariant scaling relationships across tree-dominated communities Nature 2001 410 655 660 11287945 10.1038/35070500
Bartels H Metabolic rate in mammals equals the 0.75 power of their body weight Exp Biol Med 1982 7 1 11
Andersen HT ed 1969 The Biology of Marine Mammals. New York: Academic Press
Elgar MA Harvey PH Basal metabolic rates in mammals: allometry, phyogeny and ecology Funct Ecol 1987 1 25 36
Economos AC Elastic and/or geometric similarity in mammalian design? J Theoret Biol 1983 103 167 172 6621067
Secor SM Diamond J Determinants of the postfeeding metabolic response of Burmese pythons, Python molurus Physiol Zool 1997 70 202 212 9231393
McKenzie D New clues as to why size equals destiny Science 1999 284 1607 1609 10383334 10.1126/science.284.5420.1607
Dreyer O Allometric scaling and central source systems Phys Rev Lett 2001 87 1 3
Prothero J Scaling of energy metabolism in unicellular organisms: a re-analysis Comp Biochem Physiol A 1986 83 243 248 2869865 10.1016/0300-9629(86)90569-4
Smith RJ Allometric scaling in comparative biology: problems of concept and method Am J Physiol 1984 246 R152 R160 6696141
Kaitaniemi P Testing the allometric scaling laws J Theoret Biol 2004 228 149 153 15094011 10.1016/j.jtbi.2003.12.007
Patterson MR A mass transfer explanation of metabolic scaling relationships in some aquatic invertebrates and algae Science 1992 255 1421 1423 17801232
Coulson RA Metabolic rate and the flow theory: a study in chemical engineering Comp Biochem Physiol 1986 84 217 229 10.1016/0300-9629(86)90607-9
Rubner M Über die Einflus der Körpergrösse auf Stroff und Kraftwechsel Z Biol 1883 19 535 562
Heusner AA Energy metabolism and body size. I. Is the 0.75 mass exponent of Kleiber's equation a statistical artifact? Respir Physiol 1982 48 1 12 7111915 10.1016/0034-5687(82)90046-9
Feldman HA McMahon TA The 3/4 mass exponent for energy metabolism is not a statistical artifact: Respir Physiol 1983 52 149 163 6878906 10.1016/0034-5687(83)90002-6
Heusner AA Size and power in mammals J Exp Biol 1991 160 25 54 1960515
Hayssen V Lacy RC Basal metabolic rates in mammals: taxonomic differences in the allometry of BMR and body mass Comp Biochem Physiol 1985 81A 741 754 10.1016/0300-9629(85)90904-1
White CR Seymour RS Mammalian basal metabolic rate is proportional to body mass2/3 Proc Natl Acad Sci USA 2003 100 4046 4049 12637681 10.1073/pnas.0436428100
Harvey PH Bennett PM Evolutionary biology. Brain size, energetics, ecology and life history patterns Nature 1983 306 314 315 6646214 10.1038/306314a0
Wieser W A distinction must be made between the ontogeny and the phylogeny of metabolism in order to understand the mass exponent of energy metabolism Respir Physiol 1984 55 1 9 6709979 10.1016/0034-5687(84)90112-9
Butler JP Feldman HA Fredberg JJ Dimensional analysis does not determine a mass exponent for metabolic scaling Am J Physiol 1987 253 R195 R199 3605384
McMahon TA Using body size to understand the structural design of animals: quadrupedal locomotion J Appl Physiol 1975 39 619 627 1194153
Brody S Bioenergetics and Growth, with Special Reference to the Efficiency Complex in Domestic Animals 1945 New York: Reinholt
Nevill AM The need to scale for differences in body size and mass: an explanation of Kleiber's 0.75 mass exponent J Appl Physiol 1994 77 2870 2873 7896634
Taylor CR Maloiy GMO Weibel ER Longman VA Kamau JMZ Seeherman HJ Heglund NC Design of the mammalian respiratory system. III Scaling maximum aerobic capacity to body mass: wild and domestic mammals Respir Physiol 1981 44 25 37 7232885 10.1016/0034-5687(81)90075-X
Hinds DS Baudinette RV MacMillen R Halpern EA Maximum metabolism and the aerobic factorial scope of endotherms J Exp Biol 1993 182 41 56 8228784
Bishop CM The maximum oxygen consumption and aerobic scope of birds and mammals: getting to the heart of the matter Proc Roy Soc London B 1999 266 2275 2281 10.1098/rspb.1999.0919
Weibel ER Bacigalupe LD Schmitt B Hoppeler H Allometric scaling of maximal metabolic rate in mammals: muscle aerobic capacity as determinant factor Respir Physiol Neurobiol 2004 140 115 132 15134660 10.1016/j.resp.2004.01.006
Koteja P On the relation between basal and maximum metabolic rate in mammals Comp Biochem Physiol 1987 87 205 208 10.1016/0300-9629(87)90447-6
Weibel ER Symmorphosis: on Form and Function in Shaping Life 2000 Cambridge, Ma: Harvard University Press
Economos AC Gravity, metabolic rate and body size of mammals Physiologist 1979 S71 120950
Smith AH Physiological changes associated with long-term increases in acceleration COSPAR: Life Sci Space Res 1976 14 91 100
Smith AH The roles of body mass and gravity in determining the energy requirements of homoiotherms COSPAR: Life Sci Space Res 1978 16 83 88
Pace N Rahlmann DF Smith AH Scaling of metabolic rate on body mass in small laboratory animals Physiologist 1981 S115 S116
Holmes E The Metabolism of Living Tissues 1938 Cambridge: Cambridge University Press
Altman PL Dittmer DS Metabolism 1968 Bethesda, Maryland: Fed Am Soc Exp Biol Press
Kozlowski J Konarzewski M Gawelczyk AT Cell size as a link between noncoding DNA and metabolic rate scaling Proc Natl Acad Sci USA 2003 100 14080 14085 14615584 10.1073/pnas.2334605100
Lasiewski RC Calder WA A preliminary allometric analysis of respiratory variables in resting birds Respir Physiol 1971 11 152 166 5540203 10.1016/0034-5687(71)90020-X
Tenney SM Bartlett D Comparative quantitative morphology of the mammalian lung: trachea Respir Physiol 1967 3 130 135 6058337 10.1016/0034-5687(67)90002-3
Stahl WR Scaling of respiratory variables in mammals J Appl Physiol 1967 22 453 460 6020227
Gehr P Mwangi DK Ammann A Maloiy GMO Taylor CR Weibel ER Design of the mammalian respiratory system. V. Scaling morphometric pulmonary diffusing capacity to body mass: wild and domestic mammals Respir Physiol 1981 44 61 86 7232887 10.1016/0034-5687(81)90077-3
O'Neil JJ Leith DE Lung diffusing capacity scaled in mammals from 25 g to 500 kg Fed Proc 1980 39 972
Weibel ER Taylor CR Gehr P Hoppeler H Mathieu O Maloiy GMO Design of the mammalian respiratory system. IX. Functional and structural limits for oxygen flow Respir Physiol 1981 44 151 164 7232884 10.1016/0034-5687(81)90081-5
Taylor CR Weibel ER Design of the mammalian respiratory system. I. Problem and strategy Respir Physiol 1981 44 1 10 7232879 10.1016/0034-5687(81)90073-6
Schmidt-Nielsen K Pennycuik P Capillary density in mammals in relation to body size and oxygen consumption Am J Physiol 1961 200 746 750 13748094
Prothero J Heart weight as a function of body weight in mammals Growth 1980 43 139 50 510954
Buttgereit F Brand MD A hierarchy of ATP-consuming processes in mammalian cells Biochem J 1995 312 163 167 7492307
Krebs HA (1950) Body size and tissue respiration Biochim Biophys Acta 1950 4 249 269 15403932 10.1016/0006-3002(50)90032-1
Kleiber M Body size and metabolism of liver slices in vitro Proc Soc Exp Biol Med 1961 48 419 423
Couture P Hulbert JA On the relationship between body mass, tissue metabolic rate and sodium pump activity in mammalian liver and kidney cortex Am J Physiol 1995 268 R641 R650 7900906
Martin AW Fuhrman FA The relationship between summated tissue respiration and metabolic rate in the mouse and dog Physiol Zool 1955 28 18 34
Spaargen DH Metabolic rate and body size: a new view on the 'surface law' for basic metabolic rate Acta Biotheor 1994 42 263 269 7801728 10.1007/BF00707392
Wang Z O'Connor T Heshk S Heymsfield SB The reconstruction of Kleiber's law at the organ-tissue level J Nutr 2001 131 2967 2970 11694627
Porter RK Brand MD Cellular oxygen consumption depends on body mass Am J Physiol 1995 269 R226 R228 7631898
Smith RE Quantitative relations between liver mitochondria metabolism and total weight in mammals Ann New York Acad Sci 1956 62 403 422
Schwertzmann K Hoppeler H Kayar SP Weibel EP Oxidative capacity of muscle and mitochondria: correlation of physiological, biochemical and morphological characteristics Proc Natl Acad Sci USA 1989 86 1583 1587 2922400
Else PL Hulbert AJ Mammals: an allometric study of metabolism at tissue and mitochondrial level Am J Physiol 1985 248 R415 R421 3985183
Porter RK Hulbert AJ Brand MD Allometry of a mitochondrial proton leak: influence of surface area and fatty acid composition Am J Physiol 1996 271 R1550 R1560 8997352
Brand MD Chien L-F Diolez P Experimental discrimination between proton leak and redox slip during mitochondrial electron transport Biochem J 1994 310 477 481
Stahl WR Organ weights in primates and other mammals Science 1965 150 1039 1042 5843618
Drabkin DL The distribution of the chromoproteins, hemoglobin, myoglobin, and cytochrome c, in the tissues of different species, and the relationship of the content of each chromoprotein to body mass J Biol Chem 1950 182 317 333
Kunkel HO Spalding JF Franciscis G Futrell MF Cytochrome oxidase activity and body weight in rats and in species of large animals Am J Physiol 1956 186 203 206 13362505
Jansky L Total cytochrome oxidase activity and its relation to basal and maximal metabolism Nature 1961 189 921 922 13789490
Guan M-X Fischel-Ghodsian N Attardi G Biochemical evidence for nuclear gene involvement in phenotype of non-syndromic deafness associated with mitochondrial 12S rRNA mutation Hum Molec Genet 1996 5 963 971 8817331 10.1093/hmg/5.7.963
Villani G Attardi G In vivo control of respiration by cytochrome c oxidase in wild-type and mitochondrial DNA mutation-carrying human cells Proc Natl Acad Sci USA 1997 94 1166 1171 9037024 10.1073/pnas.94.4.1166
Terroine EF Roche J La respiration des tissus. I. Production calorique des homeotherms et intensité de la respiration in vitro des tissu homologues Arch Intern Physiol 1925 24 356 399
Rumsey WL Schloss C Nuutinen EM Robiolo M Wilson DF Cellular energetics and the oxygen dependence of respiration in cardiac myocytes isolated from adult rat J Biol Chem 1990 265 15392 15399 2394731
West GB Woodruff WH Brown JH Allometric scaling of metabolic rate from molecules and mitochondria to cells and mammals Proc Natl Acad Sci USA 2002 99 2473 2478 11875197 10.1073/pnas.012579799
Wheatley DN Clegg JS What determines the basal metabolic rate of vertebrate cells in vivo? BioSystems 1994 32 83 92 8043754 10.1016/0303-2647(94)90033-7
Wheatley DN On the possible importance of an intracellular circulation Life Sci 1985 36 299 307 4038419 10.1016/0024-3205(85)90114-6
Clegg JS Wheatley DN Intracellular organization: evolutionary origins and possible consequences to metabolic rate control in vertebrates Am Zool 1991 31 504 513
Wheatley DN On the vital role of fluid movement in organisms and cells: a brief historical account from Harvey to Coulson, extending the hypothesis of circulation Medical Hypotheses 1999 52 275 284 10465662 10.1054/mehy.1997.0661
Hochachka PW The metabolic implications of intracellular circulation Proc Natl Acad Sci USA 1999 96 12233 12239 10535904 10.1073/pnas.96.22.12233
Banavar JR Naritan A Rinaldo A Size and form in efficient transportation networks Nature 1999 399 130 132 10335841 10.1038/20144
Banavar JR Dalmuth J Maritan A Rinaldo A Supply-demand balance and metabolic scaling Proc Natl Acad Sci USA 2002 99 10506 10509 12149461 10.1073/pnas.162216899
Dreyer O Puzio R Allometric scaling in animals and plants J Math Biol 2001 43 144 156 11570589 10.1007/s002850170001
Haff PK Rivers, blood and transportation networks Nature 2000 408 159 160 11089962 10.1038/35041633
Rau AR Biological scaling and physics J Biosci 2002 27 475 478 12381870
Blum JJ On the geometry of four-dimensions and the relationship between metabolism and body mass J Theoret Biol 1977 64 599 602 839824
Speakman JR On Blum's four-dimensional geometric explanation for the 0.75 exponent in metabolic allometry J Theoret Biol 1990 144 139 141 2385111
West GB Brown JH Enquist BJ A general model for the origin of allometric scaling laws in biology Science 1997 276 122 126 9082983 10.1126/science.276.5309.122
West GB Brown JH Enquist BJ The fourth dimension of life: fractal geometry and allometric scaling of organisms Science 1999 284 1677 1679 10356399 10.1126/science.284.5420.1677
Agutter PS Wheatley DN Random walks and cell size BioEssays 2001 22 1018 1023 11056478
Santillan M Allometric scaling law in a simple oxygen exchanging network: possible implications on the biological allometric scaling laws J Theor Biol 2003 223 249 257 12814606 10.1016/S0022-5193(03)00097-3
Gillooly JF Brown JH West GB Savage VM Charnov EL Effects of size and temperature on metabolic rate Science 2001 293 2248 2251 11567137 10.1126/science.1061967
Hochachka PW Somero GN Biochemical Adaptation: Mechanism and Process in Physiological Evolution 2001 Oxford: Oxford University Press
Carbone C Gittleman JL A common rule for the scaling of carnivore density Science 2002 295 2273 2276 11910114 10.1126/science.1067994
Enquist BJ Universal scaling in tree and vascular plant allometry: towards a general quantitative theory linking plant form and function from cells to ecosystems Tree Physiol 2002 22 1045 1064 12414366
Richie ME Olff H Spatial scaling laws yield a synthetic theory of biodiversity Nature 1999 400 557 560 10448857 10.1038/23010
Garlaschelli D Caldarelli G Pietronero L Universal scaling relations in food webs Nature 2003 423 165 168 12736684 10.1038/nature01604
Fell D Understanding the Control of Metabolism 1999 London: Portland Press
Darveau C-A Suarez RK Andrews RD Hochachka PW Allometric cascade as a unifying principle of body mass effects on metabolism Nature 2002 417 166 170 12000958 10.1038/417166a
Jones JH Optimization of the mammalian respiratory system: symmorphosis versus single-species adaptation Comp Biochem Physiol 1998 120B 125 138
Burness GP Ecology. Elephants, mice, and red herrings Science 2002 296 1245 1247 12016294 10.1126/science.1072300
Hochachka PW Darveau CA Andrews RD Suarez RK Allometric cascade: a model for resolving body mass effects on metabolism Comp Biochem Physiol 2003 134 675 691
Batterham AM Jackson AS Validity of the allometric cascade model at submaximal and maximal metabolic rates in exercising men Respir Physiol Neurobiol 2003 135 103 106 12706070 10.1016/S1569-9048(03)00027-2
Heusner AA Body size, energy metabolism, and the lungs J Appl Physiol 1983 5 867 873 6853292
Lindstedt L Schaeffer PJ Use of allometry in predicting anatomical and physiological parameters of mammals Lab Anim 2002 36 1 19 11833526 10.1258/0023677021911731
Bokma F Evidence against universal metabolic allometry Func Ecol 2004 18 184 187 10.1111/j.0269-8463.2004.00817.x
| 15546492 | PMC539293 | CC BY | 2021-01-04 16:39:22 | no | Theor Biol Med Model. 2004 Nov 16; 1:13 | utf-8 | Theor Biol Med Model | 2,004 | 10.1186/1742-4682-1-13 | oa_comm |
==== Front
Mol CancerMolecular Cancer1476-4598BioMed Central London 1476-4598-3-341557596410.1186/1476-4598-3-34ResearchExpression of the Tpl2/Cot oncogene in human T-cell neoplasias Christoforidou Anna V [email protected] Helen A [email protected] Andrew N [email protected] George D [email protected] Christos [email protected] Department of Clinical Chemistry-Biochemistry, School of Medicine, University of Crete and University Hospital of Heraklion, 71110 Heraklion, Crete, Greece2 Department of Hematology, School of Medicine, University of Crete and University Hospital of Heraclion, 71110 Heraclion, Crete, Greece2004 3 12 2004 3 34 34 9 9 2004 3 12 2004 Copyright © 2004 Christoforidou et al; licensee BioMed Central Ltd.2004Christoforidou 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
Tpl2/Cot oncogene has been identified in murine T-cell lymphomas as a target of MoMuLV insertion. Animal and tissue culture studies have shown that Tpl2/Cot is involved in interleukin-2 (IL-2) and tumor necrosis factor-α (TNF-α) production by T-cells contributing to T-cell proliferation. In the present report we examined a series of 12 adult patients with various T-cell malignancies, all with predominant leukemic expression in the periphery, for the expression of Tpl2/Cot oncogene in order to determine a possible involvement of Tpl2/Cot in the pathogenesis of these neoplasms.
Results
Our results showed that Tpl2/Cot was overexpressed in all four patients with Large Granular Lymphocyte proliferative disorders (LGL-PDs) but in none of the remaining eight patients with other T-cell neoplasias. Interestingly, three of the LGL-PD patients displayed neutropenia, one in association with sarcoidosis. Serum TNF-α levels were increased in all Tpl2/Cot overexpressing patients while serum IL-2 was undetectable in all subjects studied. Genomic DNA analysis revealed no DNA amplification at the Tpl2/Cot locus in any of the samples analyzed.
Conclusions
We conclude that Tpl2/Cot, a gene extensively studied in animal and tissue culture T-cell models may be also involved in the development of human LGL-PD and may have a role in the pathogenesis of immune manifestations associated with these diseases. This is the first report implicating Tpl2/Cot in human T-cell neoplasias and provides a novel molecular event in the development of LGL-PDs.
==== Body
Background
Cells may transform to a malignant phenotype following accumulation of distinct genetic events that result in altered protein expression pattern, thus facilitating uncontrolled proliferation. Such genetic events target specific oncogenes that act in concert to provide the malignant phenotype.
Tpl2/Cot oncogene was initially cloned as a MoMuLV proviral integration locus in murine T-cell lymphoma cells, resulting in its carboxy-terminal truncation[1,2]. Expression of the truncated form of Tpl2 as a transgene in T-cells under the control of the lck promoter in mice results in rapid development of T-cell lymphomas [3]. Expression of Tpl2 is associated with T-cell activation. Overexpression of the wild type Tpl2 in the Jurkat T-cell leukemia cell line results in NFkB and NFAT activation and subsequent IL-2 and TNF-α expression [4-7]. In the CTLL2 IL-2 dependent cell line Tpl2 promotes cell proliferation by activating E2F-dependent transcription [8]. Tpl2/Cot is, therefore, tightly associated with T-cell neoplasms and T-cell activation and proliferation.
Studies in human tumor specimens have shown that Tpl2/Cot is overexpressed in early stage breast cancer [9], in EBV-related Hodgkin lymphomas and nasopharyngeal carcinomas [10] and occasionally in gastric and colon adenocarcinomas [11]. To our knowledge, no available data exist on human hematologic neoplasias, other than Hodgkin lymphoma.
Given the compelling evidence of the importance of Tpl2/Cot in experimental and tissue culture models of T-cell neoplasias, we designed a study to investigate possible involvement of Tpl2/Cot in the pathogenesis of human T-cell neoplasias. Specifically, we studied 12 adults with various T-cell neoplasias to obtain a broad spectrum of T-cell malignancies, all with predominant leukemic expression, and examined whether Tpl2/Cot expression is deregulated in the transformed cells. The expression levels of Tpl2/Cot were quantitated by SybrGreen real-time RT-PCR using three different quantitation approaches (standard curve[12], absolute fluorescence increase [13] and the M.W.Pfaffl method [14]) as well as the conventional semi-quantitative RT-PCR.
Results
Evaluation of Tpl2/Cot mRNA expression in T-cell neoplasias
To determine the levels of expression of Tpl2/Cot mRNA we first established and validated a real time PCR approach. Melting curves showed that there were no by-products in both Tpl2/Cot and GAPDH reactions (Figure 1B). CVs of mean triplicate Ct (threshold cycle) ranged from 0.1% to 0.92% which account for a low intra-assay variability. A series of five 10 to 20-fold dilutions of a standard cDNA from different cDNA preparations were also run multiple times to determine primer efficiencies. Linear regression analysis of the standard curves [mean Ct plotted against the log(RNA input)] showed high linearity, with regression coefficients greater than 0.997 (Figure 1A). We used the standard curve slope in the equation
Figure 1 Real-Time PCR validation A. Serial dilutions of a standard cDNA duplicates for the construction of standard curves for GAPDH and Tpl2/Cot. The curve slopes shown on the upper right corner of each plot are -3.61 and -3.59 respectively. Black arrows correspond to dilution 1:10; white arrows to 1:20; left-pointed arrow is NTC (no-template control). X-axes represent the Log of the dilution factor, Y-axes the mean Ct of duplicates. B. Dissociation (melting) curve of the PCR products, showing a peak at 81.7°C for both Tpl2/Cot(black arrow) and GAPDH(white arrow), while NTCs have either no peak or a peak at a much lower temperature (thin arrow).
(1) E = 10-1/slope
to calculate the mean efficiency of Tpl2/Cot and GAPDH primers. The slopes were almost equal (from -3.59 to -3.62) for both primers which showed that we could use the Pfaffl method [14] without the need of a standard curve in every set of reactions. Specifically, to determine the ratio (R) of the normalized Tpl-2/Cot expression of sample vs control we used the equation
where Etarget and Eref are the Efficiencies of the target (Tpl2) and reference genes(GAPDH) respectively which both were equal to a mean of 1.89 and ÄCt is the difference between the mean Ct of control cDNA(CTR) and patient cDNA (Sample). To confirm our results we also tried the Absolute Fluorescence Increase method using the LinRegPCR software v.7.5 which measures the actual efficiency of each amplification curve by fitting its linear part in a simulation plot of the Log (fluorescence) versus Cycle and calculates the efficiency from the slope of a linear regression model of the simulation curve [13]. As control we used triplicates of cDNA from 3 healthy individuals, for which we calculated the mean Ct. The control samples were representative among 22 control specimens with similar values. Results were similar to those obtained by the Standard Curve method [12] (data not shown).
A total of 12 adult samples with T- and NK-cell neoplasias were analyzed according to the described method. They all had leukemic expression in the periphery. Morphology was assessed by light microscopy on peripheral blood smears, including measurement of absolute LGL number. Four out of twelve (33%) patients tested markedly overexpressed Tpl2/Cot (p = 0.034), as determined by either conventional RT-PCR or real-time qPCR (Tables 2 and 3, Figure 2). Interestingly, all the Tpl2/Cot overexpressing patients had LGL-PD, three with the phenotype of CD3+ T-LGL leukemia and one with the CD3- pattern of chronic NK-lymphocytosis. Three of these patients displayed neutropenia not attributable to BM infiltration, one in association with sarcoidosis (Table 1). The fourth patient with monoclonal T-LGL lymphocytosis had a co-current cutaneous T-cell lymphoma which, during follow-up, demanded systemic chemotherapy.
Table 1 Patient characteristics
Patient no. Disease Sex Age %T+NKcells/PBMC Disease state Co-existent conditions
1 SS F 88 87 PP
2 CTCL with monoclonal T- LGL lymphocytosis M 63 86 RD
3 T-LGL leykemia F 75 93 RD Neutropenia
4 MF M 91 71 RD
5 MF M 85 71 RD
6 TLL F 35 77 RD
7 T-PLL M 85 67 RD Myositis
8 Chronic NK -lymphocytosis F 67 64 Stable for two years Neutropenia
9 T-ALL M 16 95 LB RD
10 T-LGL leukaemia with reactive NK lymphocytosis F 60 71 Stable for 10 years Sarcoidosis (past), neutropenia
11 PTCL secondary to MF M 52 85 PR
12 Pre-T-ALL M 37 95 LB RD
Abbreviations: SS, Sezary syndrome; CTCL, Cutaneous T cell Lymphoma; T-LGL, T-large granular lymphocyte; MF, Mucosis Fungoides; NK, natural killer cell; TLL, T-lymphoblastic lymphoma; T-PLL, T-Prolymphocytic leukemia; T-ALL, T-acute lymphoblastic leukemia; LB, Lymphoblast ;PP, primary progressive; RD, recently diagnosed; PR, partial response.
Table 2 Tpl-2/Cot expression in PBMC
Patient no. Tpl2-Norma Fold increase (R)b CV(%)-Rc
Ctr (n = 3) 0.04 ± 0.019 1.00 ± 0.119 45.3
1 0.02 ± 0.000 0.59 ± 0.009 1.6
2 0.13 ± 0.005 3.88 ± 0.149 3.8
3 0.13 ± 0.005 4.04 ± 0.143 3.5
4 0.02 ± 0.003 0.66 ± 0.092 13.9
5 0.04 ± 0.005 1.22 ± 0.092 11.8
6 0.06 ± 0.003 1.70 ± 0.085 5.0
7 0.06 ± 0.002 1.81 ± 0.076 4.2
8 0.17 ± 0.008 5.21 ± 0.251 4.8
9 0.01 ± 0.000 0.32 ± 0.007 2.2
10 0.16 ± 0.006 5.04 ± 0.194 3.8
11 0.04 ± 0.001 1.09 ± 0.037 3.4
12 0.04 ± 0.004 1.22 ± 0.108 8.9
a. Tpl2/Cot normalized to GAPDH by equation (3) in the text ± SD.
b. Tpl2/Cot-Norm (sample) ratio to Tpl2/Cot-Norm (control) by equation (2) in the text ± SD. Samples in which Tpl2/Cot is overexpressed are shown in bold.
c. % Coefficient variance of (R).
Table 3 Summary of the results
Patient no. Immunophenotype Serum TNF-α (pg/ml) Tpl2/Cot (fold increase)
1 CD3+, CD4+, CD7- 1.7 0.6
2 CD3+, CD8+, CD56+, CD57+, TCRαβ+ 1.9 3.9
3 CD3+, CD8+, CD16, 56-, CD57+, TCRαβ+ 1.8 4
4 CD3+, CD4+, CD5+, CD7+ 1.5 0.7
5 CD3+, CD4+, CD7- 1.9 1.2
6 TdT+, CD5+, CD2+ 2.5 1.7
7 CD3+CD4+CD25-TCRγ+ 3.8 1.8
8 CD2+, CD3-, CD16+, CD56-, CD57-, TCR- 6.2 5.2
9 TdT+, CD2+, CD3+, CD7+, CD4-, CD8- 2.6 0.3
10 2 populations: a) T-LGL CD3+, CD8+, CD56+, CD57+, TCR γ+ b) NK CD2+, CD3-, CD16, 56+, 57- 1.9 5
11 CD3+, CD4+, CD7-, CD25-, TCRαβ+ 1.2 1.1
12 TdT+, cCD3+, CD7+, CD4-CD8- 1.5 1.2
Figure 2 Tpl2/Cot mRNA expression in T-cell neoplasms. Representative semi-quantitative RT PCR for Tpl-2 mRNA expression in patients and controls. A. Tpl2/Cot PCR product of 228 bp B. β-Actin PCR product of 214 bp. Samples no 2, 3, 8 and 10 are overexpressed compared to the control and correspond to the LGL-PD patients shown in Table 1.
Evaluation of serum TNF-α and IL-2 levels
Overexpression of Tpl2/Cot in Jurkat T-cells induces TNF-α expression [7]. Tpl2 also regulates TNF-α expression in macrophages by activating ERK and thus controlling the posttranscriptional modification of the TNF-α mRNA, which is necessary for its export from the nucleus [15]. We, therefore, evaluated serum TNF-α levels in all patients studied. In order to have an internal negative control in the study we analyzed 8 samples from normal blood donors. The mean TNF-α concentration in patient sera was 2.37 ± 1.4 pg/ml with a range between 1.2 pg/ml and 6.2 pg/ml, while in the control group it was 0.6 ± 0.2 pg/m. The mean patient TNF-α value was statistically significant higher than the respective of the healthy controls (p = 0.002), as was TNF-α in the LGL-PD group alone compared to the control group (p = 0.006). It is of interest that the patient with chronic NK-lymphocytosis and neutropenia displayed the highest TNF-α value (6.2 pg/ml) that was associated with the highest Tpl2/Cot expression (5.2 fold compared to control) (Table 3) suggesting a relationship between Tpl2/Cot overexpression and TNF-α overproduction in this patient. There was no difference in TNF-α levels between the group of LGL-PD patients and the group of patients with the remaining T-cell other neoplasias (p = 0,392).
Tpl2/Cot overexpression in Jurkat and EL-4 T-cells induces IL-2 secretion in culture. We, therefore, examined the expression levels of IL-2 in the sera from the 12 patients studied and 8 control samples. No circulating IL-2 was detected in the sera of the patients or the control donors (data not shown).
Overexpression of Tpl2/Cot in LGL-PD is not associated with gene amplification
Overexpression of Tpl2/Cot in human breast cancer has been associated with amplification of the tpl2 genomic locus [9]. We, thus, evaluated whether the overexpression of Tpl2/Cot in T-cell malignancies is associated with amplification of the genomic tpl2 locus. For this purpose genomic DNA was isolated from PBMCs of the same patients and part of the genomic tpl2 locus was amplified using multiplex PCR. As reference gene we used the IFN-γ gene. The results showed a similar tpl2/IFN-γ ratio in all cases (Figure 3) indicating that overexpression of Tpl2/Cot in LGL-PDs is not due to gene amplification.
Figure 3 The tpl2/cot genomic locus is not amplified in the T-cell neoplasms analyzed. Multiplex PCR for the quantification of the Tpl-2/Cot gene load, relative to IFN-γ gene in patient and control DNA: tpl2/cot PCR product detected at 139 bp; IFN-γ PCR product at 250 bp. No significant difference between samples is evident.
Discussion
Experimental data based on in vitro and animal model have shown that Tpl2/Cot is an important regulator in the transduction of signals leading to T-cell activation [16]. Overexpression of this kinase results in increased proliferation of T-cells by activating E2F-dependent transcriptional activity [8]. A truncated form present in rodents exhibits increased catalytic activity, and when overexprerssed as a T-cell-specific transgene in mice it induces tumors within 3–9 months [3]. Tpl2/Cot activates the transcription factors NFAT and NFkB in T-cells, which drive the transcription of several cytokine genes such as IL-2 and TNF-α [4,7]. In macrophages, Tpl2/Cot is essential for the activation of ERK by LPS via TLR4 and the export of TNF-α mRNA from the nucleus [15]. The preceding evidence supports a possible involvement of Tpl2/Cot in human T-cell neoplasias.
Investigation of Tpl2/Cot expression in human tumor specimen has shown that it is occasionally overexpressed in colon and gastric adenocarcinomas [11] and human breast cancer tissues [9]. Tpl2/Cot overexpression has also been detected in a hepatocellular carcinoma cell line [17] and in patient tumor tissue from EBV-related Hodgkin lymphomas and nasopharyngeal carcinomas[10]. In the present study we analyzed peripheral blood from patients with T- and NK-cell lymphoproliferative diseases at the time where the patients were not receiving any treatment and had profound leukemic expression in the periphery. Out of the 12 cases analyzed, Tpl2/Cot was found overexpressed in three T-LGL-leukemias and in one chronic NK-lymphocytosis accounting for all four LGL-PDs studied. These findings suggest that Tpl2/Cot deregulation may be a defining molecular event for the development of this type of neoplasias. Given the role of Tpl2/Cot in T-cell proliferation via activation of E2F transcription [8], overexpression of this kinase may contribute to neoplastic cell proliferation.
Large granular lymphocyte lymphoproliferative diseases (LGL-PD) are relatively rare and not well defined disorders, frequently associated with autoimmune diseases or immune mediated manifestations such as rheumatoid arthritis (pseudo-Felty syndrome [18]), RF positivity, neutropenia and pure red cell aplasia [19,20]. They present clinical, morphological and immunological distinct features, resulting from chronic proliferation of CD3+ or CD3- granular lymphocytes. In the CD3+ cases the proliferating cells express CD8 and NK-associated surface antigens such as CD16, the LGL-specific CD57 antigen and CD45RA, and display also clonal rearrangement of the TCR alpha-beta or, less often, gamma-delta chains, thus representing cytotoxic effector T-cells[21]. The T-LGL leukemias are by definition indolent, but there are rare reports indicating evolution to high grade lymphoma [22]. Limited data on recurrent chromosomal aberrations exist [23,24]. In the rare CD3- cases the cells are TCR-, CD2+, CD16+ and CD56+ representing, therefore, true NK-cell proliferations corresponding to the aggressive NK-cell leukemias or to the – usually benign – chronic NK-lymphocytosis [19,20]. Lack of Tpl2/Cot gene amplification in our LGL-PD patients indicate that overexpression is either due to changes in the regulation of Tpl2/Cot gene activation or mRNA stability. Such changes can be either primary (i.e. mutations that affect the promoter or the mRNA stability) or secondary (i.e. activation of transcription factors that affect the Tpl2/Cot promoter or signaling molecules that affect the stability of its mRNA).
There is accumulating evidence suggesting that patients with LGL-PDs display frequently immune manifestations and increased TNFα production by the neoplastic T-cells has been reported to play a role in their pathogenesis[20,25]. Interestingly, three of our LGL-PD patients displayed neutropenia not attributable to BM infiltration while one of these patients displayed also sarcoidosis. Evaluation of circulating TNF-α level in the patients showed that the highest TNF-α value was found in the LGL-PD patients that displayed also the highest Tpl2/Cot expression among the subjects studied. These findings are in agreement with previous reports demonstrating that Tpl2/Cot is involved in TNF-α expression and secretion [7,15] while provide evidence for a causal relationship between the Tpl2/Cot overexpression, the TNF-α overproduction and the pathogenesis of neutropenia in LGL-PD patients [25]. TNF-α was increased in the sera of patients where Tpl2/Cot expression was not elevated, indicating that in these patients TNF-α may be upregulated via alternative pathways not associated with overexpression of Tpl2/Cot.
Conclusions
In conclusion, data from the present study suggest that Tpl2/Cot overexpression may have a role in the development of certain types of human T-cell neoplasms thus confirming experimental data on animal and tissue culture models for the role of Tpl2/Cot in T-cell malignancies.
Methods
Patients
Peripheral blood samples from 12 adults aged 16–88 years (median age 64 years) with various T- and NK-cell neoplasias with peripheral blood leukemic expression were collected during a two-year time period. Patients with signs of infection or recently subjected to cytotoxic therapy were excluded from the study. Diagnosis was based on morphological, immunophenotypic and genomic studies and histological findings of bone marrow and/or lymph node biopsies and disease was classified according to the WHO classification [26]. There were three patients with T-LGL leukemia, one patient with chronic NK lymphocytosis, one patient with Sezary syndrome (SS), two patients with Mucosis Fungoides (MF), one patient with T-Prolymphocytic Leukemia (T-PLL), two patients with T-acute Lymphoblastic Leukemia (T-ALL), one patient with T-lymphoblastic Lymphoma (TLL) and one patient with Peripheral T-Cell Lymphoma (PTCL) secondary to MF. Detailed patient characteristics are presented in Table 1. Complete blood counts and flow cytometric analysis of peripheral blood lymphocytes were performed at the day of blood collection for the molecular study. In addition, patient sera were obtained by centrifugation of 4 ml of non-anticoagulated blood at 3000 rpm for 10 min and were stored at -80°C for IL-2 and TNF-α measurement. Peripheral blood specimens from 22 healthy volunteers age- and sex-matched with the patients were collected and used as controls. This research project was subjected to and approved by the Ethics Committee of the University Hospital of Heraclion.
Peripheral Blood Mononuclear Cell isolation and RNA extraction
Peripheral blood mononuclear cells (PBMC) were isolated from 9 ml of fresh EDTA-K3 anti-coagulated peripheral blood samples by Lymphoprep density centrifugation (Nycomed Pharma AS, Norway). PBMCs were immediately lyzed in suitable volume of Trizol LS reagent (Invitrogen, UK) and mRNA was isolated according to the manufacturers' protocol. An aliquot of 1 μg of total RNA was treated with 1 IU DNase I, Amplification Grade (Invitrogen, UK) to eliminate any traces of genomic DNA.
Semi-quantitative RT-PCR
First strand cDNA was synthesized by reverse transcription of 1 μg total RNA using the Thermoscript™ RT kit (Invitrogen, UK). 0.8 μl of cDNA were amplified in a 20 μL PCR reaction containing 250 nM of each primer, 200 nM dNTPs, 0.5 IU Taq polymerase and 2 μl of 10X reaction buffer (Platinum Taq DNA Polymerase kit, Invitrogen, UK). Reactions were first optimized for annealing temperature, Mg and primer concentration (data not shown). Primers for Tpl2/Cot detection were derived from exons 3 and 4 of the human Tpl2/Cot gene (Genbank accession no: AL547407), spanning an 8.5 kb intron to prevent co-current genomic DNA amplification and their sequences were: forward 5'-CAG TAA TCA AAA CGA TGA GCG TTC TAA-3', reverse 5'-GAA CAT CGG AAT CTA TTT GGT AAC GTC-3' producing a 228 bp-length amplicon. For normalization of mRNA input differences human beta actin mRNA (Genbank accession no: BC013835) was detected using the following primers: forward 5'-CCG GCC AGC CAG GTC CAG A-3', reverse 5'-CAA GGC CAA CCG CGA GAA GAT G-3', amplifying a 214 bp cDNA fragment. In each reaction two negative controls were included by either omitting reverse transcriptase at the RT step or cDNA template respectively.
PCR reactions were performed on a thermal cycler (PTC-200 MJ Reasearch with heated lid) and repeated 3 times with different cDNAs from the same mRNA. Expression of Tpl2/Cot was determined by semi-quantitative, relative RT-PCR. An amplification curve for each gene was acquired by performing the reaction with increasing number of PCR annealing cycles to identify the exponential phase of the reaction. The thermocycling parameters were as follows: initial denaturation at 94°C for 5 min followed by 34 for Tpl2 or 23 for actin cycles at 94°C for 30 sec, 54°C for 30 sec, 72°C for 30 sec and a final extension at 72°C for 7 min.
The RT-PCR products were analysed by electrophoresis in a 2.5% agarose gel, stained with 0.2 μg/ml ethidium bromide and visualized in a UVP transiluminator (Gel Doc 1000 Bio Rad). The band intensity was analysed by a densitometric image analysis system (TINA scan v2.07) and the results were expressed as a ratio between Tpl2/Cot and β-actin band intensity.
Real-time PCR
Primers for real-time PCR were designed with the Primer Express Software v.2 (Applied Biosystems) and selected so that they amplify a region of no more than 150 bp, they have similar GC content, same Tm, no more than 3 G or C's at the 3' end and no secondary structure formation. To exclude primers with more than 3 consecutive complementary bases between them we used Qiagen Oligo Analysis & Plotting Tool (Qiagen, Germany). Primers forTpl2/Cot annealed to exons 6–7 and their sequences were: forward 5'-TCC TAA GGA CCT CCG AGG AAC-3', reverse 5'-CCC AGG CTG TAG ATG TCT GCT-3', amplifying a 93 bp region. GAPDH was used as a reference gene to compensate for mRNA input differences. Primers for GAPDH derived from exons 2–3 (Genbank accession no: BC023632) and their sequences were: forward 5'-GGA AGG TGA AGG TCG GAG TCA-3', reverse 5'-GTC ATT GAT GGC AAC AAT ATC CAC T-3', amplifying a 101 bp region.
Primer concentration, cDNA dilution, Mg concentration and annealing temperature were optimized so that a maximum fluorescent signal with no inhibition from RT components and a similar reaction yield from both primer sets could be obtained. For the real-time PCR study cDNA from first strand synthesis treated with RNase H for 20' at 37°C was diluted 1:20 with DNAse-free water and 5 μl were used in a 20-μl reaction mixture consisting of 10.4 μl 2x SybrGreen PCR Master mix, 6 mM final concentration of MgCl2 and 500 nM of the Tpl2/Cot primers or 100 nM of GAPDH primers. Reactions were carried out using an ABI Prism 7000 sequence detector (Applied Biosystems, Foster City, CA, USA) according to manufacturer's instructions. The thermal profile used consisted of 2 min at 50°C,10 min at 95°C and 40 repeats of denaturation at 95°C for 15 sec and annealing-extension-fluorescence data acquisition at 60°C for 1 min. A post-PCR Melting Curve Analysis was performed by a 20-min slow ramp from 60° to 95°C to confirm that there were no by-products. Samples were run in a 3% agarose gel after the end of reaction to confirm specificity. All samples were run in triplicates and two negative controls with either no reverse transcriptase or no cDNA template were included. Reaction was repeated twice in different days to estimate inter-assay variation. Results were analyzed using the ABI Prism 7000 SDS software (v.1.1, Applied Biosystems) and Excel for further quantitative study.
DNA extraction and multiplex PCR
To detect possible gene amplification in cases with Tpl2/Cot overexpression we used a multiplex DNA PCR protocol with interferon-γ (IFN-γ) as a reference gene amplified at the same reaction tube. High molecular weight DNA was isolated from PBMC by proteinase K digestion and phenol chloroform extraction with ethanol precipitation and diluted in TE buffer. The primers for Tpl2/Cot were: forward 5'-GCG ACG GAT TGA GGT TTG-3', reverse 5'-GCG TTT CAG GCG TAT GGA-3' amplifying a 139 bp region of intron 1(Genbank accession no.AY309013) and the primers for IFN-γ were: forward 5'-ATG CAG GTC ATT CAG ATG TAG C-3', reverse 5'-TTG GAT GCT CTG GTC ATC TTT A-3' amplifying a 250 bp fragment containing intron and exon sequences between exons 2–3 (Genbank accession no: J00219). 50 ng of DNA were used in a 20 μl reaction containing 10 μl Platinum qPCR Supermix-UDG (Invitrogen, UK), 250 nM Tpl2/Cot primers and 100 nM IFNγ primers. The number of amplification cycles was adjusted so that the reaction terminated at the middle of the exponential phase of both products (data not shown). The thermal profile consisted of a denaturation step at 95° for 10 min, followed by 29 repeats at 95°C for 15 sec, 58°C for 30 sec and 72°C for 30 sec and a final extension at 72°C for 7 min. PCR products were analysed in a 3% agarose gel, visualized and scanned as described earlier and the Cot/IFNγ ratio was determined.
Peripheral blood lymphocyte immunophenotype and assessment of T-cell clonality
Two-color flow cytometry was used for the analysis of peripheral blood lymphocyte subpopulations. In brief, 100 μL aliquots of EDTA-anticoagulated peripheral blood samples were surface stained with each of the following PE- or FITC-conjugated mouse antihuman monoclonal antibodies: anti-CD2, anti-CD3, anti-CD4, anti-CD8, anti-CD5, anti-CD7, anti-CD16, anti-CD56, anti-CD57, anti-CD19, anti-CD25, anti-CD11b, anti-CD79a, anti-FMC7 and anti-HLA-DR (Beckman Coulter, France). Cells were also stained for intracellular Terminal deoxy-transferase (TdT) (Beckman Coulter) and CD3 using the IntraPrep intracellular staining kit (Beckman Coulter). PE- or FITC-conjugated mouse IgG of appropriate isotype served as negative controls. Following 30 min incubation at room temperature and two washes with phosphate buffer saline (PBS)-1% fetal bovine serum (FBS)-0.05% azide, contaminating red cells were lysed with 0.12% formic acid and samples were fixed in 0.2% parafolmadeyde using the Q-prep reagent system (Coulter, Louton, England). Analysis on 10,000 events was performed in an Epics Elite model flow cytometer (Coulter) in the lymphocyte gate.
Clonality assessment of peripheral blood T-cells was performed by analysing quantitatively different variable regions of the T-cell receptor (TCR) β chain (Vβ repertoire) of CD3+ cells by means of flow cytometry using the IOTest Beta Mark kit (Beckman Coulter), according to the manufacturer's instructions [27]. T-cell clonality assessment was also performed by PCR analysis using primers for the TCR V(D)J junction in PBMC derived DNA, according to standard methods in a reference laboratory.
ELISA
TNF-α concentration was determined using the High Sensitivity human TNF-α ELISA kit (R&D, USA) or the IL-2 ELISA kit (R&D, USA) according to the manufacturer's instruction. According to the manufacturer, the sensitivity of these assays are 0.12 pg/ml and 7 pg/ml respectively.
Statistical analysis
The comparison of Tpl2/Cot mRNA expression between patient and control samples was performed by means of the nonparametric Mann-Whitney U test using as variables the mean normalized Tpl-2 expression of each sample and control, described by the equation
(3) Tpl-2 norm = (E)-(mean CtTpl2-meanCtGAPDH)
where E is the mean common efficiency of the reference and target gene that we calculated (E = 1.89, see section Results)[14]. The Mann-Whitney U test was also used for the comparison of the TNF a and IL-2 levels between the Tpl-2 overexpressing patients, the controls and the non-overexpressing patients.
Authors' contributions
AVC collected specimens and patient data and performed the major body of the experimental work as well as the data analysis and the preparation of the manuscript. HAP was involved in flow cytometric analysis, patient selection and manuscript preparation. ANM and GDE were involved in the study design, the interpretation and evaluation of the clinical data and manuscript preparation. CT conceived the project, performed the ELISAs, was responsible for the coordination of the experimental procedures and was involved in the preparation of the manuscript.
Acknowledgements
We thank Christiana Choulaki and Ioannis Diamantis for their help and critical comments on Real-time PCR methodology and Ariadne Androulidaki and Erini Dermitzaki for their help in semi-quantitative PCR analysis. AVC was partly supported by the Hellenic Hematology Association. This work was partly supported by AICR grand # 03–061 and AICR grand # 03–016.
==== Refs
Makris A Patriotis C Bear SE Tsichlis PN Genomic organization and expression of Tpl-2 in normal cells and Moloney murine leukemia virus-induced rat T-cell lymphomas: activation by provirus insertion J Virol 1993 67 4283 4289 8510223
Patriotis C Makris A Bear SE Tsichlis PN Tumor progression locus 2 (Tpl-2) encodes a protein kinase involved in the progression of rodent T-cell lymphomas and in T-cell activation Proc Natl Acad Sci U S A 1993 90 2251 2255 7681591
Ceci JD Patriotis CP Tsatsanis C Makris AM Kovatch R Swing DA Jenkins NA Tsichlis PN Copeland NG Tpl-2 is an oncogenic kinase that is activated by carboxy-terminal truncation Genes Dev 1997 11 688 700 9087424
Tsatsanis C Patriotis C Tsichlis PN Tpl-2 induces IL-2 expression in T-cell lines by triggering multiple signaling pathways that activate NFAT and NF-kappaB Oncogene 1998 17 2609 2618 9840924 10.1038/sj.onc.1202460
Tsatsanis C Patriotis C Bear SE Tsichlis PN The Tpl-2 protooncoprotein activates the nuclear factor of activated T cells and induces interleukin 2 expression in T cell lines Proc Natl Acad Sci U S A 1998 95 3827 3832 9520452 10.1073/pnas.95.7.3827
Ballester A Tobena R Lisbona C Calvo V Alemany S Cot kinase regulation of IL-2 production in Jurkat T cells J Immunol 1997 159 1613 1618 9257820
Ballester A Velasco A Tobena R Alemany S Cot kinase activates tumor necrosis factor-alpha gene expression in a cyclosporin A-resistant manner J Biol Chem 1998 273 14099 14106 9603908 10.1074/jbc.273.23.14099
Velasco-Sampayo A Alemany S p27kip protein levels and E2F activity are targets of Cot kinase during G1 phase progression in T cells J Immunol 2001 166 6084 6090 11342626
Sourvinos G Tsatsanis C Spandidos DA Overexpression of the Tpl-2/Cot oncogene in human breast cancer Oncogene 1999 18 4968 4973 10490831 10.1038/sj.onc.1202891
Eliopoulos AG Davies C Blake SS Murray P Najafipour S Tsichlis PN Young LS The oncogenic protein kinase Tpl-2/Cot contributes to Epstein-Barr virus-encoded latent infection membrane protein 1-induced NF-kappaB signaling downstream of TRAF2 J Virol 2002 76 4567 4579 11932422 10.1128/JVI.76.9.4567-4579.2002
Ohara R Hirota S Onoue H Nomura S Kitamura Y Toyoshima K Identification of the cells expressing cot proto-oncogene mRNA J Cell Sci 1995 108 ( Pt 1) 97 103 7738120
Biosystems PEA ABI Prism 7700 Sequence Detection System. User bulletin 2. 2001
Ramakers C Ruijter JM Deprez RH Moorman AF Assumption-free analysis of quantitative real-time polymerase chain reaction (PCR) data Neurosci Lett 2003 339 62 66 12618301 10.1016/S0304-3940(02)01423-4
Pfaffl MW A new mathematical model for relative quantification in real-time RT-PCR Nucleic Acids Res 2001 29 e45 11328886 10.1093/nar/29.9.e45
Dumitru CD Ceci JD Tsatsanis C Kontoyiannis D Stamatakis K Lin JH Patriotis C Jenkins NA Copeland NG Kollias G Tsichlis PN TNF-alpha induction by LPS is regulated posttranscriptionally via a Tpl2/ERK-dependent pathway Cell 2000 103 1071 1083 11163183 10.1016/S0092-8674(00)00210-5
Tsatsanis C Spandidos DA The role of oncogenic kinases in human cancer (Review) Int J Mol Med 2000 5 583 590 10812005
Aoki M Hamada F Sugimoto T Sumida S Akiyama T Toyoshima K The human cot proto-oncogene encodes two protein serine/threonine kinases with different transforming activities by alternative initiation of translation J Biol Chem 1993 268 22723 22732 8226782
Toussirot E Lafforgue P Harle JR Kaplanski S Mannoni P Acquaviva PC Pseudo Felty's syndrome. A polyclonal disease with a favorable prognosis. Report of two cases with Southern blot analysis of TCR Clin Exp Rheumatol 1993 11 591 595 8299249
Kingreen D Siegert W Chronic lymphatic leukemias of T and NK cell type Leukemia 1997 11 Suppl 2 46 49
Lamy T Loughran TPJ Current concepts: large granular lymphocyte leukemia Blood Rev 1999 13 230 240 10741898 10.1054/blre.1999.0118
Bigouret V Hoffmann T Arlettaz L Villard J Colonna M Ticheli A Gratwohl A Samii K Chapuis B Rufer N Roosnek E Monoclonal T-cell expansions in asymptomatic individuals and in patients with large granular leukemia consist of cytotoxic effector T cells expressing the activating CD94:NKG2C/E and NKD2D killer cell receptors Blood 2003 101 3198 3204 12480700 10.1182/blood-2002-08-2408
Matutes E Wotherspoon AC Parker NE Osuji N Isaacson PG Catovsky D Transformation of T-cell large granular lymphocyte leukaemia into a high-grade large T-cell lymphoma Br J Haematol 2001 115 801 806 11843812 10.1046/j.1365-2141.2001.03220.x
Man C Au WY Pang A Kwong YL Deletion 6q as a recurrent chromosomal aberration in T-cell large granular lymphocyte leukemia Cancer Genet Cytogenet 2002 139 71 74 12547164 10.1016/S0165-4608(02)00605-2
Wong KF Chan JC Liu HS Man C Kwong YL Chromosomal abnormalities in T-cell large granular lymphocyte leukaemia: report of two cases and review of the literature Br J Haematol 2002 116 598 600 11849217 10.1046/j.0007-1048.2001.03315.x
Shvidel L Duksin C Tzimanis A Shtalrid M Klepfish A Sigler E Haran M Eilat E Berrebi A Cytokine release by activated T-cells in large granular lymphocytic leukemia associated with autoimmune disorders Hematol J 2002 3 32 37 11960393 10.1038/sj.thj.6200149
Harris NL Jaffe ES Diebold J Flandrin G Vardiman J Lister TA Bloomfield CD The World Health Organization classification of neoplasms of the hematopoietic and lymphoid tissues: report of the Clinical Advisory Committee meeting-AirlieHouse,Virginia,November,1997 The Hematology Journal 2000 (2000) 53 66 11920170 10.1038/sj.thj.6200013
Lima M Almeida J Santos AH dos Anjos TM Alguero MC Queiros ML Balanzategui A Justica B Gonzalez M San Miguel JF Orfao A Immunophenotypic analysis of the TCR-Vbeta repertoire in 98 persistent expansions of CD3(+)/TCR-alphabeta(+) large granular lymphocytes: utility in assessing clonality and insights into the pathogenesis of the disease Am J Pathol 2001 159 1861 1868 11696446
| 15575964 | PMC539294 | CC BY | 2021-01-04 16:36:33 | no | Mol Cancer. 2004 Dec 3; 3:34 | utf-8 | Mol Cancer | 2,004 | 10.1186/1476-4598-3-34 | oa_comm |
==== Front
Cardiovasc UltrasoundCardiovascular Ultrasound1476-7120BioMed Central London 1476-7120-2-291559835210.1186/1476-7120-2-29ReviewUltrasound imaging versus morphopathology in cardiovascular diseases. Coronary atherosclerotic plaque Baroldi Giorgio [email protected] Riccardo [email protected] Lauro [email protected] Institute of Clinical Physiology, National Research Council, Milan and Pisa, Italy2 University School of Medicine and "A. De Gasperis" Foundation, Niguarda Hospital. Milan, Italy3 Cardiovascular Unit, "Campo di Marte" Hospital, Lucca, Italy2004 14 12 2004 2 29 29 9 8 2004 14 12 2004 Copyright © 2004 Baroldi et al; licensee BioMed Central Ltd.2004Baroldi 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.
This review article is aimed at comparing the results of histopathological and clinical imaging studies to assess coronary atherosclerotic plaques in humans. In particular, the gap between the two techniques and its effect on the understanding of the pathophysiological basis of coronary artery disease is critically discussed.
==== Body
Introduction
Amongst the clinical approaches ultrasound imaging is one of the more promising technique to understand dysfunction. The need is to compare morphopathological counterpart to have a correct pathophysiological interpretation.
In four reviews, morphopathology of the main cardiovascular disorders, in relation to the status of art of clinical imaging will be presented. The aim is to recall the pathological anatomy to stimulate ultrasound experts to further sharpen their technology till "histological" perfection.
The present first review concerns the coronary atherosclerosis since the current dogma of "unifying theory" assumes that the acute coronary syndromes, namely unstable angina, myocardial infarct and sudden death, are caused by atherosclerotic plaque rupture or "explosion" with occlusive thrombus formation preceded by intramyocardial emboli. An assumption which implies to discover a clinical imaging able to show when a coronary atherosclerotic plaque becomes vulnerable i.e. prone to rupture.
The risk of the latter has been correlated with large lipid core (atheroma in our definition), thin fibrous cap (< 65 μm) covering atheroma. Therefore any imaging should have a 50 μm resolution to identify a fibrous cap prone to rupture, 100 μm or 150 μm thickness being respectively at low or minimal risk. Matrix-digesting enzymes released from inflammatory cells (monocytes, macrophages, T-cells, B-cells, neutrophils, mastcells) may contribute to plaque rupture. An attractive approach since, despite many years of preventive and therapeutical attempts, coronary heart disease (CHD) remains the main cause of death and morbidity in advanced societies and selects people at the top of their work skillness and productivity.
The first question is whether ultrasound imaging may or may not discriminate extent and morphology of plaque variables seen within the intima.
Coronary atherosclerotic plaque
Physiological intimal thickening
Morphology of the atherosclerotic plaque has been described in textbook and articles [1-4]. In comparing many contributions, the major difficulty is to discriminate among different morphologic patterns selected in different groups of patients according to unclear definitions and without distinction between the plaque obtained by hypercholesterol diet and that found in the general population. The other need is to consider the evolution of a plaque in each single arterial system since anyone has its own peculiarity with different response to blood flow dynamics. In this sense, only the coronary arteries have a diphasic blood flow in relation to the cycle of myocardial contraction, i.e. filling of extramural vessels without intramural flow in systole because contracted myocardium compresses intramyocardial vessels. The result is an excess of systolic radial, circumferential, longitudinal and drag pressures on the wall of arteries and branches free to expand on the cardiac surface. In turn, this diphasic hemodynamic induces a structural response of the coronary intima which starts as smooth muscle cell proliferation from the tunica media, followed by elastic fiber hyperplasia ending in fibrosis of the whole intima without lumen reduction (Fig 1). First described by Wolkoff in 1929, such thickening becomes greater in the second decade in contrast to its absence in other human muscular arteries (e.g. brain arteries) or in animal with a similar coronary artery anatomy and diphasic flow (Fig. 2). The latter finding suggests that hemodynamic pressures may have the intimal hyperplastic effect in association with the neurovegetative regulation of arterial wall tone particularly active in humans in relation to heart function. Any clinical imaging of the coronary wall should consider this physiological intimal thickening [3-5] which in normal adult hearts measures about 200 μm, does not show any atherosclerotic variable, may become thicker in hypertrophic hearts with normal coronary artery and is greatly reduced or absent in segments of extramural coronary artery embedded ("mural artery") within the myocardium. The latter abolishes the systolic arterial wall expansion (Fig. 2).
Figure 1 Coronary physiologic intimal thickening. This changes starts as nodular (already visible at birth at the site of vessel bifurcation) smooth muscle cells (A) and elastic fibrils (B) hyperplasia which in the second decade is diffuse to the whole intimal surface of all extramural arterial vessels. With aging there is a progressive increase of fibrous tissue which substitutes myo-elastic tissue (C, D) with final total, anelastic, fibrosis (E, F). Arteriosclerosis distinct from atherosclerosis
Figure 2 Comparison between the intimal thickening of the LAD (A) and the middle cerebral artery (B) of the same 18-year old subject. In the latter artery the intimal thickening is minimal in contrast to that of LAD which is circumferential with a thickness greater than tunica media. C), difference in maximal thickness in microns found in main coronary arteries and branches in respect of the middle cerebral artery. D), absence of intimal thickening in the LAD of dogs, despite and identical morpho-function. This suggests a possible role of the neurogenic control of coronary arteries in humans. On the other hand the absence of intimal thickening in the "mural tract" of coronary arterial vessels (E) emphasizes the role of systolic dynamic stresses on arterial wall free to expande versus those protected by encircling contracted myocardium.
The first conclusion is that such intimal thickening is a physiologic structural respons to hemodynamics and not the initial phase of atherosclerosis as claimed by some authors.
Coronary atherosclerotic intimal thickening
In contrast to the uniformely diffuse physiological intimal thickening, the atherosclerotic one is focal and protrudes within the lumen which is progressively reduced. In order to quantitatively study this progression, we sampled systematically in each heart the first tract of the main left trunk (LCA), left descending (LAD) and circumflex (LCX) branches, right coronary artery (RCA), posterior descending branch (PD) and the middle tract of LAD and marginal and posterior tracts of RCA. These selected tracts correspond to the sites where atherosclerotic changes generally occur. The coronary arterial sampling was performed in 100 fatal cases of acute myocardial infarct without other diseases and not undergone invasive techniques; 208 cases of sudden and unexpected coronary death (SUD) which occurred in apparently normal people, acting their usual life, without resuscitation attempts and autopsy findings limited to coronary atherosclerosis of any degree, myocardial necrosis or scar, with or without cardiac hypertrophy: 50 cases with chronic angina pectoris who died within 25 day after coronary by-pass surgery; and 97 normal subjects who died by accident without pathological findings at autopsy but coronary atherosclerosis. In a total of 3,640 coronary sections the following variables were quantified:
1. Lumen reduction calculated in percent of the normal diameter measured in normal coronary arteries and branches injected by plastic substance under pressure. Measurement often referred to the cross-sectional area within the internal elastic membrane may result in severe stenosis despite a normal lumen since the atherosclerotic plaque may enlarge the cross-section.
2. Shape of residual lumen: concentric if encircled by pathological intima or semilunar when an arch of the wall was normal.
3. Length calculated in number of segments involved by the plaque, all extramural coronary arteries being sistematically cross-sectioned at 3 mm interval.
4. Intimal and tunica media thickness measured in microns.
5. Atherosclerotic changes within the intima: fibrosis alone, basophilia i.e. proteoglycan accumulation, atheroma or lipoprotein/cholesterol material, calcification, vascularization, hemorrhage, adventitial and intimal lymphocytic infiltration. All these variables were expressed in percent of the total intima but vascularization calculated in number of vessels found.
Amongst 1,519 sections without lumen reduction with an intimal physiological thickness less of 300 μm we never found subendothelial or internal lipoprotein/cholesterol infiltration or deposit (fatty streaks), monocytes or macrophages or foam cells, platelet aggregates, fibrin-platelet thrombi or inflammatory elements. A similar negative finding was observed in 1,315 coronary sections with a lumen reduction less than 69% and pathological intimal thickness less than 600 μm. In all these sections we were unable to demonstrate an intimal fissuration. We must emphasize that in selecting our material cases of familial hypercholesterolemia were excluded.
In general the atherosclerotic intimal variables increased in frequency and extent in parallel with the lumen reduction and pathological intimal thickeness with the exception of proteoglycan accumulation less found in stenoses >90% and intimal thickness >2000 μm. Of 990 sections with calcification, 488 (49%) had mild stenosis, calcification being severe in 162 (33%). This finding indicates that calcification per se does not necessarily means a severe lumen reduction. The less frequent variable was intimal hemorrhage mainly seen in plaques located in a vessel tributary to an acute infarct.
When different groups of CHD and normal people were matched, an excess of atheroma, hemorrhage, calcification and adventitial/intimal lymphocytic inflammation was observed in AMI group while a significant defect was present in normal subjects with the same degree of stenosis. In synthesis, two other main findings are worth of mention: 1) proteoglycan accumulation is a relatively late event which occurred in the deep layer of the intima near to the tunica media and below the fibrous cap of a plaque. Lipoprotein/cholesterol plus macrophages (foam cell) and/or calcium salts appeared only in this proteoglycan pool in agreement with their interaction with glycosaminoglycans (Fig. 3); 2) adventitial inflammation showed a peculiar tropism for the nervous structures related to the media at plaque level only (medial neuritis). An inflammatory process which involved all plaques present in each CHD patient while absent or limited to one plaque in normal controls.
Figure 3 Natural history of the coronary atherosclerotic plaque in general population, including most of CHD patients. The starting point is a nodular hyperplasia of smooth muscle cells and elastic tissue with progressive fibrous replacement. No other changes as subendothelial lipo-protein-cholesterol storage, inflammatory process of any type, platelet aggregation and/or fibrin-platelet thrombi are found (A). Proteoglycan accumulation in the deep intima between tunica media and the fibrous cap is the second step (B). In this proteoglycan pool, lipo-protein/cholesterol cleft, in macrophages ("foam cells") and/or Ca++ salts appear. Vascularization of the plaque and hemorrhage (C) follow. In the stage of proteoglycan accumulation, lympho-plasmacellular infiltrates occur in the adventitia and intima (C) with specific localization, around adventitial nerves closed to the tunica media (medial neuritis) (D, E). This natural history is totally different from that obtained experimentally by hypercholesterol diet in animals free of spontaneous atherosclerosis or in the small group of patients with familial hypercholesterolemia (F), in which transendothelial lipo-protein insudation is the starting point.
This study induced the recognition of two types of coronary atherosclerotic plaque: one, which belongs to the general population, (including CHD patients) and starts as nodular intraluminal proliferation of smooth muscle cells followed by elastic tissue hyperplasia and final substitution by fibrous tissue. Subsequently, a deep proteoglycan pool forms and becomes a deposit of lipoprotein/cholesterol and/or calcium salts. The recurrence of these phenomena explains the radial, circumferential, longitudinal progression of the coronary plaque resulting in increasing lumen reduction. This type of myohyperplastic plaque is totally different from the hypercholesterol plaque obtained experimentally by hypercholesterol diet in animals free of atherosclerosis or observed in a small group of patients with familial hypercholesterolemia. In literature too often the hypercholesterol plaque is taken as a model of an atherosclerotic plaque in man [1]. In timing the sequence of the events is important to stress that the inflammatory lymphocytic-plasmacellular process (autoimmune phenomenon?) starts after the proteoglycan insudation, being a relatively late complication.
The recurrent basic changes in myohyperplastic plaque (smooth muscle cell hyperplasia, fibrosis, proteoglycan accumulation with atheroma and/or calcification) explain the various intimal aspects amongst different plaques and different tracts of the same plaque (Fig. 4). A synopsis comparing dogma versus our findings is given in Table 1:
Figure 4 Coronary atherosclerosis. Different aspect of a severe, pin-point lesion (arrow). Plaque with prevailing atheroma (A) or fibrosis (B). Plaque with pale, large zone of proteoglycan accumulation (C) or with small atheroma plus hemorrhage and proteoglycans associated with critical stenosis occluded by an acute thrombus (D). Sequence in the same plaque of "rupture" (E) followed by severe hemorrhagic atheroma with minimal, linear lumen (arrow) without occlusion(F). Occlusive thrombosis connected with hemorrhagic atheromasia at the site of a critical stenosis (G). Semilunar stenosis (H) with a normal half wall and minor lumen reduction. The concept of vessel wall remodeling to compensate plaque growth has not any support (very low frequency of this type of lesion versus severe concentric lumen reduction in the natural history of coronary heart disease).
Targets of ultrasound diagnosis: the present and the future
Different echocardiographic techniques have been employed in the attempt to provide adequate visualization of coronary arteries (Table 2). However, intracoronary ultrasound (ICUS) represents the most valuable method to assess plaque morphology. Given to its limited resolution (approximately 0.3 mm), angiography is a fairly imprecise measure of luminal morphology and size. In particular, it is deficient in providing adequate distinction between plaque and lumen irregularities and assessing the extent of atherosclerotic disease [6]. Both these issues appear to be much well defined by ICUS. Firstly, the tomographic orientation of ultrasound enables a visualization of the full circumference of the vessel wall and, therefore, a more accurate assessment of size [7-9]. In addition, it allow us to overcome the false assumption that the nonstenotic region surrounding a discrete stenosis is normal and, therefore, to obtain an unbiased assessment of the plaque burden at the site of the stenosis. Finally, the penetrating nature of ultrasound provides unique images of the atherosclerotic plaque.
Table 2 Echocardiographic approach to coronary arteries.
Technique % success Image quality Anatomic information (plaque) Functional information (flow)
Transtoracic 20 ± ± ±
Transesophageal 80 + + ++
Epicardic 90 ++ ++ -
Intracoronary 95 +++ +++ -
- = poor; + = sufficient; ++ = good; +++ = very good
ICUS image analysis has been extensively used for determining plaque composition [10-13]. However, there are some limitations to this approach: 1) digitizing videotapes is time-consuming and therefore not suitable for real-time analysis; (2) image resolution is reduced to that of videotape, approximately 330 μm; (3) parameters such as gain, including time gain compensation and intensity, can be adjusted by the operator, thereby adding variability to the data set; (4) the dynamic range, pre- and postprocessing of the images depend on the analog-to-digital converters used in the ICUS consoles; (5) finally, due to the small dimension of the transducer, the transmitted acustic energy is low. Thus, some concerns still exist on whether this technique is ready to go for clinical use. In particular, definition of plaque composition seems not enough reproducible to provide an alternative independent standard to quantitative histology [14].
New technical development based on noninvasive molecular imaging [15,16], such as the use of novel targeted contrast agents able to identify fibrin deposited within plaque microfissures [17], adhesion or thrombogenic molecules expressed on endothelium of vulnerable plaques [18-22], matrix metalloproteinases in the cores of progressing lesions [23], or even early angiogenic expansion of the vasa vasorum that supports plaque development [24], will contribute to fill the gap between information derived from direct, quantitative histology and ultrasound imaging. Moreover, spectral analysis of the radiofrequency signal allows a more detailed analysis of various vessel components than does image analysis of digitized videotape images and can be potentially employed in real-time. This approach is expected to improve tissue characterization [25].
Comment
The working hypothesis is to stabilize the atherosclerotic plaque by increasing the thickness of the fibrous cap or by regression of atheroma burden (1). However, present imaging techniques (coronary cineangiography, angioscopy, contrast magnetic resonance, contrast echocardiography, nuclear scintigraphy, etc.) cannot provide adequate clinical evaluation of plaque vulnerability. In particular, ICUS is unable to provide discrimination between physiological and pathological intimal thickening and to define the shape of plaque, i.e. concentric or semilunar. In fact, amongst 2121 coronary sections at the site of maximal lumen-diameter reduction the stenosis was concentric in 70% of the cases (99% in supplying vessels of an acute myocardial infarct). Furthermore, in 408 CHD patients the maximal stenosis in each single case was less than 69% in 68, 70%, in 67, 80% in 109 and >90% in 164. These data mean that the residual lumen ranged from 900 to less than 50 μm and catheter of 1500 μm in most instances must break the plaque, being the residual lumen too small. Therefore, shape and contour of a plaque can be altered with a misleading higher frequency of semilunar stenosis giving an erroneous support to the questionable concept of vessel wall remodelling following an atherosclerotic plaque formation.
According to the previous data the main conclusions are: 1. The natural history of coronary atherosclerotic plaque in the general population, inclusive of CHD patients, is different from plaques obtained by experimental hypercholesterol diet or found in familial hypercholesterolemia. Most data refer to the latter as a valid model for the human plaque. In particular, fatty streaks is not the starting change of the myohyperplastic atherosclerotic plaque; 2. Emphasis is given to a "macrophagic inflammation" as source of proteolytic and/or thrombogenic moleculae causing plaque rupture. However, macrophagic reaction belongs to a repair process to digest necrotic or extraneous material rather than typical elements of an inflammatory process, as lymphocytes, neutrophils. The assumption that on increased number of labelled macrophages may indicate a risk of rupture is questionable in human coronary myohyperplastic plaque.
In the present review we have discussed the behaviour and meaning of components of the human coronary atherosclerotic plaque to emphasize the inconsistency of the current myths:
1. Experimental hypercholesterol model and correspondent human conditions do not represent the natural history of atherosclerosis in coronary arteries in the general population.
2. Physiological intimal thickening can not be interpreted as starting point of the atherosclerotic process.
3. Fatty streak does not represent the early atherosclerotic lesion.
4. Calcification is not synonymous of severe stenosis.
5. Hemorrhage is not consequent to endothelial fissuration.
6. Prevention of macrophage "inflammation" as source of substances able to disrupt the fibrous cap allowing rupture and thrombosis as well as identification of the thickness of fibrous cap to diagnose a vulnerable plaque may have little, if any, sense. Rupture and thrombosis may be secondary phenomena and not the cause of an acute coronary syndrome.
7. Degree and number of severe coronary plaques do not predict onset, course, complications and death in CHD.
Authors' contributions
Prof. Giorgio Baroldi contributed to the conception and organization of this review and to the final comments.
Dr. Riccardo Bigi and Dr. Lauro Cortigiani summarized the use of ultrasound techniques in atherosclerotic plaque imaging
Table 1 Natural History of Human Coronary Atherosclerotic Plaque
Beliefs Facts
Transendothelial lipoprotein/cholesterol infiltration Nodular smooth myocell-elastic hyperplasia protruding in lumen
Fatty streaks Fibrous substitution
Macrophagic "inflammation" Proteoglycan accumulation below fibrous cap between media/intima
Necrotic core-atheroma under fibrous cap Interstitial/macrofagic (foam cells) storage of lipoprotein-cholesterol and/or calcium salts in proteoglycan pool
Rupture fibrous cap Plaque tridimensional growth by recurrence of previous phenomena
Thrombosis-Embolization Late vascularization of atherosclerotic intima
Divergency between dogma and heretic view is easily explained by the fact that the former is founded on experimental hypercholesterolemic plaque which more or less may correspond to plaques in the relatively small group of familial hypercholesterolemia in humans. Furthermore, the main arterial vessel examined was the aorta and in human pathology advanced plaque was not studied comparing study in different CHD patterns and pathologic and normal controls.
==== Refs
Aikawa M Libby P The vulnerable atherosclerotic plaque. Pathogenesis and therapeutical approach Cardiov Pathol 2004 13 125 138 10.1016/S1054-8807(04)00004-3
Burke AP Virmani R Galis Z Haudenschild CC Muller JE Task Force. 34th Bethesda Conference: Task force #2– What is the pathologic basis for new atherosclerosis imaging technique J Am Coll Cardiol 2003 41 1874 1886 12798554 10.1016/S0735-1097(03)00359-0
Baroldi G Silver MD Mariani F Giuliano G Correlation of morphological variables in the coronary atherosclerotic plaque with clinical patterns of ischemic heart disease Am J Cardiovasc Pathol 1988 2 159 172 3207498
Baroldi G Silver MD The etiopathogenesis of coronary heart disease: a heretical theory based on morphology Medical Intelligence Unit, Eurekah-Com, Landes Bioscience, Georgetown, Texas, USA 2004 2
Baroldi G Silver MD, Gotlieb AI, Schoen FJ Myocardial cell death, including ischemic heart disease and its complications In Cardiovascular Pathology 2001 3 Churchill Livingstone New York 198
Mintz GS Popma JJ Pichard AD Kent KM Satler LF Chuang YC DeFalco RA Leon MB Limitations of angiography in the assessment of plaque distribution in coronary artery disease: a systematic study of target lesion eccentricity in 1446 lesions Circulation 1996 93 924 931 8598083
Yock P Fitzgerald P Popp R Intravascular ultrasound Sci Am Science Med 1995 2 68 72
Gussenhoven WJ Essed CE Frietman P Mastik F Lancee C Slager C Serruys P Gerritsen P Pieterman H Bom N Intravascular echographic assessment of vessel wall characteristics: a correlation with histology Int J Card Imaging 1989 4 105 116 2527913
Nishimura RA Edwards WD Warnes CA Reeder GS Holmes DR JrTajik AJ Yock PG Intravascular ultrasound imaging: in vitro validation and pathologic correlation J Am Coll Cardiol 1990 16 145 154 2193046
Gussenhoven EJ Essed CE Frietman P van Egmond F Lancee CT van Kappellen WH Roelandt J Serruys PW Gerritsen GP van Urk H Intravascular ultrasonic imaging: Histologic and echographic correlation Eur J Vasc Surg 1989 3 571 576 2696650
Dixon KJ Vince DG Cothren RM Cornhill JF Characterization of coronary plaque in intravascular ultrasound using histological correlation Annu Int Conf IEEE Eng Med Biol Proc 1997 2 530 533
Vince DG Dixon KJ Cothren RM Cornhill JF Comparison of texture analysis methods for the characterization of coronary plaques in intravascular ultrasound images Comput Med Imaging Graph 2000 24 221 229 10842046 10.1016/S0895-6111(00)00011-2
Coronary artery imaging with intravascular high-frequency ultrasound. [see comments] Circulation 1990 81 1575 1585 2184946
Zhang X McKay C Sonka M Tissue characterization in intravascular ultrasound images IEEE Trans Med Imaging 1998 17 889 899 10048846 10.1109/42.746622
Weissleder R Molecular imaging: exploring the next frontier Radiology 1999 212 609 614 10478223
Wickline SA Lanza GM Molecular imaging, targeted therapeutics, and nanoscience J Cell Biochem 2002 39 90 97 10.1002/jcb.10422
Flacke S Fischer S Scott MJ Fuhrhop RJ Allen JS McLean M Winter P Sicard GA Gaffney PJ Wickline SA Lanza GM Novel MRI contrast agent for molecular imaging of fibrin: implications for detecting vulnerable plaques Circulation 2001 104 1280 1285 11551880
Lanza GM Abendschein DR Hall CS Marsh JN Scott MJ Scherrer DE Wickline SA Molecular imaging of stretch-induced tissue factor expression in carotid arteries with intravascular ultrasound Invest Radiol 2000 35 227 234 10764091 10.1097/00004424-200004000-00003
Lanza GM Abendschein DR Hall CS Scott MJ Scherrer DE Houseman A Miller JG Wickline SA In vivo molecular imaging of stretch-induced tissue factor in carotid arteries with ligand-targeted nanoparticles J Am Soc Echocardiogr 2000 13 608 14 10849515 10.1067/mje.2000.105840
Kang HW Josephson L Petrovsky A Weissleder R Bogdanov AJ Magnetic resonance imaging of inducible E-selectin expression in human endothelial cell culture Bioconj Chem 2002 13 122 127 10.1021/bc0155521
Development of echogenic, plasmid-incorporated, tissue-targeted cationic liposomes that can be used for directed gene delivery Invest Radiol 2000 35 732 738 11204800 10.1097/00004424-200012000-00007
Lindner JR Song J Christiansen J Klibanov AL Xu F Ley K Ultrasound assessment of inflammation and renal tissue injury with microbubbles targeted to P-selectin Circulation 2001 104 2107 2112 11673354
Chen J Tung CH Mahmood U Ntziachristos V Gyurko R Fishman MC Huang PL Weissleder R In vivo imaging of proteolytic activity in atherosclerosis Circulation 2002 105 2766 2771 12057992 10.1161/01.CIR.0000017860.20619.23
Winter PM Morawski AM Caruthers SD Fuhrhop RW Zhang H Williams TA Allen JS Lacy EK Robertson JD Lanza GM Wickline SA Molecular imaging of angiogenesis in early-stage atherosclerosis with avb3-integrintargeted nanoparticles Circulation 2003 108 2270 2274 14557370 10.1161/01.CIR.0000093185.16083.95
Kawasaki M Takatsu H Noda T Sano K Ito Y Hayakawa K Tsuchiya K Arai M Nishigaki K Takemura G Minatoguchi S Fujiwara T Fujiwara H In vivo quantitative tissue characterization of human coronary arterial plaques by use of integrated backscatter intravascular ultrasound and comparison with angioscopic findings Circulation 2000 105 2487 2492 12034654 10.1161/01.CIR.0000017200.47342.10
| 15598352 | PMC539295 | CC BY | 2021-01-04 16:38:29 | no | Cardiovasc Ultrasound. 2004 Dec 14; 2:29 | utf-8 | Cardiovasc Ultrasound | 2,004 | 10.1186/1476-7120-2-29 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-5-1921558828810.1186/1471-2105-5-192SoftwareABC: software for interactive browsing of genomic multiple sequence alignment data Cooper Gregory M [email protected] Senthil AG [email protected] Arend [email protected] Department of Genetics, Stanford University, Stanford, CA 94305-9010, USA2 Department of Pathology, Stanford University, Stanford, CA 94305-5324, USA2004 8 12 2004 5 192 192 22 10 2004 8 12 2004 Copyright © 2004 Cooper 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
Alignment and comparison of related genome sequences is a powerful method to identify regions likely to contain functional elements. Such analyses are data intensive, requiring the inclusion of genomic multiple sequence alignments, sequence annotations, and scores describing regional attributes of columns in the alignment. Visualization and browsing of results can be difficult, and there are currently limited software options for performing this task.
Results
The Application for Browsing Constraints (ABC) is interactive Java software for intuitive and efficient exploration of multiple sequence alignments and data typically associated with alignments. It is used to move quickly from a summary view of the entire alignment via arbitrary levels of resolution to individual alignment columns. It allows for the simultaneous display of quantitative data, (e.g., sequence similarity or evolutionary rates) and annotation data (e.g. the locations of genes, repeats, and constrained elements). It can be used to facilitate basic comparative sequence tasks, such as export of data in plain-text formats, visualization of phylogenetic trees, and generation of alignment summary graphics.
Conclusions
The ABC is a lightweight, stand-alone, and flexible graphical user interface for browsing genomic multiple sequence alignments of specific loci, up to hundreds of kilobases or a few megabases in length. It is coded in Java for cross-platform use and the program and source code are freely available under the General Public License. Documentation and a sample data set are also available .
==== Body
Background
Functional elements in a genome accumulate inter-specific substitutions more slowly than neutral DNA throughout evolution [1]. Therefore, comparing orthologous genomic sequences from related species is useful for the identification of elements that play important roles in the biology of an organism [2-7]. While the statistical and computational methods for extracting comparative information are variable, the types of data involved are generally quite similar. First, a multiple sequence alignment is necessary. Second, a vector of quantitative scores is produced that describes the similarity of the nucleotides observed in small windows, or individual columns, of the alignment; percent identity is the metric used by the popular program VISTA [8], while a variety of other scoring methods also exist [9-12]. Third, annotations are generated that highlight regions of the alignment that are under constraint or meet some other quantitative threshold. Fourth, annotations of features like transcripts, promoters, coding exons, and repeats provide functional context. Finally, the phylogenetic tree that relates the aligned sequences is important for both performing comparative analyses and for interpreting their results.
Simultaneous visualization of complex data such as these is of utmost importance both for experimentalists and for computational biologists. Several options currently exist for such visualization, but there are a variety of characteristics that distinguish the ABC from them. VISTA, for example, generates a static image that is not interactive [8]. Other popular browsers such as phylo-VISTA [13] and PipMaker [14,15] require the use of a particular alignment program and scoring scheme. Also, the ABC is not suited for genome-wide visualization. Other tools exist for this and are quite useful for the browsing of very large genomic intervals and major evolutionary events such as genomic rearrangements [16-19]. However, these programs are generally part of larger, more complex interfaces that are not necessarily ideal for targeted analysis of an individual alignment or genomic locus. Finally, we note that the ABC allows many annotation types and colors, is not web-based and can be used on a local machine as intensively as necessary, and the source code is open and freely available allowing users to modify and add features if desired.
Implementation
The ABC requires Java 1.4 or later and has been successfully tested on Windows, Linux, and OS X. There is no specific upper limit in the size of potential data sets, but system memory usage can be high on large alignments. The ABC can efficiently handle a 2 Mb alignment of 29 sequences on a machine with a 1 GHz processor and 1 Gb of RAM. Details about file formats and instructions for use are available in the documentation that is available along with the source code . The file formats used by the ABC are quite similar, with only minor modifications, to other standard formats, such as fasta-formatted sequence files and standard parenthesis notation for phylogenetic tree descriptions. A sample data set is available, and is the source of the screenshot depicted in Figure 1. The sequence data are derived from a previously published analysis [20]; it includes ~300 kb of sequence from 9 mammals, centered around the ST7 gene in the human genome, near the CFTR gene. Repeats were identified in the human sequence using RepeatMasker [21] and genes identified using RefSeq annotations from the UCSC genome browser [17]. The alignment was generated using MLAGAN [22], and has been compressed so that the human sequence is ungapped; annotation of human sequence features is thus identical to the alignment annotations shown in Figure 1. A description of the method used to score the alignment columns and identify constrained elements, along with Perl scripts that facilitate this method, including export of results in ABC-ready formats, will be described elsewhere (in preparation). Please note that the ABC will not translate coordinates from alignment to sequence coordinates (or vice versa); the annotations that the user supplies must be appropriate to the alignment being analyzed.
Results and discussion
By default, the ABC displays graphical summaries of the quantitative information associated with the alignment. Scores are summarized regionally in consecutive non-overlapping windows. The size of these windows depends on the resolution, defined as the number of alignment columns summarized per pixel. The ABC has three distinct display modes, chosen automatically depending on the density of the information. At very low resolution, a histogram is displayed that plots the number of data points in each window that are at or below a specified value; note that regions containing many low scores will stand out as peaks in the histogram, as demonstrated by the clear association of peaks and the location of exons (Figure 1, top panel). At intermediate resolutions, a 'wiggly plot' is displayed, in which the average score for each regional window is plotted; in this case, regions containing many low scores will appear as valleys in the plot (Figure 1; middle panels). Finally, at very high resolution, the user may view the sequence data directly, along with the sequence names and a tree relating the sequences (Figure 1; lower panel).
A mobile and scalable zoom window allows for exploration of the summary views (Figure 1; upper panels). The user may drag and resize this rectangle, and when a desired region is selected a more detailed view can be obtained. This region will be expanded immediately below the parent display, with the resolution, score plot, and annotation adjusted accordingly (Figure 1; compare the start and stop coordinates of the black rectangle in the top panel to the start and stop coordinates of the entire panel immediately below). At all resolutions, annotation tracks are displayed immediately above the score/sequence display (Figure 1; all panels). An arbitrary number of tracks may be displayed, but the bottom two tracks are reserved for displaying information about transcripts with exons and introns. Colors for features can be specified individually using standard RGB notation. Other key features of the ABC include:
• Mouse-over highlighting to reveal annotations, scores, coordinates, etc
• Exporting of sequence, score, and annotation data
• Searching sequence data for particular nucleotide strings
• GoTo feature to quickly bring up a desired region
The ABC is flexible in that is has the ability to visualize diverse quantitative information and it has the capacity to display an arbitrary number of annotation types. It does not have a built-in scoring function; all data needs to be generated and formatted prior to being displayed in the ABC. While this may seem to be a drawback, it is in fact the intended function for an interface that has no preconceptions about the methodology that generated the data. Finally, the ABC is interactive, allowing the user to zoom in quickly from summary views of the comparative data to individual alignment columns. Zoom levels remain in the display, allowing the user to keep a birds-eye view of a large genomic region while focusing at much higher resolution on a small section within it.
Conclusions
The ABC is stand-alone alignment browsing software that is relatively easy to use and customize. While it was not designed as a genome-wide browser, it is well-suited for tasks associated with comparative sequence analysis: exploration of alignments of individual genomic loci; analyzing the relationship between known biological features and quantitative comparative data; visualizing results for researchers who develop and test methods for comparative sequence analysis; isolating sequence elements in a genomic locus for downstream applications like motif-discovery or primer design; generating graphics that characterize a multiple sequence alignment or region of an alignment; and potentially more applications that we have not yet considered. In our own research, for example, we have used it to display SNPs between different mouse strains in the context of a comparative alignment (not shown). This flexibility should be beneficial to researchers whose primary interest is comparative sequence analysis, but should also be valuable to those who use comparative analyses in support of other types of projects, such as experimental characterization of constrained elements. We also note that this flexibility distinguishes the ABC from other browsers that require built-in or specific types of score data and/or the use of a particular alignment program. The software is written in Java for cross-platform support, and the source code is freely available under the General Public License (GPL).
Availability and requirements
Project name: Application for Browsing Constraints (ABC)
Project home page:
Operating system: Platform independent
Programming language: Java
Other requirements: Java 1.4 or later
License: GPL
Abbrevations
ABC: Application for Browsing Constraints
GUI: Graphical User Interface
GPL: General Public License
SNP: Single Nucleotide Polymorphism
UCSC: University of California, Santa Cruz
RGB: Red-Green-Blue
Authors' contributions
GMC and AS conceived of the project, organized the feel and design of the browser, generated the underlying data, and wrote the manuscript. SAGS wrote the Java code and provided comments on the manuscript. All authors read and approved the final manuscript.
Acknowledgments
GMC is a Howard Hughes Medical Institute Pre-doctoral fellow. AS acknowledges support from NIH/NHGRI. We thank two anonymous reviewers for their comments.
Figures and Tables
Figure 1 A screenshot of the ABC. The upper-most panel consists of a histogram describing the regional density of columns with low scores, corresponding to regions of high sequence similarity, across the length of the entire alignment. Note the association of peaks in sequence similarity near the exons annotated in the lowest tracks ('Trx 1' and 'Trx 2'). The middle panels consist of 'wiggly plots' of the alignment scores; note the association in these panels between valleys in the plot and the location of exons. The black rectangle can be moved and resized to highlight particular regions. The panels represent increasingly zoomed in views of the regions highlighted within the rectangle, moving from 348 alignment columns per pixel at the top to 4 positions per pixel in the third panel. The bottom panel shows a view of individual alignment columns corresponding to a small region of the alignment. If supplied, a topology describing the relationships of the aligned sequences is also displayed here. Above all panels, four annotation tracks are displayed, the bottom two being reserved for transcripts with exons and introns. In this image, the third and fourth tracks reveal conserved regions and repeats, respectively.
==== Refs
Kimura M The neutral theory of molecular evolution 1983 Cambridge [Cambridgeshire] ; New York, Cambridge University Press xv, 367
Boffelli D Nobrega MA Rubin EM Comparative genomics at the vertebrate extremes Nat Rev Genet 2004 5 456 465 15153998 10.1038/nrg1350
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
Dermitzakis ET Reymond A Scamuffa N Ucla C Kirkness E Rossier C Antonarakis SE Evolutionary discrimination of mammalian conserved non-genic sequences (CNGs) Science 2003 302 1033 1035 14526086 10.1126/science.1087047
Göttgens B Barton LM Chapman MA Sinclair AM Knudsen B Grafham D Gilbert JG Rogers J Bentley DR Green AR Transcriptional regulation of the stem cell leukemia gene (SCL)--comparative analysis of five vertebrate SCL loci Genome Res 2002 12 749 759 11997341 10.1101/gr.45502
Hardison RC Comparative genomics PLoS Biol 2003 1 E58 14624258 10.1371/journal.pbio.0000058
Sidow A Sequence first. Ask questions later Cell 2002 111 13 12372296 10.1016/S0092-8674(02)01003-6
Mayor C Brudno M Schwartz JR Poliakov A Rubin EM Frazer KA Pachter LS Dubchak I VISTA : visualizing global DNA sequence alignments of arbitrary length Bioinformatics 2000 16 1046 1047 11159318 10.1093/bioinformatics/16.11.1046
Sumiyama K Kim CB Ruddle FH An efficient cis-element discovery method using multiple sequence comparisons based on evolutionary relationships Genomics 2001 71 260 262 11161821 10.1006/geno.2000.6422
Boffelli D McAuliffe J Ovcharenko D Lewis KD Ovcharenko I Pachter L Rubin EM Phylogenetic shadowing of primate sequences to find functional regions of the human genome Science 2003 299 1391 1394 12610304 10.1126/science.1081331
Margulies EH Blanchette M Haussler D Green ED Identification and characterization of multi-species conserved sequences Genome Res 2003 13 2507 2518 14656959 10.1101/gr.1602203
Cooper GM Brudno M Stone EA Dubchak I Batzoglou S Sidow A Characterization of evolutionary rates and constraints in three Mammalian genomes Genome Res 2004 14 539 548 15059994 10.1101/gr.2034704
Shah N Couronne O Pennacchio LA Brudno M Batzoglou S Bethel EW Rubin EM Hamann B Dubchak I Phylo-VISTA: interactive visualization of multiple DNA sequence alignments Bioinformatics 2004 20 636 643 15033870 10.1093/bioinformatics/btg459
Schwartz S Elnitski L Li M Weirauch M Riemer C Smit A Green ED Hardison RC Miller W MultiPipMaker and supporting tools: Alignments and analysis of multiple genomic DNA sequences Nucleic Acids Res 2003 31 3518 3524 12824357 10.1093/nar/gkg579
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
Kalafus KJ Jackson AR Milosavljevic A Pash: efficient genome-scale sequence anchoring by Positional Hashing Genome Res 2004 14 672 678 15060009 10.1101/gr.1963804
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
Kozik A Kochetkova E Michelmore R GenomePixelizer--a visualization program for comparative genomics within and between species Bioinformatics 2002 18 335 336 11847088 10.1093/bioinformatics/18.2.335
Chakrabarti K Pachter L Visualization of multiple genome annotations and alignments with the K-BROWSER Genome Res 2004 14 716 720 15060015 10.1101/gr.1957004
Cooper GM Brudno M Program NC Green ED Batzoglou S Sidow A Quantitative Estimates of Sequence Divergence for Comparative Analyses of Mammalian Genomes Genome Res 2003 13 813 820 12727901 10.1101/gr.1064503
Smit AFA Hubley R Green P RepeatMasker Open-3.0. <http://wwwrepeatmaskerorg> 1996
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
| 15588288 | PMC539296 | CC BY | 2021-01-04 16:02:45 | no | BMC Bioinformatics. 2004 Dec 8; 5:192 | utf-8 | BMC Bioinformatics | 2,004 | 10.1186/1471-2105-5-192 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-5-571558832910.1186/1471-2202-5-57Research ArticleMembrane trafficking and mitochondrial abnormalities precede subunit c deposition in a cerebellar cell model of juvenile neuronal ceroid lipofuscinosis Fossale Elisa [email protected] Pavlina [email protected] Janice A [email protected] Tanya [email protected] Allison M [email protected] Hanlin [email protected] Dorotea [email protected] Elena [email protected] Marcy E [email protected] Susan L [email protected] Molecular Neurogenetics Unit of Department of Neurology and Center for Human Genetic Research, Massachusetts General Hospital, Charlestown, MA, USA2 Department of Pharmacological Sciences and Center of Excellence on Neurodegenerative Diseases, University of Milano, Milan, Italy2004 10 12 2004 5 57 57 4 10 2004 10 12 2004 Copyright © 2004 Fossale et al; licensee BioMed Central Ltd.2004Fossale 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
JNCL is a recessively inherited, childhood-onset neurodegenerative disease most-commonly caused by a ~1 kb CLN3 mutation. The resulting loss of battenin activity leads to deposition of mitochondrial ATP synthase, subunit c and a specific loss of CNS neurons. We previously generated Cln3Δex7/8 knock-in mice, which replicate the common JNCL mutation, express mutant battenin and display JNCL-like pathology.
Results
To elucidate the consequences of the common JNCL mutation in neuronal cells, we used P4 knock-in mouse cerebella to establish conditionally immortalized CbCln3 wild-type, heterozygous, and homozygous neuronal precursor cell lines, which can be differentiated into MAP-2 and NeuN-positive, neuron-like cells. Homozygous CbCln3Δex7/8 precursor cells express low levels of mutant battenin and, when aged at confluency, accumulate ATPase subunit c. Recessive phenotypes are also observed at sub-confluent growth; cathepsin D transport and processing are altered, although enzyme activity is not significantly affected, lysosomal size and distribution are altered, and endocytosis is reduced. In addition, mitochondria are abnormally elongated, cellular ATP levels are decreased, and survival following oxidative stress is reduced.
Conclusions
These findings reveal that battenin is required for intracellular membrane trafficking and mitochondrial function. Moreover, these deficiencies are likely to be early events in the JNCL disease process and may particularly impact neuronal survival.
==== Body
Background
Juvenile neuronal ceroid lipofuscinosis (JNCL), or Batten disease, is a recessively inherited childhood-onset neurodegenerative disorder characterized by progressive blindness, seizures, motor and cognitive decline, and early death [1]. The primary genetic defect (>80% disease chromosomes) leading to JNCL is a 1.02 kb genomic DNA deletion in the CLN3 gene, which eliminates exons 7 and 8 and surrounding intronic DNA, predicting a non-functional protein product [2].
The pathological hallmark of JNCL is autofluorescent ceroid lipofuscin deposits within autolysosomes that are enriched in subunit c of the mitochondrial ATP synthase complex [3-5]. Remarkably, these deposits are not only found in CNS neurons but are also abundant in non-neuronal cells outside of the nervous system. The relationship of subunit c deposits to the JNCL disease process, and the underlying reason for the neuronal specificity of the disease remain poorly understood.
The CLN3-encoded protein (battenin, also called CLN3 or cln3 p) is a highly conserved, ubiquitously expressed, multi-pass membrane protein [6] that localizes to the lysosome and other vesicular compartments [7-9]. Battenin function remains to be elucidated, although studies of btn1, the yeast CLN3 ortholog, have implicated battenin in lysosomal pH homeostasis and amino acid transport [10,11].
To explore JNCL pathogenesis and battenin function, we previously generated a genetically precise JNCL mouse model. Cln3Δex7/8 knock-in mice harbor the ~1 kb common JNCL mutation and express a non-truncated mutant battenin isoform that is detectable with antibodies recognizing C-terminal epitopes. Homozygous Cln3Δex7/8 knock-in mice exhibit a progressive JNCL-like disease, with perinatal onset of subunit c deposition in many cell types and later onset of neuronal dysfunction and behavioral deficits [12]. These findings suggest that the major JNCL defect leads to abnormal turnover of mitochondrial subunit c, in a manner that selectively compromises CNS neurons.
Currently, there is no suitable neuronal cell system to investigate the impact of the common JNCL mutation on biological processes. Therefore, we have established cerebellar neuronal precursor cell lines from Cln3Δex7/8 knock-in mice. Homozygous CbCln3Δex7/8 cells exhibit pathological hallmarks of the disease, and a survey of membrane organelles revealed membrane trafficking defects and mitochondrial dysfunction in homozygous mutant CbCln3Δex7/8 cells.
Results
Generation of a genetically precise cerebellar JNCL cell model
To generate a precise genetic, neuron-derived JNCL cell culture system, we immortalized granule neurons cultured from postnatal day 4 (P4) cerebella of homozygous and heterozygous Cln3Δex7/8 knock-in mice, and wild-type littermates. Primary cell cultures enriched for granule neurons were transduced with retroviral vector bearing a selection cassette and temperature-sensitive tsA58 SV40 large T antigen. Growth in G418 containing medium at the permissive temperature (33°C) allowed for selection and isolation of multiple clonal nestin-positive (Fig. 1a), and GFAP-negative (Fig. 1b), cell lines for each genotype. No genotype specific differences were observed in cellular morphology or doubling time (~46 hours) (data not shown). As expected, SV40 large T antigen expression was rapidly lost and cell division ceased when cells were shifted to the non-permissive temperature (39°C) (data not shown). Upon addition of neuronal differentiation cocktail, precursor cells became neuron-like in morphology and exhibited decreased nestin expression (data not shown) and increased MAP2 and NeuN expression (Fig. 1c,1d), but not expression of the Purkinje marker, calbindin (Fig. 1e).
Figure 1 Neuronal marker expression in CbCln3+/+ cells Characterization of CbCln3+/+ cells by immunofluorescence with marker antibodies is shown. CbCln3+/+ precursors exhibit nestin expression (a) but not GFAP expression (b), consistent with a neuronal precursor identity. Upon stimulation with a differentiation cocktail (see Methods), CbCln3+/+ cells achieved neuron-like morphology, with rounded cell bodies and extension of processes, and MAP2 (c) and NeuN (d) expression was increased. CbCln3+/+ cells are negative for the Purkinje neuron marker calbindin (e). CbCln3+/Δex7/8 and CbCln3Δex7/8/Δex7/8 cell lines exhibited identical marker immunofluorescence results. a, b) 20 × magnification; c, d, e) 40 × magnification.
Homozygous CbCln3Δex7/8 cells express mutant battenin and display JNCL-like pathology
Homozygous Cb Cln3Δex7/8 cells were first examined for JNCL-like characteristics. Homozygous Cln3Δex7/8 knock-in mice express multiple Cln3 mRNA splice variants and mutant battenin protein that is detectable by batp1 antibody recognizing C-terminal epitopes [12]. To assess this molecular phenotype in CbCln3Δex7/8 cells, RT-PCR and anti-battenin (batp1) immunostaining were performed. As shown in Figure 2, Cln3 mRNA isoforms in wild-type and homozygous cells were similar to those observed in total RNA isolated from wild-type or homozygous Cln3Δex7/8 knock-in brain, respectively (Fig. 2). In addition, batp1 immunostaining detected mutant battenin product in homozygous CbCln3Δex7/8 cells, in a similar albeit reduced cytoplasmic, vesicular staining pattern as that seen in wild-type cells. Batp1 signal exhibited some overlap with the lysosomal marker, Lamp1, but had more significant overlap with early endosome antigen 1 (EEA1) and the late endosomal marker, Rab7 (Fig. 3). Only limited overlap was observed with recycling endosomes, as determined by transferrin receptor co-staining (data not shown). Intriguingly, Lamp1 and EEA1 immunocytochemical distribution were altered in homozygous CbCln3Δex7/8 cells, with less perinuclear clustering than in wild-type cells, and Rab7 staining was frequently less intense in homozygous CbCln3Δex7/8 cells (Fig. 3). Heterozygous CbCln3Δex7/8 cells contained a mixture of Cln3 mRNA products from both the wild-type allele and the mutant allele, and batp1 signal was similar to that seen in wild-type cells (data not shown).
Figure 2 RT-PCR of Cln3 mRNA in wild-type and homozygous CbCln3Δex7/8 cells Cln3 Exon1-forward, Exon 15-reverse RT-PCR products are shown, from total wild-type (+/+) or homozygous mutant (Δex7/8/Δex7/8) brain and cell line RNA. Brain and cell line RT-PCR reaction products had identical band patterns on ethidium-bromide stained agarose gels. Wild-type RT-PCR product was a single ~1.6 kb band and mutant products were ~1.6, ~1.5, ~1.4, ~1.35, and ~1.3 kb, representing multiple mutant splice variants.
Figure 3 Battenin and lysosomal and endosomal marker co-staining in wild-type and homozygous CbCln3Δex7/8 cerebellar precursor cells Batp1 immunostaining of wild-type (CbCln3+/+) and homozygous mutant (CbCln3Δex7/8/Δex7/8) cerebellar precursor cells is shown, with co-staining for lysosomes (Lamp 1), early endosomes (EEA1), and late endosomes (Rab7). Significant overlap of Batp1 signal (red) with EEA1 (green, middle panels) and Rab7 (green, bottom panels) can be seen as yellow when the two channels are merged (Merge). The degree of Batp1 overlap is greatest with Rab7. Only limited overlap between Batp1 (red) and Lamp 1 (green, top panels) can be seen. Batp1 signal in homozygous CbCln3Δex7/8 cells is significantly reduced, but significant overlap with EEA1 and Rab7, and very little Lamp 1 overlap, can be seen as yellow in the respective merged panels. Notably, Lamp 1 and EEA1 localization appear altered, and Rab7 staining was frequently less intense in homozygous CbCln3Δex7/8 cells. Wild-type and homozygous CbCln3Δex7/8 confocal images were captured with identical exposure settings. 60 × magnification.
During sub-confluent growth conditions, neither wild-type nor homozygous CbCln3Δex7/8 cells displayed autofluorescence or subunit c inclusion formation (data not shown). However, when cells were aged at confluency (3+ days post-confluency), homozygous CbCln3Δex7/8 cellular subunit c levels were elevated beyond normal wild-type levels by immunostaining (Fig. 4a) and immunoblot analysis (Fig. 4b). Autofluorescent signal sometimes overlapped with subunit c signal, but also was elevated more diffusely (Fig. 4a). Moreover, although multilamellar "fingerprint" profiles were not detected, confluency-aged homozygous CbCln3Δex7/8 cells displayed numerous ultrastructural abnormalities including electron dense inclusions characteristic of lipofuscin and large autophagosomes that contained dense core structures, degenerating mitochondria, and many smaller vesicles (Fig. 4c). Inclusion bodies and autophagosomes were infrequently observed in confluency-aged wild-type cultures (data not shown).
Figure 4 Subunit c accumulation in homozygous CbCln3Δex7/8 cerebellar precursor cells a. Subunit c immunostaining and autofluorescence of 7-day confluency-aged wild-type and homozygous CbCln3Δex7/8 cells is shown. Wild-type cultures (CbCln3+/+) exhibited limited subunit c immunostaining and autofluorescence. However, CbCln3Δex7/8/Δex7/8 cells contained numerous subunit c puncta. Autofluorescence (7 days AF) was also significantly elevated (right panels), although limited overlap with subunit c puncta was observed (arrows). 40 × magnification. b. Immunoblot analysis of subunit c protein at sub-confluency or 7-day confluency incubation is shown. Total protein extracts from sub-confluency wild-type (+/+) and homozygous mutant (Δex7/8/Δex7/8) cultures contained approximately equal levels of subunit c protein (α-sub c). 7-day confluency extract from homozygous CbCln3Δex7/8 cells (Δex7/8/Δex7/8) had elevated levels of subunit c protein (~1.5X), relative to wild-type extract (+/+). Protein levels were normalized to cytochrome c oxidase subunit IV (α-cox4). c. TEM analysis of inclusions in 7-day confluency-aged homozygous CbCln3Δex7/8 cells is shown. A large autophagosome contained by double membrane (arrows) is filled with degenerating mitochondria (Md), electron dense cores (left and right of *) and other smaller vesicular structures. A large electron-dense inclusion, with a lipofuscin (Ln) appearance, is also present. M, mitochondria. 10,000 × magnification.
Homozygous CbCln3Δex7/8 cells and Cln3Δex7/8 knock-in mice process cathepsin D abnormally
We next investigated the basis for subunit c accumulation, testing the hypothesis that cathepsin D is abnormal since it is required for ATP synthase subunit c degradation in the lysosome [13]. We first tested cathepsin D transport and processing in homozygous CbCln3Δex7/8 cells and Cln3Δex7/8 mice using anti-cathepsin D antibody that recognizes unprocessed and processed cathepsin D isoforms. Immunostaining of wild-type and homozygous CbCln3Δex7/8 cells revealed a perinuclear and punctate vesicular cathepsin D distribution, consistent with its transport and processing through the secretory pathway and delivery to the lysosome (Fig. 5a). However, in homozygous CbCln3Δex7/8 cells, cathepsin D distribution was less vesicular and more perinuclear-clustered than in wild-type cells. Immunoblots of homozygous CbCln3Δex7/8 cell and Cln3Δex7/8 tissue extracts also showed altered relative levels of cathepsin D isoforms (Fig. 5b). Cathepsin D isoforms, identified by relative molecular weights, represent the ~45 kDa precursor, the ~43 kDa intermediate single chain form of the enzyme, and the 31 kDa heavy chain of the double-chain form of the mature enzyme [14]. In homozygous CbCln3Δex7/8 cell and Cln3Δex7/8 tissue extracts, the precursor and heavy chains were reduced, and the single chain was slightly elevated compared to wild-type extracts. The cellular growth media did not contain altered levels of cathepsin D, indicating enzyme secretion was not affected. Heterozygous Cln3Δex7/8 mice and CbCln3Δex7/8 cells were indistinguishable from wild-type, as expected for a recessive disease phenotype (data not shown).
Figure 5 Cathepsin D localization and processing in wild-type and homozygous CbCln3Δex7/8 cells a. Immunostaining of wild-type and homozygous CbCln3Δex7/8 precursor cells with anti-cathepsin D antibody, recognizing unprocessed and processed forms of cathepsin D protein is shown. CbCln3+/+ cells (left panel) exhibited a perinuclear and cytoplasmic punctate signal. Cathepsin D signal in homozygous CbCln3Δex7/8 cells (right panel) was more often perinuclear, with less cytoplasmic punctate signal, compared to wild-type CbCln3+/+ cells. 40 × magnification. b. α-Cathepsin D-probed immunoblots of total wild-type versus homozygous Cln3Δex7/8 knock-in tissue or CbCln3Δex7/8 cellular extracts are shown. The ~45 kDa cathepsin D band, representing precursor, was the predominant band in wild-type (wt) tissue and cellular extracts, with lower levels of mature enzyme (single chain, ~43 kDa, and heavy chain, ~31 kDa). Conversely, homozygous Cln3Δex7/8 and CbCln3Δex7/8 mutant (m) extracts exhibited reduced levels of precursor and heavy chain of the double-chain form of the enzyme, with elevated levels of single-chain mature enzyme.
The impact of the altered cathepsin D processing on enzymatic activity was next tested to determine if altered enzymatic activity accounts for inefficient subunit c turnover. In a fluorogenic in vitro assay, cathepsin D activity in total cellular extracts was not significantly altered in homozygous CbCln3Δex7/8 cells (376 ± 89 RFU/μg total protein), versus wild-type cells (324 ± 58 RFU/μg total protein), although a consistent trend towards increased enzymatic activity in mutant cells was observed. Thus, cathepsin D transport and processing are disrupted in homozygous CbCln3Δex7/8 cells in a manner such that enzymatic activity appears to be relatively unaffected.
Homozygous CbCln3Δex7/8 cells show abnormal membrane organelles
The abnormal transport and processing of cathepsin D suggested membrane trafficking disruptions in homozygous CbCln3Δex7/8 cells; therefore, we surveyed the subcellular distribution and morphology of membrane organelles. Components of the secretory pathway, including the ER, cis-Golgi, and trans-Golgi, did not appear altered from wild-type appearance when labeled with the respective markers, protein disulfide isomerase (PDI), GM130, and VVL (data not shown). By contrast, the lysosomal markers, Lysotracker and Lamp 2 had significantly altered signal in homozygous CbCln3Δex7/8 cells, versus wild-type cells. Wild-type cells exhibited brightly stained lysosomes that were large and clustered in the perinuclear region whereas homozygous CbCln3Δex7/8 lysosomes were lightly stained, smaller vesicles that were more diffusely scattered in the cytoplasm of the cell (Fig. 6). Lamp 1 distribution was also altered, as previously noted (Fig. 3). However, Lamp 1 and Lamp 2 total protein levels were similar in wild-type and homozygous CbCln3Δex7/8 cells by immunoblot analysis, indicating the altered signal likely reflects dispersed lysosomes or altered localization and/or epitope availability (data not shown). It is noteworthy that Lysotracker dye, which selectively accumulates in acidic compartments, exhibited the most marked reduction in lysosomal labeling. This observation may reflect altered lysosomal pH, an established finding in JNCL [10,15].
Figure 6 Lysotracker and Lamp 2 labeling of wild-type and homozygous CbCln3Δex7/8 lysosomes Lysosomal labeling of wild-type and homozygous CbCln3Δex7/8 precursor cells with lysotracker and Lamp 2 antibody is shown. Lysotracker dye (top panels) labeled large, perinuclear-clustered lysosomes and scattered lysosomes in the periphery of wild-type cells (CbCln3+/+). Lysotracker stain was dramatically reduced in homozygous mutant cells (CbCln3Δex7/8/Δex7/8), with smaller labeled vesicles and less apparent perinuclear clustering. Lamp 2 (bottom panels) immunostaining also showed reduced signal intensity with less perinuclear clustering in homozygous CbCln3Δex7/8 cells, although the effect was somewhat less dramatic than that observed with Lysotracker dye. Wild-type and homozygous CbCln3Δex7/8 confocal images were captured with identical exposure settings. 60 × magnification.
Consistent with the altered early endosome marker (EEA1) signal observed by immunostaining (Fig. 3), fluid-phase endocytosis was also altered in homozygous CbCln3Δex7/8 cells, as measured by dextran-FITC uptake (Fig. 7). Following a 15-minute incubation in media containing dextran-FITC, wild-type and heterozygote cells displayed brightly stained, large endocytic vesicles that were clustered in the perinuclear region. However, homozygous CbCln3Δex7/8 cells were less brightly stained with most dextran-FITC signal localizing to smaller vesicles scattered throughout the cytoplasm of the cell.
Figure 7 Endocytosis in wild-type, heterozygous and homozygous CbCln3Δex7/8 cells Dextran-FITC uptake in wild-type, heterozygous and homozygous CbCln3Δex7/8 precursor cells is shown. In wild-type (CbCln3+/+, left panel) and heterozygous (CbCln3+/Δex7/8, middle panel) cells, dextran-FITC label was observed in a perinuclear-clustered vesicular pattern with scattered labeled vesicles also present in the periphery. In contrast, dextran-FITC label of homozygous mutant (CbCln3Δex7/8/Δex7/8, right panel) cells was reduced overall and exhibited smaller stained vesicles with less perinuclear clustering. Confocal images were captured with identical exposure settings. 40 × magnification.
Finally, because subunit c is a mitochondrial protein and its turnover proceeds through autophagic engulfment of mitochondria [13], we analyzed homozygous CbCln3Δex7/8 cell mitochondrial morphology and function. Mitochondrial distribution in homozygous CbCln3Δex7/8 cells was indistinguishable from wild-type and heterozygous cells; however, homozygous CbCln3Δex7/8 mitochondria appeared more elongated by grp75 marker immunostaining and TEM analysis (Fig. 8a). 72% of homozygous mutant mitochondria were greater than 0.6 μm in length (range = 0.26 μm to 2.75 μm), while fewer wild-type mitochondria (51%) reached this length (range = 0.15 μm to 2.29 μm). Mitochondrial width was not altered in homozygous CbCln3Δex7/8 cells (data not shown). Moreover, compared to wild-type or heterozygous cells, homozygous CbCln3Δex7/8 cells had significantly reduced cellular ATP levels (1.3 fold less, Fig. 8b) and exhibited reduced survival following hydrogen peroxide treatment (~50% of wild-type survival, Fig. 8c), suggesting impaired energy metabolism and oxidative stress response. Taken together, these data support impaired mitochondrial function in homozygous CbCln3Δex7/8 cells.
Figure 8 Mitochondrial morphology and function in wild-type, heterozygous and homozygous CbCln3Δex7/8 cells a. Confocal and TEM micrographs of wild-type and homozygous CbCln3Δex7/8 mitochondrial morphology are shown. Immunostaining with the inner mitochondrial membrane marker, grp75 (top panels) highlighted elongated mitochondria in homozygous mutant cells (CbCln3Δex7/8/Δex7/8), relative to wild-type mitochondria (CbCln3+/+) (insets, zoom = 2.75x). Mitochondrial distribution was not altered from the wild-type pattern. Elongated homozygous CbCln3Δex7/8 mitochondria were also observed by TEM analysis. 60 × magnification. b. Cellular ATP levels in wild-type, heterozygous and homozygous CbCln3Δex7/8 precursor cells are shown. Wild-type (open bar) and heterozygous (gray bar) CbCln3Δex7/8 cells contained ~39 μM ATP, while homozygous CbCln3Δex7/8 cells (black bar) contained ~1.3 fold reduced levels of ATP (~30 μM), which was statistically significant in a t-test (p < 0.0001). Wild-type and heterozygous CbCln3Δex7/8 cellular ATP levels were not statistically different from each other (p > 0.4). A representative of triplicate experiments is shown (n = 6 in each experiment). c. Cell survival following 24-hour hydrogen peroxide treatment is shown. Homozygous CbCln3Δex7/8 cells were ~2-fold more sensitive to oxidative stress by hydrogen peroxide treatment. Wild-type (circle) and heterozygous (triangle) CbCln3Δex7/8 cells exhibited ~50% survival rates with 75–100 μM H2O2, whereas homozygous CbCln3Δex7/8 cells (squares) had a ~50% survival rate with 50 μM H2O2. A representative of triplicate experiments is shown (n = 4 in each experiment).
Discussion
CbCln3Δex7/8 cerebellar precursor cells represent the first genetically accurate neuron-derived culture model of JNCL. Homozygous CbCln3Δex7/8 cells express mutant battenin and JNCL-hallmark mitochondrial ATPase subunit c accumulation, upon aging of cells at confluency. Moreover, this is the first study to indicate recessive endosomal/lysosomal membrane trafficking defects and mitochondrial dysfunction that precedes subunit c deposition in an accurate JNCL model. Importantly, these defects are likely to be early events in the JNCL disease process and may particularly impact neuronal function.
Abnormal cathepsin D localization and processing in homozygous CbCln3Δex7/8 cells and Cln3Δex7/8 mice likely reflects altered vesicular trafficking and/or lysosomal pH, which is known to impact cathepsin D processing [14,16]. Indeed, CLN3 overexpression in HEK-293 cells altered lysosomal pH and cathepsin D processing [17], and lysosomal pH homeostasis is disrupted in JNCL [10,15]. It is noteworthy that cathepsin B and the CLN2-encoded enzyme, TPPI, are also altered in JNCL [18-20]. Nevertheless, despite the cathepsin D protein alterations that are observed in homozygous CbCln3Δex7/8 cells, cathepsin D enzymatic activity does not appear to be reduced. Thus, decreased cathepsin D activity is unlikely to account for subunit c accumulation in JNCL.
Aging of homozygous CbCln3Δex7/8 cells at confluency is necessary to induce significantly accumulated subunit c protein. However, membrane organelle disruptions precede subunit c accumulation in homozygous CbCln3Δex7/8 cells, since they are observed without aging at confluency. Lysosomal and endosomal size and distribution are altered, and mitochondria are abnormally elongated and functionally compromised in sub-confluent homozygous CbCln3Δex7/8 cultures. These observations argue that membrane trafficking defects do not result from excessive subunit c accumulation compromising the lysosome, but rather are early events in the disease process preceding subunit c accumulation. Mitochondrial abnormalities, which have also been reported in JNCL patients and other animal models [21-23], may result from ineffective turnover by autophagy, a lysosomal-targeted pathway [24]. Alternatively, or simultaneously, battenin deficiency may impact mitochondrial function upstream of turnover, affecting mitochondrial biogenesis and/or altered transport and processing of mitochondrial proteins.
In wild-type CbCln3 neuronal precursor cells battenin primarily co-localizes with early and late endosomes. Battenin immunostaining in homozygous CbCln3Δex7/8 neuronal precursors is significantly reduced in abundance, but mutant signal also co-localizes with endosomal markers suggesting mutant battenin protein with C-terminal epitopes is trafficked similar to wild-type protein. In other studies, CLN3/battenin protein localization has been reported to partially overlap with late endosomes and lysosomes in non-neuronal cells [7], and to lysosomes, synaptosomes [8] and endosomes [9,25] in neurons. These data jointly indicate that battenin resides in a subset of vesicular compartments linking multiple membrane trafficking pathways, perhaps functioning in vesicular transport and/or fusion. Endocytic and lysosomal-targeted pathways, including mitochondrial autophagy, are specifically implicated in this study.
Conclusions
The membrane trafficking and mitochondrial deficits uncovered in homozygous CbCln3Δex7/8 cells are likely to particularly impact neuronal function. Neurotransmission heavily relies on membrane vesicle transport, and a high-energy metabolism may further sensitize neurons to the loss of battenin activity. Thus, our panel of wild-type, heterozygous, and homozygous CbCln3Δex7/8 cerebellar cells provide an ideal model system to further elucidate battenin function and JNCL pathogenesis.
Methods
Antibody and cell staining reagents
Nestin (clone Rat 401, 2 μg/ml), Lamp 1 (clone 1D4B, 6 μg/ml), and Lamp 2 (clone Abl-93; 6 μg/ml) monoclonal antibodies were obtained from the Developmental Studies Hybridoma Bank, maintained by The University of Iowa, Department of Biological Sciences. Batp1 (1 μg/ml) was previously described [12]. Anti-subunit c antibody (0.7 μg/ml) was kindly provided by Dr. Kominami (Juntendo University, Tokyo, Japan). Additional antibodies were as follows: GFAP, 1:2000 (DAKO Corporation); calbindin, 1:5000 (Sigma); NeuN, 10 μg/ml (Chemicon); SV40 T antigen (Pab 101), 2 μg/ml (Santa Cruz Biotechnology); cathepsin D (C-20), 2 μg/ml (Santa Cruz Biotechnology); cytochrome c oxidase subunit IV (cox4), 1:1000 (BD Biosciences Clontech); PDI (H-160), 2 μg/ml (Santa Cruz Biotechnology); GM130, 1 μg/ml (BD Transduction Labs); α-tubulin, 1:15,000 (Sigma); grp75, 1:200 (Stressgen); early endosome antigen-1 (EEA1), 2 μg/ml (Santa Cruz Biotechnology); rab 7, 4 μg/ml (Santa Cruz Biotechnology). All fluorescent secondary antibodies were obtained from Jackson ImmunoResearch Laboratories and HRP-conjugated secondary antibodies were obtained from Amersham Biosciences. Additional cell markers were as follows: VVL-biotin, 1:2000 (Vector Laboratories), 10,000 MW dextran-FITC, 1 mg/ml and Lysotracker Red DND-99, 500 nM (Molecular Probes).
Generation, maintenance and differentiation of CbCln3 cerebellar neuronal precursor lines
Cln3Δex7/8 knock-in mice have been previously described [12]. Littermate pups from heterozygote crosses were used for primary culture establishment, by previously described methods that yield cerebellar granule neuron-enriched cultures [26]. Postnatal day 4 (P4) cerebella were dissected in Hank's Balanced Salt Solution (HBSS, Sigma), supplemented with 35 mM glucose. Tail biopsies were also collected for genomic DNA isolation and genotypic analysis. Trypsin/EDTA (10 mg/ml, Sigma) and DNase I (100 μg/ml, Sigma), suspended in HBSS, helped dissociate cerebella for primary culture plating onto 0.01% poly-ornithine (Sigma) coated 100 mm dishes. Primary cultures from individual cerebella were cultured overnight at 37°C, 5% CO2, in granule neuron growth media (Dulbecco's Modified Eagle Medium [DMEM, Gibco BRL #11995-065], 10% fetal bovine serum [Sigma #F-2442], supplemented with 24 mM KCl). Infection was performed the following day with defective retrovirus transducing the temperature-sensitive tsA58/U19 large T antigen and a selection neomycin-resistance cassette [27], as previously described [28]. Following infection, cultures were shifted to the tsA58/U19 permissive growth temperature of 33°C and selection was in the same growth media as above, with 200 μg/ml G418. Conditionally immortalized colonies were isolated 3–9 weeks post-infection and expanded for frozen stocks and further sub-clone isolation. Multiple clonal lines were obtained for each genotype and all phenotypes were confirmed in at least 2 independent CbCln3 cell lines. CbCln3 cell lines were maintained on 0.01% poly-ornithine coated dishes at 30–90% confluency, in 33°C and 5% CO2 atmosphere. Passage number was recorded (up to ~20 passages), but had no apparent impact on phenotype. Neuronal differentiation was as previously described [29] with the following cocktail: 10 ng/ml α-FGF, 250 μM IBMX, 200 nM TPA, 50 μM forskolin, 5 μM dopamine (Sigma).
Genotyping and RT-PCR
Genomic DNA was extracted from tail biopsies and cell pellets as described (Cotman et al., 2002). Cln3Δex7/8 knock-in allele PCR genotyping was with wild-type primers, WtF (5'-CAGCATCTCCTCAGGGCTA-3') and WtR (5'-CCAACATAGAAAGTAGGGTGTGC-3') to yield a ~250 bp band and knock-in primers, 552F (5'-GAGCTTTGTTCTGGTTGCCTTC-3') and Ex9RA (5'-GCAGTCTCTGCCTCGTTTTCT-3') to yield a ~500 bp band. PCR cycling conditions were 95°C for 30 seconds, 58°C for 30 seconds, and 72°C for 35 seconds, repeated for 34 cycles. Total RNA isolation and Cln3 RT-PCR primers and procedures have been previously described [12].
Subunit c accumulation assay
Cells were seeded into 4-well chamber-slides (Falcon) at a density of 5 × 104 cells per well for microscopy studies, or into 100 mm dishes (Falcon) at a density of 5 × 105 cells per dish for protein extraction. Cells were typically >95% confluent one day post-plating, and the following day was considered 1-day post-confluency. At the indicated times, cells were either fixed with 4% formaldehyde in phosphate buffered saline (PBS), pH 7.4, for 20 minutes and processed for autofluorescence/subunit c immunostaining, or cell pellets were collected for total protein extraction.
Alternatively, cells were fixed with 2.5% glutaraldehyde/2% paraformaldehyde in 0.1 M cacodylate buffer, pH 7.4 for 1 hour and subsequently post-fixed and processed for TEM analysis as described [12]. In confocal microscopy studies, autofluorescent signal was observed over multiple wavelengths. For co-staining, settings were reduced such that autofluorescent signal did not contribute to antibody label signal.
Immunostaining and Immunoblot analysis
For immunostaining, cells were seeded at a density of 3–5 × 104 cells per well in 4-well chamber-slides and grown overnight at 33°C, unless indicated otherwise. Fixation was with ice-cold 4% formaldehyde in PBS, pH 7.4, for 20 minutes, or with ice-cold methanol/acetone (1:1) for 10 minutes at -20°C followed by air-drying (antibody-dependent). Cells were washed with PBS at least 2 times, 5 minutes per wash, between each of the following steps of the staining procedure: 0.1 M glycine in PBS for 5 minutes, 0.05% or 0.1% (antibody-dependent) Triton X-100 (Fisher Scientific) in PBS for 5 minutes, 2% bovine serum albumin (BSA) in PBS for 30 minutes, primary antibody diluted in 2% BSA/PBS for 90 minutes, secondary antibody diluted in 2% BSA/PBS for 60 minutes. All incubations were carried out at room temperature. Following staining procedures, slides were coverslipped with Vectashield mounting medium (Vector Laboratories) and analyzed on a BioRad Radiance 2100 confocal microscope (Biorad), with identical exposure settings for wild-type and mutant like images. All comparisons of wild-type and mutant staining were performed in Adobe Photoshop with identical brightness and contrast adjustments.
Total proteins were isolated from cell pellets by extraction with ice-cold 20 mM Tris, pH 7.4, 1% Triton X-100 (membrane-research grade, Roche), plus protease inhibitors (Complete mini tablet, 0.7 μg/ml pepstatin A, 2 μg/ml aprotinin, 5 μg/ml leupeptin [Roche]). Following homogenization through a 25-gauge needle (~10 passes), extracts were centrifuged at 1000 × g for 10 minutes, at 4°C, to remove debris. Typically, 20–40 μg of protein (determined by Bio-rad Dc Protein Assay) was separated by SDS-PAGE, for subsequent immunoblotting, as described [12]. 16.5% tris-tricine SDS-PAGE gels were used for subunit c immunoblotting experiments.
Cathepsin D activity assay
100 mm tissue culture dishes, which were approximately 80–90% confluent, were washed briefly with ice-cold PBS, and total protein extracts were isolated by scraping cells into 10 mM Tris, pH 7.4, 0.1% Triton X-100 followed by incubation on ice, for 20 minutes. The insoluble material was centrifuged at 14,000 g, the supernatant was isolated, and protein concentration was determined using the Bio-rad Dc Protein Assay. 50–70 μg of total protein extract were used to measure cathepsin D activity using the Fluorogenic Innozyme™ Cathepsin D Immunocapture Activity Assay Kit (EMD Biosciences) according to the manufacturer's recommendations. Relative fluorescence was measured using an Analyst AD plate reader (Molecular Devices) with the following filters and settings: excitation filter, 360-35; emission filter, 400-20, Flash lamp with 100 readings/well, 100 ms between readings, and 100,000 μs integration time.
Lysosomal staining and endocytic uptake
Cells were seeded at a density of 3–5 × 104 cells per well in 4-well chamber-slides and grown overnight at 33°C. Growth media was exchanged for fresh, pre-warmed growth media containing 500 nM Lysotracker or 1 mg/ml dextran-FITC, and cells were incubated at 33°C for 45 minutes or 15 minutes, respectively. Following labeling, cells were immediately placed on ice and washed for 10 minutes in ice-cold dye-free media, and fixed with 4% formaldehyde in PBS, for 20 minutes on ice. Chambers were removed and slides were coverslipped with Vectashield mounting media for confocal microscopy analysis, as described above.
Morphometric analysis of mitochondria
TEM photomicrographs (10,000 × – 40,000 × magnification) were taken from random grid fields. For length measurements, the longest side of each mitochondria was measured in centimeters, and along the length of the mitochondria, width measurements were taken every 2.5–4 mm (dependent on the magnification of the micrograph image). Following measurement, all numbers were normalized to reflect one magnification and data was analyzed using Microsoft Excel software. To ascertain unmagnified mitochondrial size, final measurement data, in centimeters, was converted to nanometers according to scale bar representation.
ATP measurement
ATP was measured by using the CellTiter-GLO® Luminescent Cell Viability kit (Promega), according to the manufacturer's recommendations. Briefly, cells were plated in a black opaque-walled 96 well plate (Packard Bioscience) at a density of 20000/well and incubated at 33°C overnight. The following day, CellTiter-GLO® Reagent was added to each well and cell lysis was induced by mixing 2 minutes. An ATP standard curve was prepared in the same plate. Before recording luminescence with a microplate luminometer (MicroLumat Plus LB 96V, Berthold Techonologies), the plate was dark adapted for 10 minutes at room temperature to stabilize the luminescence signal.
Hydrogen peroxide treatment assay
Cells were plated at a density of 10,000 cells/well in 96-well plates and incubated at 33°C overnight. The following day, fresh media containing varying concentrations of hydrogen peroxide was dispersed to each well. Cells were incubated in the presence of hydrogen peroxide for 24 hours, at 33°C, and viability was measured using the CellTiter-96® AQueous Non-Radioactive Cell Proliferation Assay (Promega), according to the manufacturer's specifications.
Authors' contributions
EF participated in establishment and characterization of cell lines and performed ATP determinations. PW participated in mitochondrial analysis and immunocytochemistry. JE, TL-N, AMT, and HG participated in genotypic and additional phenotypic analysis of cell lines. DR and EC generated virus-conditioned medium for conditional immortalization of cells. MEM co-conceived of the study and assisted on drafting of the manuscript. SLC co-conceived of the study, participated in establishment and phenotypic analysis of cell lines, and drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Dr. E. Kominami for antibody to subunit c of mitochondrial ATPase, L. Trakimas and M. Ericsson of the Harvard Medical School Electron Microscope Facility, Dr. David Sulzer for assistance with TEM analysis, Dr. Sylvie Breton for assistance with endocytosis experiments, and Dr. Andre Bernards and James Follen for assistance with the cathepsin D activity assay. This work was supported by NIH/NINDS grant # NS 33648. Dr. S.L. Cotman received fellowship funding from the Batten Disease Support and Research Association (BDSRA).
==== Refs
Rider JA Rider DL Batten disease: past, present, and future Am J Med Genet Suppl 1988 5 21 26 3146319
International Batten Disease Consortium Isolation of a novel gene underlying Batten disease, CLN3. The International Batten Disease Consortium Cell 1995 82 949 957 7553855 10.1016/0092-8674(95)90274-0
Jolly RD Martinus RD Palmer DN Sheep and other animals with ceroid-lipofuscinoses: their relevance to Batten disease Am J Med Genet 1992 42 609 614 1535180
Palmer DN Fearnley IM Walker JE Hall NA Lake BD Wolfe LS Haltia M Martinus RD Jolly RD Mitochondrial ATP synthase subunit c storage in the ceroid-lipofuscinoses (Batten disease) Am J Med Genet 1992 42 561 567 1535179
Wisniewski KE Rapin I Heaney-Kieras J Clinico-pathological variability in the childhood neuronal ceroid-lipofuscinoses and new observations on glycoprotein abnormalities American Journal of Medical Genetics - Supplement 1988 5 27 46 3146325
Mao Q Foster BJ Xia H Davidson BL Membrane topology of CLN3, the protein underlying Batten disease FEBS Letters 2003 541 40 46 12706816 10.1016/S0014-5793(03)00284-9
Ezaki J Takeda-Ezaki M Koike M Ohsawa Y Taka H Mineki R Murayama K Uchiyama Y Ueno T Kominami E Characterization of Cln3p, the gene product responsible for juvenile neuronal ceroid lipofuscinosis, as a lysosomal integral membrane glycoprotein Journal of Neurochemistry 2003 87 1296 1308 14622109 10.1046/j.1471-4159.2003.02132.x
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 Human Molecular Genetics 2001 10 2123 2131 11590129 10.1093/hmg/10.19.2123
Kyttala A Ihrke G Vesa J Schell MJ Luzio JP Two motifs target Batten disease protein CLN3 to lysosomes in transfectedd nonneuronal and neuronal cells Molecular Biology of the Cell 2004 15 1313 1323 14699076 10.1091/mbc.E03-02-0120
Pearce DA Ferea T Nosel SA Das B Sherman F Action of BTN1, the yeast orthologue of the gene mutated in Batten disease Nat Genet 1999 22 55 58 10319861 10.1038/8861
Kim Y Ramirez-Montealegre D Pearce DA A role in vacuolar arginine transport for yeast Btn1p and for human CLN3, the protein defective in Batten disease Proceedings of the National Academy of Sciences of the United States of America 2003 100 15458 15462 14660799 10.1073/pnas.2136651100
Cotman SL Vrbanac V Lebel LA Lee RL Johnson KA Donahue LR Teed AM Antonellis K Bronson RT Lerner TJ MacDonald ME Cln3∆ex7/8 knock-in mice with the common JNCL mutation exhibit progressive neurologic disease that begins before birth Human Molecular Genetics 2002 11 2709 2721 12374761 10.1093/hmg/11.22.2709
Kominami AE What are the requirements for lysosomal degradation of subunit c of mitochondrial ATPase? IUBMB Life 2002 54 89 90 12440525
Partanen S Storch S Loffler HG Hasilik A Tyynela J Braulke T A replacement of the active-site aspartic acid residue 293 in mouse cathepsin D affects its intracellular stability, processing and transport in HEK-293 cells Biochemical Journal 2003 369 55 62 12350228 10.1042/BJ20021226
Holopainen JM Saarikoski J Kinnunen PK Jarvela I Elevated lysosomal pH in neuronal ceroid lipofuscinoses (NCLs) European Journal of Biochemistry 2001 268 5851 5856 11722572 10.1046/j.0014-2956.2001.02530.x
Nishimura Y Kato K Furuno K Himeno M Biosynthesis and processing of lysosomal cathepsin D in primary cultures of rat hepatocytes Biological & Pharmaceutical Bulletin 1995 18 825 828 7550114
Golabek AA Kida E Walus M Kaczmarski W Michalewski M Wisniewski KE CLN3 protein regulates lysosomal pH and alters intracellular processing of Alzheimer's amyloid-beta protein precursor and cathepsin D in human cells Mol Genet Metab 2000 70 203 213 10924275 10.1006/mgme.2000.3006
Dawson G Glaser PT Abnormal cathepsin B activity in Batten disease Am J Med Genet Suppl 1988 5 209 220 3146318
Wisniewski KE Kida E Walus M Wujek P Kaczmarski W Golabek AA Tripeptidyl-peptidase I in neuronal ceroid lipofuscinoses and other lysosomal storage disorders European Journal of Paediatric Neurology 2001 5 Suppl A 73 79 11589013
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]. [erratum appears in Neurobiol Dis 2000 Apr;7(2):127] Neurobiol Dis 1999 6 321 334 10527801 10.1006/nbdi.1999.0267
March PA Wurzelmann S Walkley SU Morphological alterations in neocortical and cerebellar GABAergic neurons in a canine model of juvenile Batten disease Am J Med Genet 1995 57 204 212 7668331
Das AM Jolly RD Kohlschutter A Anomalies of mitochondrial ATP synthase regulation in four different types of neuronal ceroid lipofuscinosis Mol Genet Metab 1999 66 349 355 10191128 10.1006/mgme.1999.2811
Das AM von Harlem R Feist M Lucke T Kohlschutter A Altered levels of high-energy phosphate compounds in fibroblasts from different forms of neuronal ceroid lipofuscinoses: further evidence for mitochondrial involvement European Journal of Paediatric Neurology 2001 5 Suppl A 143 146 11588986
Klionsky DJ Emr SD Autophagy as a regulated pathway of cellular degradation Science 2000 290 1717 1721 11099404 10.1126/science.290.5497.1717
Persaud-Sawin DA McNamara JOII Vandongen RS Boustany R A galactosylceramide binding domain is involved in trafficking of CLN3 from golgi to rafts via recycling endosomes. Pediatric Research 2004 56 449 463 15240864
D'Mello SR Galli C Ciotti T Calissano P Induction of apoptosis in cerebellar granule neurons by low potassium: inhibition of death by insulin-like growth factor I and cAMP Proceedings of the National Academy of Sciences of the United States of America 1993 90 10989 10993 8248201
Jat PS Cepko CL Mulligan RC Sharp PA Recombinant retroviruses encoding simian virus 40 large T antigen and polyomavirus large and middle T antigens Molecular & Cellular Biology 1986 6 1204 1217 3023876
Cattaneo E Conti L Generation and characterization of embryonic striatal conditionally immortalized ST14A cells Journal of Neuroscience Research 1998 53 223 234 9671979 10.1002/(SICI)1097-4547(19980715)53:2<223::AID-JNR11>3.0.CO;2-7
Trettel F Rigamonti D Hilditch-Maguire P Wheeler VC Sharp AH Persichetti F Cattaneo E MacDonald ME Dominant phenotypes produced by the HD mutation in STHdh(Q111) striatal cells Hum Mol Genet 2000 9 2799 2809 11092756 10.1093/hmg/9.19.2799
| 15588329 | PMC539297 | CC BY | 2021-01-04 16:39:08 | no | BMC Neurosci. 2004 Dec 10; 5:57 | utf-8 | BMC Neurosci | 2,004 | 10.1186/1471-2202-5-57 | oa_comm |
==== Front
Environ HealthEnvironmental Health1476-069XBioMed Central London 1476-069X-3-151558827610.1186/1476-069X-3-15ReviewOn-call work and health: a review Nicol Anne-Marie [email protected] Jackie S [email protected] Centre for Health and Environment Research, University of British Columbia, 2206 East Mall, Vancouver, BC, V6T 1Z3, Canada2 School of Cultural and Innovation Studies, University of East London, 4–6 University Way, London, E16 2RD, UK2004 8 12 2004 3 15 15 8 7 2004 8 12 2004 Copyright © 2004 Nicol and Botterill; licensee BioMed Central Ltd.2004Nicol and Botterill; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Many professions in the fields of engineering, aviation and medicine employ this form of scheduling. However, on-call work has received significantly less research attention than other work patterns such as shift work and overtime hours. This paper reviews the current body of peer-reviewed, published research conducted on the health effects of on-call work The health effects studies done in the area of on-call work are limited to mental health, job stress, sleep disturbances and personal safety. The reviewed research suggests that on-call work scheduling can pose a risk to health, although there are critical gaps in the literature.
==== Body
Background
The question of whether work hours and schedules affect people's health has been reviewed for a range of work patterns including shift work and overtime. Research in these areas indicates that shift work, and in particular night work can interrupt sleep patterns [1], aggravate existing medical conditions and increase the risk of cardiovascular, gastrointestinal, and reproductive dysfunctions [2-4].
However, the health effects of on-call work, where workers are called to work either between regular hours or during set on-call periods, has not merited as much attention. This form of work scheduling occurs in a variety of diverse occupations, for example medical technologists, doctors, ship engineers, utility workers, electrical technicians, tug boat pilots, midwives, information technologists, media personnel and junior airline pilots. For many of these professions being on-call is not an option, but rather a component of the job. This form of scheduling is often used to provide 24 hour coverage, 7 days a week, for facilities such as hospitals and laboratories, where emergencies require personnel to immediately deal with critical situations and where the volume of evening and weekend work does not necessitate full shift coverage. Having employees on-call, even if they are being paid a stipend for their call time, is often seen as less expensive for employers than providing full shift coverage during off-peak hours [5].
While on-call work scheduling may be less expensive, it is not without human costs. On-call employees must plan their lives and the lives of their families around a call schedule. This often means limiting behaviours and obliges employees to restrict their on-call time to activities that would not interfere with their ability to work. The unpredictability of the call scheduling can also generate a great deal of stress, as home life is interrupted and workers are required to "change hats" to shift to their professional roles at any time during call. These limitations and interferences present unique challenges for on-call workers that are not encountered by those working set schedules or even people with rotating shifts. It is thus not surprising that researchers have found that on-call work patterns can have a major influence on employees' lifestyles and their interactions with family members and friends [6]. However, in addition to the impact on lifestyle and relationships, on-call work patterns may impact the health of employees.
Within the limited literature that has explored on-call work, there exists some pertinent findings concerning the impact of on-call for an employee's physical and psychological health, and social relationships, which this review seeks to bring together. Specific attention has been devoted to the areas of stress, sleep, mental health and personal safety.
Types of on-call work
The implementation of on-call schedules varies. For many occupations, workers leave their place of employment and are placed "on-call" on evenings and weekends, which means they can be called back to work during these periods. For many professions this form of scheduling is a normal component of the occupation, for example, marine pilots can spend up to 60% of their working time on-call. However, for a limited number of occupations such as airline pilots, on-call hours are reduced with seniority. Generally, but not always, employees are compensated monetarily for the period of call, usually with a stipend which is less than their hourly rate. When on-call employees are usually expected to restrict their use of alcohol and limit distance or travel time from the work-site. The on-call experience of these workers includes aspects of interruption, either of sleep or family or social life, and often includes an element of uncertainty as to the time of call or the occurrence of the call.
Other forms of on-call include work done by junior doctors during their medical training. Medical residents spend periods of time "on-call" at a hospital, where space may be provided for them to sleep. This form of on-call work is distinct because workers remain at work to undertake their call duty. During these periods, residents often put in 30–36 hour shifts with little to no sleep [7], resulting in a combination that is both a night shift and an overtime shift. Because of the intensive demands placed on medical residents during their apprenticeship, this group has received a fair amount of research attention. This has been particularly so in the 1990s as the rigors of this period in junior doctors' training has come under much scrutiny both in the UK and in the US. New working regulations have been introduced in an attempt to deal with what is considered, by many, to be harsh and unacceptable working conditions. The debate over and outcome of these interventions continues [8-10].
This review focuses on the health effects of on-call work in which an employee spends a period of time on-call outside of their workplace and/or their regular working hours. The research on medical residents has been excluded because the medical resident experience is distinctly different from that of other professions where on-call is utilized. However, research on medical residents is used to illustrate findings from other work-related areas when appropriate.
Methods
This review explores the published literature referring to on-call work patterns and health. For the purpose of this review, the on-call period may be formal (e.g. a person is designated as being on-call for the weekend or overnight) or informal (emergency call back during a crisis). Search terms for this review included "on-call" and "work schedule tolerance". The terms "stand-by" and "night call" are used by some professions to describe on-call type work, and were also used as search parameters. This literature review was undertaken on journal articles included in databases up to December 2000. Database searches were performed on the following electronic sources:
1) OVID Databases: including Medline (1966–2000) and Current Contents (1996–2000).
2) Canadian Centre for Occupational Health and Safety Database
3) Cambridge Abstracts (Environmental Science and Pollution Management) 1981–2000.
4) PsycInfo (1989–2000).
5) Web of Science: including Science Citation Index and Social Science Citation Index (1989–2000).
A manual review of the references generated from the computer-search was also done. Articles were excluded from the review if they were not original research, were not written in English or focused on medical residents experiences with on-call work.
Two reviewers read through each of the eligible research papers independently.
Results
In total, 24 papers met the search criteria. Eight (8) were excluded as they focused on the impact of on-call work patterns on patient's health and not on the health of workers. The remaining sixteen studies were used for this review. The results are divided into four health-related sections; 1) Stress, 2) Sleep, 3) Mental Health and 4) Personal Safety.
On-Call Work and Stress
Of the five studies pertaining to on-call and stress uncovered in this review, all focus on the General Practitioners (GPs) as their subject. In these studies, the relationship between on-call work and stress was measured through self-report and perceived stress.
Three of the studies were part of a major UK study carried out from 1989 to 1998 [11-13]. In the early 1990s the British health care system experienced considerable financial and administrative restructuring. This large study was conducted at different points in time to determine GP's satisfaction with the changes in their workplace. GPs were randomly selected throughout Britain in 1987, 1990 and 1998 to fill out postal questionnaires. The studies yielded sample sizes of 1817, 917, and 999 respectively, representing rather low response rates of 48%, 67% and 47%. However, the authors' assessment of all three samples found that they tended to be fairly representative of the larger population of GPs in the country [13].
In the first two studies, GPs ranked working on-call at night as one of the top two most stressful aspects of their work situation [11,12]. However results from the third study in 1998 revealed that night call was no longer a major source of stress, dropping to 12th in a ranking of 14 major stressors. The authors believe this reduction in the level of stress from on-call work could be explained by the introduction of GP co-operatives in the mid 1990s for the management of out-of-hours calls. This cooperative system allowed GPs to either do their own calls or share them with a cooperative formed by 10 or more doctors. The cooperatives gave GPs greater flexibility for how and where they saw their patients and how they implemented 24-hour care and appear to have successfully reduced the stress of night visits for GPs. Indeed, night visit stress went from being one of two top stressors for GPs in 1987 and 1990 to being one of the least stressful issues by 1998. The authors also posit that this "success may also explain the reported reduction by 1998 in stress attributable to disturbance of home/family life" [[13] pg. 370].
The fourth study also dealt with the changes in the British health care system, in particular the introduction of partial shifts to decrease long on-call periods [14]. A small sample of GPs'(n = 14 and 12) were surveyed about their stress levels before and after the new system was in place. Doctors' stress levels were significantly reduced, particularly in relation to their mental well-being and their job satisfaction.
The fifth study on GPs and stress was a qualitative analysis of 25 GPs and their spouses in Manchester [15]. This research found that for male GPs, the uncertainty of being on-call caused them to be unhappy. Some doctors spoke quite frankly about how night calls could "perturb family life and wreck personal intimacy" [[15] p. 158). The uncertainty of their on-call commitments also contributed to the male GPs' unhappiness. Female GPs were stressed by factors other than on-call, including time pressure, role conflict and work overload. They were also concerned about how their work schedule decreased the amount of time they spent with their children. These marked differences between how male and female doctors experience the stress of on-call work signals the importance of examining gender as a variable in this research.
Other studies have revealed that the amount of time spent on-call varies between male and female doctors, but no clear pattern has emerged [16,17]. It has been hypothesized that female doctors who work reduced on-call hours do so because of the dual role they must play as both worker and care-giver [17,18].
Research conducted in other professions support the idea that work patterns, particularly night shifts, can increase stress in workers and have a negative impact on family life. Working late afternoon and evening shifts has been related to increased stress for both workers and their families [2]. Variable shifts have been shown to cause more stress than regular shifts [19] and working more than 50 hours per week is associated with increased job stress [20]. Many on-call workers regularly experience variation in their work patterns, as well as being expected to work at night, and undertake greater than normal hours when called in.
On-Call Work and Sleep
Besides stress, the interruption of sleep is another major component of on-call work, particularly for those who work nights on-call and in professions that deal with emergencies that occur at all hours. Three studies have dealt specifically with the sleeping patterns and problems experienced by train and ship engineers and transplant coordinators, all of whom regularly work on-call.
The first study researched the on-call sleep patterns of 198 train engineers using prospective activity logs over a 14-day period in the United States [21]. It was determined that those working on-call had greater difficulty falling asleep and staying asleep while on-call versus when they were not on-call. Train engineers working on-call also had a greater number of days where there was less than 24 hours between the on-set of their work shifts. These engineers reported more sleep-related problems that those with at least 24 hours between the on-set of their shifts. The researchers also explored how sleeping was impacted when it was undertaken in different locations. They found that train engineers sleep varied when at home versus "away". (Engineers can finish a shift away from home, and have "away" terminals where they can sleep.) The researchers compared the amount and quality of sleep engineers had while both "at home" and "away" and found that engineers on-call slept less at home than they did "away". The authors attribute the difference to the presence of family and social obligations in the home that conflicted with the workers' ability to sleep while working on-call. However, the authors note that the response rate of this study was low, only 25% of the sampled population of approximately 800. The authors caution readers to remain critical of their findings, because their sample may be biased towards those who generally have difficulty sleeping. An analysis of the final study group did find that the responding sample reflected the age and gender distributions of the larger population, factors that the authors suggest indicate robustness even with the low response rate.
The second study of on-call and sleep explored the sleeping patterns of 53 predominantly female organ transplant coordinators in the UK, using a postal questionnaire [22]. This research determined that not only was sleep affected when people worked on-call (51% occasionally had difficulty and 6% frequently had difficulty falling asleep) but that the effects carried over to time off call as well. Sixty-eight percent of the sample reported that the time they spent on-call negatively influenced their off-call lives. Workers pointed out that after being on-call they often had to spend additional time catching up on sleep. They also complained that on-call work left them too tired to undertake social and home activities. But although the workers complained about being fatigued at home, this was not correlated with days absent from work. The authors suggest that this finding may be the result of transplant coordinators "guilt" around placing an extra burden on a co-worker if they were absent. Another possible explanation was the overall satisfaction of the type of work being done by the coordinators, a factor which may decrease their willingness to take time off.
The third study, conducted on a small sample (n = 5) of ship engineers in Sweden, measured sleep during on-call periods using electroencephalogram (EEG) and electrocardiogram (ECG) recordings and subjective ratings. [23]. This research found, like the others, that the sleep quality and quantity of the ship engineers was affected by the interruptions of being on-call. In their subjective assessments, the engineers reported being more drowsy during the day after being on-call, a finding similar to that of the transplant coordinators. But, the authors also found that the apprehension associated with the possibility of being awakened for call duty also negatively impacted sleep. On-call sleep registered less slow wave sleep (SWS) and rapid eye movement (REM) and a higher heart rate than when workers were testing during their normal sleep. Many of these conditions occurred prior to being awakened for call duty. Earlier research by the same authors examined the sleep patterns of Swedish merchant marines at sea. This population also found it difficult to fall asleep on nights when they were on watch. The anticipation of alarms that would wake them up was seen as an obstacle that prevented workers from relaxing enough to allow for normal sleep patterns to develop [24].
The impacts of sleep loss on job performance remain unclear and controversial. For example, research on the cognitive performance in sleep deprived medical residents has produced mixed results [25-27]. However, research on anaesthetists found that 86% reported fatigue related errors [28]. Job performance and fatigue have also been studied in relation to age, a factor not explored in the on-call studies. Significant changes were found between younger and older shift workers, with younger workers better able to maintain performance across day and night shifts and older shift workers prone to more sleep disruption [29].
Work-related fatigue has been related to an increase in car accidents. A review of traffic accidents determined that falling asleep while driving accounted for a major proportion of accidents while driving under monotonous conditions [30]. This finding has been corroborated with research done on medical residents working long night shifts. Seventy-five percent of accidents incurred by a population of emergency medicine residents happened after working a night shift [31]. In this study, the number of motor vehicle accidents and near misses was positively correlated to the number of nights worked per month. A similar study done on paediatric residents indicated that residents fell asleep at the wheel significantly more than other professionals, with 90% of these events occurring after a night on-call [32].
On Call Work and Mental Health
Six studies were found that examined the impact of on-call work schedules on mental health. All of these studies used self-reported questionnaires and/or mood diaries. Five studies were conducted on GPs in the UK and one examined gas and electrical employees in France.
Two surveys were conducted by Chambers et al. [33,34] on GPs in Staffordshire, UK. The first survey, conducted in 1994 (n = 704), was designed to research the factors predictive of anxiety and depression in GPs [33]. The study determined that working one or more nights on-call per week was significantly predictive of anxiety. Other factors predictive of anxiety were depression and three or more weekdays feeling exhausted or stressed. Males and females showed no significant differences in anxiety or depression determinants.
The second survey conducted by Chambers et al in 1996 (n = 620) employed the Hospital Anxiety and Depression scale to assess the mental health of GPs [34]. It was determined that both anxiety and depression were associated with the amount of on-call duties undertaken. Findings revealed that both anxiety and depression increased with the frequency of time spent on-call per month. Again, the results were the same for both male and female GPs, and the authors conclude that GPs' mental ill health is associated with workload, of which on-call is a major factor.
A third survey done on GPs in Leeds in 1993 was designed to determine the psychological symptoms and sources of stress among 268 GPs [35]. This survey used the UK General Health Questionnaire as well as qualitative questions regarding mental health and workload. Problems with physical and mental health were significantly associated with several aspects of workload, including the amount of time spent on-call per month. The study also found that those GPs who spent more time on-call each month were more likely to feel their work affected their physical health. Males and females reported differences in the sources of their stress, with females showing greater job satisfaction than males. The authors suggest that this finding may be due to the fact that, for this study population, female doctors worked fewer hours and spent significantly fewer nights on-call [34].
The fourth study in this area surveyed mental health and job stress on 414 GPs in England in 1992 [36]. This research determined that interruptions, a category which included taking night calls, remaining alert on-call, 24 hour patient responsibility and telephone interruption of family life, was a predictive factor for decreased mental health, depression and somatic anxiety. These factors were similar for men and women, although their contribution to each condition varied by gender.
A pilot study of 44 male and female volunteer GPs using cognitive behavioural diaries assessed self-reported emotional states recorded in conjunction with hourly activities over 2 days [37]. Doctors' moods were significantly lowered when on-call as compared to off-call. Doctors on-call also had significantly increased tension and frustration. The main reported cause of dissatisfaction was the uncertain nature of the doctors working hours [37].
The sixth study examined male gas and electrical employees working in France [38]. Employees who worked on-call (n = 145) were assessed for health status and psychological problems and were compared to those not working on-call (n = 195). Workers were also questioned about the impact of their job on their family life. Although no particular mental or health disorder was found to be more frequent in the on-call group, the psychological equilibrium of the on-call workers was significantly worse than the comparison group. On-call workers also reported significantly worse global-well being and indicated significantly higher levels of social disturbance. On-call workers reported that their family and social life were acutely disturbed and they were significantly less likely to be involved in clubs or take on outside responsibilities.
The research conducted on GPs in the UK supports a negative role of on-call work related to mental health. However, the results from the gas and electrical workers do not reflect the same findings from the research on GPs. This may be the result of either a difference in study methodology or a difference that is profession-specific. On-call gas and electrical workers did experience psychological disruption and the lack of significant diagnostic findings may be a function of other factors, such as self-selection, in this profession, where those most affected opt out early on. The on-call gas and electrical workers experience of family and social life disruption does mirror the experiences of doctors and transplant coordinators as discussed previously [13,15,22].
On-call Work and Personal Security
Working on-call often necessitates leaving home alone, at night, to attend work, conditions that can jeopardize personal safety. Unfortunately, there is only sparse data regarding this issue. A study done in the north west of England, in a hospital where the on-call sleeping quarters were separate from the hospital found that 40% of anaesthetists feared for their safety while walking through hospital grounds at night [39].
In medical professions, patients can also present a danger to those working on-call. Doctors have cited fear of violence from night call visits as a significant stressor [11]. A study of 327 nurses in remote areas who worked on-call found increased incidences of violent acts perpetrated by patients, particularly in smaller communities [40]. This study found that working on-call increased the number of incidents ranging from verbal abuse to property crime and physical assault compared to working regular shifts.
This issue has only been peripherally studied and further attention needs to be given to personal safety, particularly when being called in at night.
Discussion
What emerges from this review is the limited research that has been done in the area of on-call work. Preliminary work done in the areas of stress and mental health suggests that on-call work may play a role in increasing stress and decreasing mental well-being. The three studies that examined sleep indicate that on-call work does decrease the quality and quantity of sleep for workers and can leave people feeling fatigued for periods after their on-call work.
The current body of literature on the health effects of on-call work is limited in part due to the narrow range of professions studied. The majority of research done to date has been on general practitioners. It is reasonable to assume that the effects of on-call will vary across occupations, given the host of other factors that can influence occupational health. However, the degree to which this variation exists might only be determined by examining a wider occupational base. The need to undertake more on-call research across a greater variety of occupational groups is suggested given that this form of work scheduling touches many occupations, and given that on-call work is estimated to continue to increase in many sectors in the future [6].
There is also an obvious lack of research focusing on the impact of on-call shifts on psychosocial factors. Given the very disruptive and limiting nature of on-call schedules, it would not be surprising that workers' family and social life suffer due to this type of scheduling. The results of the research addressing gender (discussed above) do suggest, albeit indirectly, that such social and familial impacts may be significant. However, without more research, it is not possible to determine the magnitude of these effects, nor the relative importance compared to other factors such as physiological responses.
More rigorous methodological designs are needed for future research in the area of on-call work and health. The current research is predominantly cross-sectional in nature, a factor that makes it difficult to determine causality. Only two studies employed external comparison groups [21,38] and only a limited number have measured effects in workers on-call versus off-call (own-controls) [22,23]. Additionally, most of the measurement has been subjective in nature and often the operationalization of on-call work is not clear. In the GP studies, on-call is generally measured as the "number of nights spent on-call" either per week or per month. Some attempt is made to measure the amount of sleep during these periods, but there is little refinement of factors such as whether the subject were actually called in to work and for how long. Additionally, little attention has been paid to the amount of time worked or sleep obtained prior to the on-call shifts or factors such as second jobs or outside work, variables that may confound the outcomes. Other factors, such as age and personality type, that have been shown to be significant variables in other areas of work scheduling [41,42] also need to be explored. Attention also needs to be paid to the possible self-selection of workers out of on-call professions or adaptive strategies that workers may employ to cope with on-call (such as the sharing of on-call shifts). More controlled research that includes both subjective and objective measures would provide better evidence regarding the effects of on-call work.
Future research on the health effects of on-call work also needs to examine the role of gender, not only from a physiological standpoint, (e.g. reproductive issues), but also from a psychosocial perspective. Many of the articles reviewed above indicate differences in how males and females experience the stress of on-call work [11,15,17,36]. Research in other work-related areas suggests that males and females cope differently with the impact of job schedules [43-45]. While gender may be a factor that directly mediates health effects, it may also be an indirect measure of other phenomena such as the division of labour outside of the workplace. More careful research is needed to illuminate the role gender may play in the effects of on-call work.
The range of health effects studied in relation to on-call work has to date been inadequate. Health conditions such as cardiovascular disease, reproductive problems, gastrointestinal issues and overall mortality need to be explored as has been done in conjunction with work patterns such as overtime and shift work [41,45]. Factors such as personal safety and car accidents have only briefly been touched upon, and merit more attention.
Conclusions
While the results of this review are limited, initial research in this area suggest that being on-call can have negative impacts on workers' sleep patterns, mental health and personal life. Further research in this area is required to provide a clear picture of the risks of this form of work scheduling.
List of abbreviations
UK, United Kingdom
US, United States of America
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AMN designed the research project, carried out the literature search, reviewed articles and drafted the manuscript. JSB reviewed articles and edited the manuscript. Both authors approved the final manuscript.
Acknowledgements
This research was supported in part by the Occupational Health and Safety Association for Healthcare in BC (OHSAH), a non-profit agency. The authors wish to thank Dr. Kay Teschke (UBC), Rachel Notley and Carol Riviere (Health Sciences Association of BC) and David Murphy (SFU) for their assistance with this review.
==== Refs
Kuhn G Circadian Rhythm, shift work and emergency medicine Ann Emerg Med 2001 37 88 98 11145778 10.1067/mem.2001.111571
Scott AJ Shift work and health Prim Care 2000 27 1057 11072299
Nurimen T Shift work and Reproductive Health Scand J Work, Environ Health 1998 28 34 9916814
Harrington J Working Long Hours and Health BMJ 1994 308 1581 1582 8025420
Mabon J Call-back – the hidden issues Can J Med Technol 1995 57 116 117 10142998
Berger Y Standby Periods Aust Safety News 1999 63 3 10
Ozkarahan I A scheduling model for hospital residents J Med Syst 1994 18 251 265 7861103 10.1007/BF00996605
Thorpe K House Staff Supervision and Working Hours: Implications of Regulatory Change in New York State JAMA 1990 263 3177 3181 2348528 10.1001/jama.263.23.3177
Last GC Curley P Galloway JM Wilkinson A Impact of the New Deal on vascular surgical training Ann R Coll Surg Engl 1996 78 263 266 8944496
Conigliaro J Frishman WH Lazar EJ Croen L Internal medicine housestaff and attending physician perceptions of the impact of the New York State Section 405 regulations on working conditions and supervision of residents in two training programs J Gen Intern Med 1993 8 502 507 8410422
Cooper CL Rout U Faragher B Mental health, job satisfaction, and job stress among general practitioners BMJ 1989 298 366 370 2493939
Sutherland VJ Cooper CL Job Stress, Satisfaction and Mental Health among general practitioners before and after introduction of new contract BMJ 1992 304 1545 1548 1628056
Sibbald B Enzer I Cooper C Rout U Sutherland V GP job satisfaction in 1990 and 1998: lessons for the future? Fam Pract 2000 17 364 371 11021893 10.1093/fampra/17.5.364
Reid N Moss PJ The impact of the New Deal: Doctors' stress levels and their views Stress Med 1999 15 9 15 10.1002/(SICI)1099-1700(199901)15:1<9::AID-SMI780>3.0.CO;2-O
Rout U Stress among general practitioners and their spouses: a qualitative study Br J Gen Pract 1996 46 157 160 8731621
Hooper J Full-Time women general practitioners – an invaluable asset J R Coll Gen Pract 1989 39 289 291 2556544
Chambers R Campbell I Gender differences in general practitioners at work Br J Gen Pract 1996 46 291 293 8762745
St-Laurent-Gagnon T Duval R Lippe J Cote-Boileau T Women in pediatrics: the experience in Quebec CMAJ 1993 148 773 778 8439936
Gordon NP Cleary PD Parker CE Czeisler CA The prevalence and health impact of shift work Am J Public Health 1986 76 1225 1228 3752325
Spurgeon A Harrington J Cooper C Health and Safety Problems associated with long working hours a review of the current position Occup Environ Med 1997 54 367 375 9245942
Pilcher JJ Coplen MK Work/rest cycles in railroad operations: effects of shorter than 24-h shift work schedules and on-call schedules on sleep Ergonomics 2000 43 573 588 10877477 10.1080/001401300184260
Smithers F The pattern and effect of on call work in transplant coordinators in the United Kingdom Int J Nurs Stud 1995 32 469 483 8550307 10.1016/0020-7489(95)00008-L
Torsvall L Akerstedt T Disturbed sleep while being on-call: an EEG study of ships' engineers Sleep 1988 11 35 38 3363267
Torsvall L Castenfors K Akerstedt T Froberg J Sleep at Sea Ergonomics 1987 30 1335 1340
Browne BJ Vansusteren T Onsager DR Simpson D Salaymeh B Condon RE Influence of sleep-deprivation on learning among surgical house staff and medical students Surgery 1994 115 604 610 8178259
Jacques C Lynch JC Samkoff JS The Effects of Sleep Loss on Cognitive Performance of Resident Physicians J Fam Pract 1990 30 223 229 2299318
Robbins J Gottlieb F Sleep deprivation and cognitive testing in internal medicine house staff West J Med 1990 152 82 86 2309486
Gander P Merry A Millar MM Wellers J Hours of work and fatigue-related error: a survey of New Zealand Anaesthetists Anaesth Intensive Care 2000 28 178 183 10788970
Reid K Dawson D Comparing performance on simulated 12 hour shifts rotation in young and older subjects Occup Environ Med 2003 60 17 10.1136/oem.60.12.e17
Horne J Reyner L Vehicle Accidents related to sleep: a review Occup Environ Med 1999 56 289 294 10472301
Steele MT Ma OJ Watson WA Thomas HA Emergency medicine residents' shift work tolerance and preference Acad Emerg Med 2000 7 670 673 10905646
Marcus CL Loughlin GM Effect of sleep deprivation on driving safety in housestaff Sleep 1996 19 763 766 9085483
Chambers R Campbell I Anxiety and depression in general practitioners: associations with type of practice, fund-holding, gender and other personal characteristics Fam Pract 1996 13 170 173
Chambers R Belcher J Predicting mental health problems in general practitioners Occup Med (Lond) 1994 44 212 216 7949065
Appleton K House A Dowell A A survey of job satisfaction, sources of stress and psychological symptoms among general practitioners in Leeds Br J Gen Pract 1998 48 1059 1063 9624747
Rout U Cooper CL Rout JK Job stress among British general practitioners: Predictors of job dissatisfaction and mental ill-health Stress Med 1996 12 155 166 10.1002/(SICI)1099-1700(199607)12:3<155::AID-SMI687>3.3.CO;2-1
Rankin H Serieys N Elliott-Binns C Determinants of mood in general practitioners BMJ 1987 294 618 620 3103834
Imbernon E Warret G Roitg C Chastang J Goldberg M Effects of Health and Social Well-being of On-call Shifts J Occup Med 1993 35 1131 1137 8295038
Masterson GR Ashcroft GS Shah R Factors important in determining trainee Anaesthetists' quality of life Anaesthesia 1994 49 991 995 7802249
Fisher J Bradshaw J Currie BA Klotz J Searl KR Smith J Violence and remote area nursing Aust J Rural Health 1996 4 190 199 9437143
Harrington J Health Effects of Shift Work and Extended hours of Work Occup Environ Med 2001 58 68 72 10.1136/oem.58.1.68
Walters V Lenton R French S Eyles J Mayr J Newbold B Paid work, unpaid work and social support: a study of the health of male and female nurses Soc Sci Med 1996 43 627 636 10.1016/S0277-9536(96)00067-6
Oginska H Pokorski J Oginski A Gender, ageing, and shift work intolerance Ergonomics 1993 36 161 168 8440214
Gross E Gender differences in physician stress: why the discrepant findings? Women Health 1997 26 1 14 9501398
Nylen L Voss M Floderus B Mortality among women and men relative to unemployment, part time work, overtime work and extra work: a study based on data from the Swedish twin registry Occup Environ Med 2001 58 52 57 11119635 10.1136/oem.58.1.52
| 15588276 | PMC539298 | CC BY | 2021-01-04 16:36:31 | no | Environ Health. 2004 Dec 8; 3:15 | utf-8 | Environ Health | 2,004 | 10.1186/1476-069X-3-15 | oa_comm |
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central 1465-9921-5-251557595610.1186/1465-9921-5-25ResearchValue of supplemental interventions to enhance the effectiveness of physical exercise during respiratory rehabilitation in COPD patients. A Systematic Review Puhan Milo A [email protected]ünemann Holger J [email protected] Martin [email protected] Lucas M [email protected] University of Zurich, Horten Centre, Switzerland2 University at Buffalo, Departments of Medicine and of Social & Preventive Medicine, New York, USA3 McMaster University, Department of Clinical Epidemiology and Biostatistics, Hamilton, Ontario, Canada4 Klinik Barmelweid, Department of Respiratory Medicine, Barmelweid, Switzerland5 University of Berne, Department of Social and Preventive Medicine, Berne, Switzerland2004 2 12 2004 5 1 25 25 24 7 2004 2 12 2004 Copyright © 2004 Puhan et al; licensee BioMed Central Ltd.2004Puhan et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
There is a controversy about the additional benefit of various supplemental interventions used in clinical practice to further enhance the effectiveness of respiratory rehabilitation in patients with Chronic obstructive pulmonary disease (COPD). The aim of this research was to assess randomised controlled trials (RCTs) testing the additional benefit of supplemental interventions during respiratory rehabilitation in COPD patients.
Methods
Systematic review with literature searches in six electronic databases, extensive hand-searching and contacting of authors. Two reviewers selected independently eligible RCTs, rated the methodological quality and extracted the data, which were analyzed considering the minimal important difference of patient-important outcomes where possible.
Findings
We identified 20 RCTs whereof 18 provided sufficient data for analysis. The methodological quality was low and sample sizes were too small for most trials to produce meaningful results (median total sample size = 28). Data from five trials showed that supplemental oxygen during exercise did not have clinically meaningful effects on health-related quality of life while improvements of exercise capacity may be even larger for patients exercising on room air. RCTs of adding assisted ventilation, nutritional supplements or a number of anabolically acting drugs do not provide sufficient evidence for or against the use any of these supplemental interventions.
Interpretation
There is insufficient evidence for most supplemental interventions during respiratory rehabilitation to estimate their additional value, partly due to methodological shortcomings of included RCTs. Current data do not suggest benefit from supplemental oxygen during exercise, although the methodological quality of included trials limits conclusions. To appropriately assess any of the various supplemental interventions used in clinical practice, pragmatic trials on respiratory rehabilitation of COPD patients need to consider methodological aspects as well as appropriate sample sizes.
==== Body
Introduction
Chronic obstructive pulmonary disease (COPD) has a large impact on health-related quality of life (HRQL) and represents a major health burden in industrialized and developing countries [1-4]. A systematic review including 23 randomized controlled trials (RCTs) has shown that patients with COPD improve their HRQL and exercise capacity during respiratory rehabilitation[5]. Recent data on long-term outcomes after respiratory rehabilitation show reductions of exacerbations and hospitalizations [6-8].
Physical exercise is the central component of respiratory rehabilitation programs because it reverses peripheral muscle dysfunction[9], a highly prevalent comorbidity of COPD associated with increased risk of exacerbations and mortality[10,11]. While respiratory rehabilitation including physical exercise has become a cornerstone of COPD management [12-14], there is controversy about the additional value of several supplemental interventions to support exercise programs such as oxygen during exercise[15] or anabolically acting hormones[16].
Clinicians, who consider these supplemental interventions during respiratory rehabilitation programs, should know their benefits and downsides. They need evidence from RCTs directly comparing respiratory rehabilitation with or without supplements in order to carefully discuss these benefits and downsides with their patients. Therefore, we conducted a systematic review of pragmatic RCTs comparing the effects respiratory rehabilitation with and without any supplemental intervention to assess their added value in HRQL and exercise capacity improvement.
Methods
Identification of studies
We performed electronic database searches in MEDLINE (Ovid version, New York, New York, from inception to May 2004), EMBASE (DataStar version, Cary, North Carolina from inception to December 2003), PEDRO (online version, University of Sydney, Australia, December 2003) and the Cochrane Central Register of Controlled Trials (Oxford, United Kingdom, 2003, Issue 4). We also searched the Science Citation Index database (Web of Science, Thomson ISI, Philadelphia, Pennsylvania) and the "related articles" function of PubMed (National Library of Medicine, Washington, Maryland) by entering all included studies. In addition, we hand searched the bibliographies of all included studies, of reviews on respiratory rehabilitation or physical exercise in patients with COPD that were identified in the literature search, as well as the Proceedings of the International Conferences of the American Thoracic Society and the congress of the European Respiratory Society to identify further relevant studies. We also contacted authors in the field to ask for published or unpublished data.
Selection criteria
We included RCTs investigating any supplemental intervention added to respiratory rehabilitation that included a standardized physical exercise program. We focused on standardized exercise programs because only these allow reproduction in clinical practice. A standardized exercise protocol was defined as the use of an identical exercise activity for all patients (e.g. treadmill walking or cycle ergometer training) at measurable exercise intensity (e.g. in Watts, metabolic equivalents or kilograms). We included studies if more than 90% of study participants patients had COPD according to the following criteria: (1) a clinical diagnosis of COPD, (2) irreversible airways obstruction and (3) one of the following: (a) best recorded FEV1/FVC ratio of individual patients < 0.7; (b) best recorded FEV1 of individual patients < 70% of predicted value. We considered the following outcome measures: HRQL as measured by generic (e.g. SF-36) or disease-specific (e.g. St. George Respiratory Questionnaire) questionnaires, symptom scales, functional exercise capacity as measured walk tests and results from cardiopulmonary exercise testing. We did not apply any language restrictions.
We excluded studies that compared any exercise program versus usual care (i.e. no exercise) or studies that used unstandardised exercise protocols (e.g. home exercise programs).
Data extraction and quality assessment
The bibliographic details of all retrieved articles were stored in a Reference Manager file. We removed duplicate records resulting from the various database searches. Two members of the review team independently scrutinized the titles and abstracts of all identified citations (see figure 1). We ordered the full text of any article that was deemed potentially eligible by one of the reviewers. The two reviewers then evaluated the full text of all retrieved papers, made a decision on in- or exclusion and discussed the decisions. Any disagreement was resolved by consensus with close attention to the inclusion and exclusion criteria. Final decisions on papers were recorded in the Reference Manager file and bibliographic details as well as the reasons for exclusion. We recorded the initial degree of agreement between the reviewers and corrected discordant scores based on obvious errors. We resolved discordant scores based on real differences in interpretation through consensus.
Figure 1 Study flow from identification to final inclusion of studies.
Details about study patients, interventions and outcome measures as well as the results were extracted onto a predefined data form. We pilot tested the data forms using five studies with high likelihood for inclusion.
Two reviewers independently evaluated the methodological quality of included trials reported in full reports using a detailed list of quality items assessing components of internal validity[17] (table 3, see Additional file 3). We also contacted the authors of the primary studies to obtain missing information.
Data synthesis and interpretation
We summarized the results of the data extraction and assessment of study validity in structured tables to allow looking at the variation in patient characteristics, interventions, outcome measures, study quality and results. In addition, we used forest plots to compare results across the trials. If appropriate we planned to explore sources of heterogeneity (i.e. differences between studies) using multivariable regression models (study level meta-regression analysis) where clinical and methodological items would act as explanatory variables. No pooling was undertaken in the presence of significant source heterogeneity.
Whenever possible, for each outcome, estimates and confidence limits was related to its minimal important difference[18]. We assessed whether the estimates and 95% confidence limits for the difference between study groups exceeded the minimal important difference (for the Six-minute walk distance ± 50 meters[19], Chronic Respiratory Questionnaire ± 0.5 points[20] and St. George Respiratory Questionnaire ± 4 points)[21].
Data were analyzed using STATA (version 8.2, Stata Corp., College Station, Texas).
Results
Study selection
Figure 1 shows the study selection process and agreement on study inclusion. Main reasons for study exclusion (Appendix, see Additional file 6) were that patients did not have an exercise programme but only exercise testing with or without oxygen (n = 12), studies were not RCTs (n = 8) and that the control group had no exercise programme (n = 5). We excluded only one study because of an undefined exercise programme. We excluded two trials[22,23] from the analysis because the abstract provided little information and the authors did not provide further details. Initially, we excluded another abstract, but since this trial was published in the meantime[24], we could include it in the analysis.
Quality assessment
Table 3 (see Additional file 3) shows a detailed assessment of the methodological quality of the included trials. Interrater agreement for all items of the quality assessment was 87% (chance corrected agreement: κ = 0.76). In general, most included trials scored poorly on the checklist used. Important methodological aspects that bear on the validity such as blinding of outcome assessors were not or just partially addressed in most trials.
Supplemental oxygen during exercise
The characteristics of the five trials on supplemental oxygen [25-29] are summarized in table 1 (see Additional file 1) and the results shown in figures 2 and 3. There was a trend towards larger improvements of HRQL and exercise duration in constant work rate tests in the groups with oxygen, but patients exercising on room air had larger improvements of the walking distance. Emtner[25] reported that the use of oxygen enabled patients to exercise at higher intensity (mean 62 Watt [SD 19] corresponding to 138% of baseline maximum exercise capacity) compared with patients on room air (52 Watt [SD 22] corresponding to 96% of baseline maximum exercise capacity, p < 0.01 for difference between groups). In the trial by Rooyackers[28], patients achieved mean exercise intensities corresponding to 124% of maximum exercise capacity in the group with oxygen and 114% of maximum exercise capacity in the group without oxygen (p = 0.12). Two trials reported on safety of exercise with oxygen or room air. Rooyackers[28] assessed whether oxygen prevented the development of pulmonary hypertension. The investigators did not find any differences between groups in resting mean pulmonary artery pressure measured by Doppler echocardiography. Waddell[29] did not find significant CO2 retention during walking tests despite high oxygen flow of 5 l/minute.
Figure 2 Effect of supplemental oxygen on health-related quality of life. The forest plot shows the results from three trials comparing physical exercise with and without oxygen, separately for each domains of the Chronic Respiratory Questionnaire (CRQ). The x-axis represents the difference in change scores between study groups with negative values favoring exercise on room air and positive values favoring exercise with supplemental oxygen. A difference of 0 means that both study groups changed to the same amount. Boxes with 95% confidence intervals represent point estimates for the difference between the CRQ change scores (from baseline to follow-up) of the study groups. Dotted lines represent the minimal important difference of the CRQ (change of 0.5). On the right of the forest plot, point estimates for differences between groups and 95% confidence intervals are shown.
Figure 3 Effect of supplemental oxygen on exercise capacity. The forest plot shows the results from five trials comparing respiratory rehabilitation with and without oxygen. Walking tests, incremental and constant work rate exercise tests were used to assess the additional effect of supplemental oxygen during exercise. The x-axis represents the difference in change scores between study groups with negative values favoring exercise on room air and positive values favoring exercise with supplemental oxygen. A difference of 0 means that both study groups changed to the same amount. Boxes with 95% confidence intervals represent point estimates for the difference between the walking distance and maximum exercise capacity change scores (from baseline to follow-up) of the study groups. Dotted lines represent the minimal important difference of the six-minute walking distance (53 meters). On the right of the forest plot, point estimates for differences between groups and 95% confidence intervals are shown.
Assisted ventilation
Two trials[30,31] evaluated proportional assist ventilation during exercise and did not find an additional benefit (tables 1 and 4, see Additional files 1 and 4). Only 50%[30] and 71.4%[31] of patients exercising with positive pressure ventilation and 67%[30] and 60%[31] exercising without positive pressure ventilation completed these trials.
Garrod[32] assessed the benefit of overnight non-invasive positive pressure ventilation at home during the training period. They found a statistically significant improvement of the walking distance for patients assigned to overnight non-invasive positive pressure ventilation. HRQL improvements also tended to be larger for patients with ventilation, but the difference reached only statistical significance for the fatigue domain and total score of the CRQ.
Johnson[33] evaluated the effect of ventilation and Heliox during exercise on exercise duration and intensity. They found a small, but statistically not significant increase in exercise duration and intensity for patients exercising with ventilation and Heliox. Patient satisfaction for overall condition, exercise capability and breathing ability measured with global ratings of change did not differ significantly between groups (exact data not available). In this trial, 73.3% of patients with ventilation, 90.9% of patients with Heliox and 84.6% of patients without a supplement completed the trial.
Nutritional supplements
We identified two RCTs that assess the additional benefit of nutritional supplements during respiratory rehabilitation (table 2, see Additional file 2)[34,35]. Steiner[35] did not find statistically significant differences for HRQL and exercise capacity (table 5, see Additional file 5). In a subgroup of patients with a BMI>19 kg/m2 (22 in group with supplement and 30 in group with placebo) the difference between groups was 27 meters (95% CI 1–53) in the incremental and 121 seconds (95% CI -44–286) in the endurance shuttle walk test. Patients with the carbohydrate-rich diet increased their body weight compared to the placebo group by 1.23 kg (95% CI 0.42–2.05), which occurred mainly because of an increase of the fat mass (difference between groups 1.46 kg, 95% CI 0.65–2.27). There was a dropout rate of 40% in the group with and of 16% in the group without carbohydrate-rich diet.
Another RCT[34] found not significant differences between patients supplemented with an additional fat-rich diet, but did not report the results in detail and could not provide these data for our review. Compared to placebo, non-depleted patients increased their body weight by 1.5 kg (95% CI 0.4–2.6) when receiving a fat-rich diet and by 1.6 kg (95% CI 0.39–2.81) when receiving a fat-rich diet plus anabolic steroids.
Anabolic steroids
Creutzberg[36] (table 2, see Additional file 2) found that only patients receiving nandrolone improved their HRQL, whereas patients following the respiratory rehabilitation program without nandrolone did not change. This trend was consistent for all domains of the St. George Respiratory Questionnaire, but only statistically significant for the symptom domain (table 5, see Additional file 5). For the subgroup of patients receiving maintenance treatment with oral glucocorticosteroids, patients with nandrolone improved their maximum exercise capacity significantly more. Isometric leg strength and isokinetic legwork improved in both groups, but did not differ significantly between groups. There was a trend in erythropoetic parameters towards an increase of erythrocyte count, hematocrit and hemoglobin in patients treated with nandrolone compared to those treated with placebo. No changes in blood pressures and any androgenic effects or fluid retention were registered in either group.
Casaburi[24] assessed the additional benefit of testosterone for male COPD patients with low testosterone levels who followed a strength exercise program (table 2, see Additional file 2). The group with testosterone had larger increases in exercise capacity and muscle strength, but none of the differences reached statistical significance (table 5, see Additional file 5). Total lean mass increased and total fat mass decreased more in patients with supplemental testosterone, but differences between groups were not significantly different (mean difference in changes between groups in lean mass 3.09 kg, p > 0.05, and total fat mass -1.28 kg, p > 0.05). Casaburi found, like Creutzberg[36], differences in hemoglobin changes between groups (mean difference in hemoglobin change between the testosterone and placebo group 1.4 g/dL, p < 0.05). They observed neither adverse events nor any differences in most safety measures between groups (prostate specific antigen, liver enzymes, alkaline phosphates, cholesterol and high-density lipoprotein cholesterol) between groups. Serum creatinine levels, however, increased in the testosterone group by 0.12 mg/dL and decreased in the placebo group by 0.05 mg/dL (difference between groups 0.17 mg/dL, p < 0.05).
Tiotropium, Creatine, Coenzyme Q10, and growth hormone
Casaburi[37] assessed the additional benefit of tiotropium in 47 patients and found a significantly increased exercise endurance time compared to patients who received placebo (n = 44, tables 2 and 5, see Additional files 2 and 5). Further results were not available. Four small RCTs evaluated the additional benefit of creatine[38], coenzyme Q10[39] and growth hormone[40,41] during respiratory rehabilitation, but did not find any additional benefit on respiratory or peripheral muscle function or HRQL (tables 2 and 5, see Additional files 2 and 5). Casaburi et al[41] reported that no adverse effects of growth hormone occurred.
Discussion
There are three main results from this systematic review. First, evidence suggests that supplemental oxygen during physical exercise does not provide a clinically relevant benefit. Second, the evidence for any other supplemental intervention is not strong enough to recommend or discourage their use in clinical practice and third, there were major methodological limitations in most trials that may explain some of the inconclusive findings. We discuss each of these results in turn.
Cotes[42] reported in 1956 that oxygen increased exercise performance in patients with COPD. Since then, many investigators assessed the short-term effect of increased oxygen availability during exercise[15]. Some investigators argue that patients tolerate higher exercise intensities or longer exercise time with supplemental oxygen leading to larger training effects[43,44]. Others believe that only without oxygen, an adequate hypoxemic stimulus is provided for peripheral muscles to improve exercise capacity.
The studies by Emtner[25] and Rooyackers[28] demonstrated that patients indeed tolerate higher exercise intensities if supplemented by oxygen. Mean differences on the CRQ domain scores, however, showed a slight but clinically not meaningful trend towards a benefit with oxygen supplementation (figure 2). The trial by Emtner[25] was the only one that showed a consistent trend towards a small benefit of oxygen on HRQL and exercise capacity. Across all studies, however, results from exercise testing were contradicting. Supplemental oxygen did prolong exercise duration in constant work rate tests, but led to considerably smaller improvements of functional exercise capacity (figure 3). It was hypothesized earlier that those patients with the highest oxygen desaturation during exercise would benefit most from supplemental oxygen[45]. The trials do not provide sufficient evidence for or against this hypothesis.
There is limited evidence on the safety of oxygen during exercise and on the safety of exercise without oxygen in patients with desaturation. Clinicians may have concerns about training in hypoxemia because of adverse events and will encourage oxygen supplementation in patients with desaturation during exercise. In theory, oxygen carries the risk of CO2 retention in COPD patients. The only trial reporting on CO2 retention[29] did not observe significant differences of CO2 levels during exercise tests with oxygen compared with exercise on room air. However, exercise tests may have been too short to assess the effect of CO2 retention. Exercise is a risk indicator for unmasking latent pulmonary hypertension[46], but supplemental oxygen may reduce this risk by decreasing the sympathetic tone and the respiratory rate allowing for less end-expiratory pressure[47]. Rooyackers[28] did not find any differences in resting mean pulmonary artery pressure between patients with and without oxygen. However, patients stopped exercising when oxygen saturation fell below 90% so that the risk of the exercise program under hypoxemic conditions on the development of pulmonary hypertensions could not be studied.
Several studies found a positive acute effect of oxygen during exercise testing on exercise capacity and a number of physiologic mechanisms for the effects of oxygen have been proposed [48-50]. However, these results on the short-term benefit of oxygen during exercise testing seem not to translate into improvements of clinically relevant outcomes during exercise programs. Current data do not suggest benefit from the use of oxygen during exercise to enhance training effects (figure 3), but show some benefit in terms of HRQL (figure 2) Given the limited methodological quality of trials, any conclusions are vague. The general use of oxygen is only justified, if larger trials of good quality show its benefit on clinically relevant outcomes. The mechanisms of the effects of oxygen during exercise are still insufficiently understood and call for more basic research[15].
Assisted ventilation also aims at increasing oxygen availability during exercise, but the trials indicated no additional benefit. An exception may represent overnight non-invasive positive pressure ventilation. This treatment may improve quality of sleep as well as daytime gas levels and respiratory muscle function thereby providing a better milieu (pH, PaO2, PaCO2) for peripheral muscle function. One trial[32] found statistically significant improvements of functional exercise capacity and also large improvements of HRQL (mean differences between groups 0.45 to 0.85 in CRQ domain scores, table 4, see Additional file 4) with additional non-invasive positive pressure ventilation. These results support the hypothesis formulated by authors of a recent meta-analysis showing that nocturnal non-invasive positive pressure ventilation alone has no effect on exercise capacity and HRQL, but may be beneficial as an adjunct to respiratory rehabilitation[51]. The eight trials that assessed various supplemental interventions during rehabilitation produced inconclusive results that do not allow recommendations for clinical practice yet.
An important result of this systematic review with implications for future research is the low methodological quality and small sample sizes. For example, the majority of trials did not consider stratification for important prognostic factors such as exercise capacity[52] for randomization. In some trials there were baseline imbalances between groups, for example in terms of exercise capacity[27,28,32,33,40]. The influence of these imbalances on the results was not investigated in any of the trials. Concealment of random allocation and blinding of treatment providers or outcome assessors was also not addressed in most trials.
Sample sizes were small except in three trials[34,35,37]. Pragmatic trials comparing active interventions, as included in this systematic review, are very useful for clinical practice when clinicians are confronted with the choice between interventions[53]. However, small sample sizes are problematic in pragmatic trials for at least two reasons: First, differences between study groups tend to be smaller in pragmatic trials than in trials comparing an active intervention with placebo or a sham intervention. Figure 4 shows the results and 95% confidence intervals of a trial comparing respiratory rehabilitation with usual care and of a trial comparing respiratory rehabilitation with and without a supplemental intervention with different sample sizes. It illustrates the importance of sufficient sample sizes in pragmatic trials by showing that for pragmatic RCTs in respiratory rehabilitation, in which widely established patient-important outcomes such as HRQL are used, sample sizes of up to 40 per group will produce imprecise results (large confidence intervals). This imprecision hinders interpretation. Another reason for sufficient samples sizes is that in pragmatic trials patient profiles are usually more variable than in explanatory trials reflecting the wide patient spectrum encountered in clinical practice[53]. The greater variability in patient profiles carries, on one side, a greater risk for confounding and, on the other side, subgroup analyses will be important to assess whether the effects differ between patient subgroups (effect modification). Subgroup analyses based on prognostically important patient characteristics will provide more differentiated evaluations than one mean for the whole study group, but they require sufficient sample for well-balanced intervention groups.
Figure 4 Sample size and interpretation of randomized controlled trials in respiratory rehabilitation. Forest plot with simulated results from two trials with varying sample size, in which the CRQ was used. Boxes with 95% confidence intervals represent point estimates for the difference between CRQ change scores (from baseline to follow-up) of the study groups. Dotted lines represent the minimal important difference of the CRQ (change of 0.5). Trial 1 shows the results from a typical explanatory trial comparing respiratory rehabilitation and no respiratory rehabilitation (usual care) with differences in CRQ change scores around 0.75[5]. Because of the large effect, trial results are interpretable also with imprecise results. Trial 2 shows the results from a pragmatic trial assessing the additional effect of a supplemental intervention (for example oxygen during exercise). The difference between groups is 0.3 and sample size must be large (80 per group) to produce results that are precise enough to allow interpretation.
We propose that investigators consider the following aspects in future pragmatic trials on respiratory rehabilitation: First, preliminary sample size considerations should be based on realistic estimates for expected differences between groups, which are typically smaller than in trials without active comparators. To better understand what these sample sizes mean, 95% confidence intervals around the predicted point estimate can be calculated as shown in figure 4. This approach will help to better foresee the consequence of different sample sizes on interpretation of the data[54]. Second, COPD patients represent a heterogeneous group and stratification for prognostically important variables should be considered to avoid baseline imbalances that bear on outcomes[55], as seen in some trials included in this review[27,28,32,33,40]. Third, more attention needs to be paid to general requirements for RCTs of high quality like method of randomisation, concealment of random allocation and blinding of those who assess treatment effects.
The strengths of our systematic review include the broad literature search including several databases and extensive hand searching for trials with direct comparisons of interventions that are used in clinical practice. In addition, we contacted authors for additional data and received them from the majority of investigators. This greatly enhanced the informativeness of included studies and thereby of this review. A weakness of this review includes the discussion that is limited to the best-investigated supplements because of the number of interventions included in this review. However, the aim of this review was to analyze current evidence from a meta-epidemiological perspective not giving to much emphasis to single studies. Some may criticize that we did not pool the results from trials on supplemental oxygen during exercise using meta-analysis. However, desaturation or no desaturation during exercise was an important inclusion criterion in four of the five trials and investigators wanted to learn about the effect of supplemental oxygen in these subgroups, in particular. Therefore, we considered the patient profiles of these trials to be too different to provide meaningful pooled estimates. Instead, we provided forest plots (figures 2 and 3) to show the individual studies' point estimates and 95% confidence intervals for the CRQ domains and the exercise tests to allow comparisons across studies.
In conclusion, data for most supplemental interventions during respiratory rehabilitation are inconclusive. Oxygen during exercise does not seem to provide a patient-important additional benefit for COPD patients during a respiratory rehabilitation, but methodological shortcomings of the trials on supplemental oxygen do not allow conclusive answers. Future trials should pay careful attention to the methodological trial design and to sufficient sample sizes.
Abbreviations
COPD: Chronic obstructive pulmonary disease
RCT: Randomised controlled trial
HRQL: Health-related quality of life
CRQ: Chronic Respiratory Questionnaire
Conflict of interest
The authors declare that they have no competing interests.
Contributions
Protocol writing: Puhan, Bachmann, Schunemann
Acquisition of data: Puhan, Bachmann
Analysis and interpretation of data: Puhan, Bachmann, Schunemann, Frey
Drafting of manuscript: Puhan, Bachmann
Critical revision of manuscript for important intellectual content: Puhan, Bachmann, Schunemann, Frey
Funding
Helmut Horten Foundation
Lucas M. Bachmann: Swiss National Science Foundation Research Fellow (PROSPER programme)
Supplementary Material
Additional File 3
Table 3: Internal validity of included studies
Click here for file
Additional File 6
Appendix: The appendix lists all studies that were excluded after full text assessment. The full reference and the reason for exclusion are given.
Click here for file
Additional File 1
Table 1: Characteristics of randomised controlled trials investigating supplemental oxygen and assisted ventilation
Click here for file
Additional File 4
Table 4: Effect of assisted ventilation on HRQL and exercise capacity
Click here for file
Additional File 2
Table 2: Characteristics of randomised controlled trials investigating drug and nutritional supplements
Click here for file
Additional File 5
Table 5: Effect of drug and nutritional interventions on HRQL and exercise capacity
Click here for file
==== Refs
Seemungal T Donaldson G Paul E Bestall J Jeffries D Wedzicha JADW Effect of Exacerbation on Quality of Life in Patients with Chronic Obstructive Pulmonary Disease Am J Respir Crit Care Med 1998 157 1418 1422 9603117
Sin DD Stafinski T Ng YC Bell NR Jacobs P The Impact of Chronic Obstructive Pulmonary Disease on Work Loss in the United States Am J Respir Crit Care Med 2002 165 704 707 11874818
Sullivan SD Ramsey SD Lee TA The economic burden of COPD Chest 2000 117 5S 9S 10673466 10.1378/chest.117.2_suppl.5S
Chan-Yeung M Ait-Khaled N White N Ip MS Tan WC The burden and impact of COPD in Asia and Africa Int J Tuberc Lung Dis 2004 8 2 14 14974740
Lacasse Y Brosseau L Milne S Martin S Wong E Guyatt GH Goldstein RS Pulmonary rehabilitation for chronic obstructive pulmonary disease Cochrane Database Syst Rev 2002 CD003793 12137716
Behnke M Jorres RA Kirsten D Magnussen H Clinical benefits of a combined hospital and home-based exercise programme over 18 months in patients with severe COPD Monaldi Arch Chest Dis 2003 59 44 51 14533282
Griffiths TL Burr ML Campbell IA Lewis-Jenkins V Mullins J Shiels K Turner-Lawlor PJ Payne N Newcombe RG Ionescu AA Thomas J Tunbridge J Lonescu AA Results at 1 year of outpatient multidisciplinary pulmonary rehabilitation: a randomised controlled trial Lancet 2000 355 362 368 10665556 10.1016/S0140-6736(99)07042-7
Troosters T Gosselink R Decramer M Short- and long-term effects of outpatient rehabilitation in patients with chronic obstructive pulmonary disease: a randomized trial Am J Med 2000 109 207 212 10974183 10.1016/S0002-9343(00)00472-1
Maltais F LeBlanc P Simard C Jobin J Berube C Bruneau J Carrier L Belleau R Skeletal muscle adaptation to endurance training in patients with chronic obstructive pulmonary disease Am J Respir Crit Care Med 1996 154 442 447 8756820
Celli BR Cote CG Marin JM Casanova C Montes O Mendez RA Pinto PV Cabral HJ The body-mass index, airflow obstruction, dyspnea, and exercise capacity index in chronic obstructive pulmonary disease N Engl J Med 2004 350 1005 1012 14999112 10.1056/NEJMoa021322
Decramer M Gosselink R Troosters T Verschueren M Evers G Muscle weakness is related to utilization of health care resources in COPD patients Eur Respir J 1997 10 417 423 9042643 10.1183/09031936.97.10020417
BTS Guidelines for the Management of Chronic Obstructive Pulmonary Disease Thorax 1997 52 Suppl 5 S1 28 9474238
Pulmonary rehabilitation-1999. American Thoracic Society Am J Respir Crit Care Med 1999 159 1666 1682 10228143
National Heart Lung and Blood Institute Global strategy for the diagnosis, management and prevention of Chronic Obstructive Pulmonary Disease National Institutes of Health 2003
Snider GL Enhancement of exercise performance in COPD patients by hyperoxia: a call for research Chest 2002 122 1830 1836 12426288 10.1378/chest.122.5.1830
Ferreira IM Verreschi IT Nery LE Goldstein RS Zamel N Brooks D Jardim JR The influence of 6 months of oral anabolic steroids on body mass and respiratory muscles in undernourished COPD patients Chest 1998 114 19 28 9674442
ter-Riet-G. Kessels-A-G-H Does electroacupuncture reduce craving for alcohol? A randomized controlled study Complementary Therapies in Medicine 1997 116 118
Guyatt GH Osoba D Wu AW Wyrwich KW Norman GR Methods to explain the clinical significance of health status measures Mayo Clin Proc 2002 77 371 383 11936935
Redelmeier DA Bayoumi AM Goldstein RS Guyatt GH Interpreting small differences in functional status: the Six Minute Walk test in chronic lung disease patients Am J Respir Crit Care Med 1997 155 1278 1282 9105067
Jaeschke R Singer J Guyatt GH Measurement of health status. Ascertaining the minimal clinically important difference Control Clin Trials 1989 10 407 415 2691207 10.1016/0197-2456(89)90005-6
Jones PW Interpreting thresholds for a clinically significant change in health status in asthma and COPD Eur Respir J 2002 19 398 404 11936514 10.1183/09031936.02.00063702
Patessio A Casaburi R Carone M Appendini L Purro A Gudjonsdottir M Donner CF Wasserman K Effects of supplemental oxygen vs room air breathing during training on exercise responses in non desaturating COPD patients European Respiratory Journal 1996 9 379s
Shu MF Kao CH Kuo HP Upper arm exercise improves exercise tolerance and dyspnea sensation in patients with chronic obstructive airway disease (COAD) European Respiratory Journal 1998 12 (suppl) 406S
Casaburi R Bhasin S Cosentino L Porszasz J Somfay A Lewis MI Fournier M Storer TW Effects of testosterone and resistance training in men with chronic obstructive pulmonary disease Am J Respir Crit Care Med 2004 170 870 878 15271690 10.1164/rccm.200305-617OC
Emtner M Porszasz J Burns M Somfay A Casaburi R Benefits of supplemental oxygen in exercise training in nonhypoxemic chronic obstructive pulmonary disease patients Am J Respir Crit Care Med 2003 168 1034 1042 12869359 10.1164/rccm.200212-1525OC
Fichter J Fleckenstein J Stahl C Sybrecht GW Effect of oxygen (FI02: 0.35) on the aerobic capacity in patients with COPD Pneumologie 1999 53 121 126 10226473
Garrod R Paul EA Wedzicha JA Supplemental oxygen during pulmonary rehabilitation in patients with COPD with exercise hypoxaemia. Thorax 2000 55 539 543 10856310 10.1136/thorax.55.7.539
Rooyackers JM Dekhuijzen PN van Herwaarden CL Folgering HT Training with supplemental oxygen in patients with COPD and hypoxaemia at peak exercise European Respiratory Journal 1997 10 1278 1284 9192929 10.1183/09031936.97.10061278
Wadell K HenrikssonLarsen K Lundgren R Physical training with and without oxygen in patients with chronic obstructive pulmonary disease and exercise-induced hypoxaemia Journal of Rehabilitation Medicine 2001 33 200 205 11585150 10.1080/165019701750419581
Bianchi L Foglio K Porta R Baiardi R Vitacca M Ambrosino N Lack of additional effect of adjunct of assisted ventilation to pulmonary rehabilitation in mild COPD patients Respiratory Medicine 2002 96 359 367 12113387 10.1053/rmed.2001.1287
Hawkins P Johnson LC Nikoletou D Hamnegard CH Sherwood R Polkey MI Moxham J Proportional assist ventilation as an aid to exercise training in severe chronic obstructive pulmonary disease Thorax 2002 57 853 859 12324670 10.1136/thorax.57.10.853
Garrod R Mikelsons C Paul EA Wedzicha JA Randomized controlled trial of domiciliary noninvasive positive pressure ventilation and physical training in severe chronic obstructive pulmonary disease Am J Respir Crit Care Med 2000 162 4 pt 1 1335 1341 11029341
Johnson JE Gavin DJ Dramiga S Effects of training with heliox and noninvasive positive pressure ventilation on exercise ability in patients with severe COPD Chest 2002 122 464 472 12171818 10.1378/chest.122.2.464
Schols AM Soeters PB Mostert R Pluymers RJ Wouters EF Physiologic effects of nutritional support and anabolic steroids in patients with chronic obstructive pulmonary disease. A placebo-controlled randomized trial Am J Respir Crit Care Med 1995 152 4 pt 1 1268 1274 7551381
Steiner MC Barton RL Singh SJ Morgan MD Nutritional enhancement of exercise performance in chronic obstructive pulmonary disease: a randomised controlled trial. Thorax 2003 58 745 751 12947128 10.1136/thorax.58.9.745
Creutzberg EC Wouters EF Mostert R Pluymers RJ Schols AM A role for anabolic steroids in the rehabilitation of patients with COPD? A double-blind, placebo-controlled, randomized trial Chest 2003 124 1733 1742 14605042 10.1378/chest.124.5.1733
Casaburi R Kukafka D Cooper CB Kesten S Improvement in Exercise Endurance with the Combination of Tiotropium and Rehabilitative Exercise Training in COPD Patients American Journal of Respiratory & Critical Care Medicine 2004 169 A756
Gosselink R Spruit MA Troosters T Kladka P Sliwinski P Nowinski A Zielinski J Decramer M Oral Creatine Supplementation (CR) in COPD Exercise Training: A Randomized, Double Blind, Placebo (PL) Controlled Trial American Journal of Respiratory & Critical Care Medicine 2001 163 A961
Satta A Grandi M Landoni CV Migliori GB Spanevello A Vocaturo G Neri M Effects of ubidecarenone in an exercise training program for patients with chronic obstructive pulmonary diseases Clinical Therapeutics 1991 13 754 757 1790550
Burdet L De Muralt B Schutz Y Pichard C Fitting JW Administration of growth hormone to underweight patients with chronic obstructive pulmonary disease. A prospective, randomized, controlled study Am J Respir Crit Care Med 1997 156 1800 1806 9412558
Casaburi R Carithers J Tosoline R Randomized controlled trial of growth hormone in severe COPD patients undergoing endurance exercise training American Journal of Respiratory & Critical Care Medicine 1997 155 A498
Cotes JE Gilson JC Effect of Oxygen on Exercise Ability in Chronic Respiratory Insufficiency - Use of Portable Apparatus Lancet 1956 270 872 876 13320918 10.1016/S0140-6736(56)91341-1
Casaburi R Patessio A Ioli F Zanaboni S Donner CF Wasserman K Reductions in exercise lactic acidosis and ventilation as a result of exercise training in patients with obstructive lung disease American Review of Respiratory Disease 1991 143 9 18 1986689
Hardman AE Issues of fractionization of exercise (short vs long bouts) Med Sci Sports Exerc 2001 33 S421 S427 11427766
Donner CF Howard P Pulmonary rehabilitation in chronic obstructive pulmonary disease (COPD) with recommendations for its use. Report of the European Respiratory Society Rehabilitation and Chronic Care Scientific Group (S.E.P.C.R. Rehabilitation Working Group) Eur Respir J 1992 5 266 275 1559592
Burrows B Kettel LJ Niden AH Rabinowitz M Diener CF Patterns of cardiovascular dysfunction in chronic obstructive lung disease N Engl J Med 1972 286 912 918 5013974
Voelkel NF Cool CD Pulmonary vascular involvement in chronic obstructive pulmonary disease Eur Respir J Suppl 2003 46 28s 32s 14621104
Dean NC Brown JK Himelman RB Doherty JJ Gold WM Stulbarg MS Oxygen may improve dyspnea and endurance in patients with chronic obstructive pulmonary disease and only mild hypoxemia Am Rev Respir Dis 1992 146 941 945 1416422
Maltais F Simon M Jobin J Desmeules M Sullivan MJ Belanger M LeBlanc P Effects of oxygen on lower limb blood flow and O2 uptake during exercise in COPD Med Sci Sports Exerc 2001 33 916 922 11404656 10.1097/00005768-200105001-01404
Somfay A Porszasz J Lee SM Casaburi R Dose-response effect of oxygen on hyperinflation and exercise endurance in nonhypoxaemic COPD patients Eur Respir J 2001 18 77 84 11510809 10.1183/09031936.01.00082201
Wijkstra PJ Lacasse Y Guyatt GH Casanova C Gay PC Meecham JJ Goldstein RS A meta-analysis of nocturnal noninvasive positive pressure ventilation in patients with stable COPD Chest 2003 124 337 343 12853542 10.1378/chest.124.1.337
Zu Wallack RL Patel K Reardon JZ Clark BAIII Normandin EA Predictors of improvement in the 12-minute walking distance following a six-week outpatient pulmonary rehabilitation program Chest 1991 99 805 808 2009778
Tunis SR Stryer DB Clancy CM Practical clinical trials: increasing the value of clinical research for decision making in clinical and health policy JAMA 2003 290 1624 1632 14506122 10.1001/jama.290.12.1624
Goodman SN Berlin JA The use of predicted confidence intervals when planning experiments and the misuse of power when interpreting results Ann Intern Med 1994 121 200 206 8017747
Puhan MA Busching G VanOort E Zaugg C Schunemann HJ Frey M Interval exercise versus continuous exercise in patients with moderate to severe chronic obstructive pulmonary disease - study protocol for a randomised controlled trial [SRCTN11611768] BMC Pulm Med 2004 4 5 15310394 10.1186/1471-2466-4-5
| 15575956 | PMC539299 | CC BY | 2021-01-04 16:47:22 | no | Respir Res. 2004 Dec 2; 5(1):25 | utf-8 | Respir Res | 2,004 | 10.1186/1465-9921-5-25 | oa_comm |
==== Front
BMC Emerg MedBMC Emergency Medicine1471-227XBioMed Central London 1471-227X-4-51558505610.1186/1471-227X-4-5Research ArticleManagement of acute renal colic in the UK: a questionnaire survey Lasoye Tunji A [email protected] Philip M [email protected] Nilay [email protected] Chas [email protected] Nadeem [email protected] Accident and Emergency Department, King's College Hospital, London, England, UK2 Department of Community Health Sciences, St George's Hospital Medical School, London, England, UK3 Department of Urology, Churchill Hospital, Oxford, England, UK4 Department of Psychology, University of Southampton, Southampton, England, UK5 Accident and Emergency Department, University Hospital Lewisham, London, England, UK2004 7 12 2004 4 5 5 2 4 2004 7 12 2004 Copyright © 2004 Lasoye et al; licensee BioMed Central Ltd.2004Lasoye 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 great variation in the Accident and Emergency workload and location of Urology services in UK hospitals. This study investigated the relationship of the initial management of acute renal colic with the department workload plus local facilities including location of X-ray and urology services in UK Accident and Emergency (A&E) departments.
Methods
A&E departments in each of the 11 UK Deanery regions were stratified based on departmental workload, namely <30,000 (small); 30,000 to 50,000 (medium); 50,000 to 80,000 (large) and >80,000 (very large) patients per year. One third of departments were selected in each group leading to a sample size of 106. A questionnaire was administered. Associations between categorical variables were investigated using the chi-squared test and when not valid, Fisher's Exact test was employed. Differences between groups in ordinal variables were investigated using the Mann-Whitney test.
Results
All questionnaires were returned. Twenty-nine units (27.4%) did not perform any radiological investigation on renal colic patients. The number of radiological investigations that were available to departments was associated with workload (P = 0.003); with 57.1% of the small departments performing none and at least 82.8% of units in the other categories performing at least one. Of those departments with X-ray facilities in or adjacent to the department, 63% performed an intravenous urography (IVU) compared to 25% of those departments without (P = 0.026). Of those departments with on-site urology services, 86% performed at least one radiological investigation compared to 52% of units without such services (P = 0.001). Department workload was associated with the first choice analgesia (NSAIDs or parenteral opiates) (P = 0.011). Of the small departments, 64.3% used NSAIDs, 21.4% used parenteral opiates and 14.3% used neither. In comparison, NSAIDS were used by at least 87%, and opiates by at most 12.5% of units in each of the other three categories of department workload.
Conclusions
Over a quarter of UK A&E departments did not perform any radiological investigations and some departments do not even offer renal colic patients any analgesia. Patient management was associated with departmental workload, location of X-ray and Urology services. National guidelines are needed to ensure optimum care for all patients.
==== Body
Background
Upon presentation to the A&E department, suspected acute renal colic patients must have a clinical examination and radiological investigations to confirm the diagnosis. Without radiological investigations, life-threatening conditions such as abdominal aortic aneurysm and ectopic pregnancy may be misdiagnosed as renal colic. However, a delay in the diagnosis is possible as the facilities needed for the diagnosis are sometimes not based in the same hospital as the A&E department. In the UK, great variation exists between Accident and Emergency services in their workload as measured by the number of new patients seen per year[1]. Larger A&E departments tend to be located in those teaching hospitals that have most specialist services on-site, therefore facilitating adequate investigation of suspected renal colic. Every A&E department in the UK should be able to perform initial assessment and investigation of suspected renal colic, provide pain relief and refer appropriately, irrespective of the location of urology services. If acute renal colic presented to primary care then the patient would be rapidly referred to secondary care, namely an A&E department. The aims of this study were to investigate the initial management of acute renal colic in UK A&E departments and if practice was related to department workload, plus the location of X-ray and urological services in relation to the A&E department.
Methods
The handbook of the British Association for Accident and Emergency lists a total of 311 A&E departments in the UK[1]. The UK is divided into 11 so-called Deanary regions that represent geographical areas. These units were categorised according to their workload (number of new patients seen per year) as follows: Small- less than 30,000, Medium- 30,000 to 50,000, Large -50,000 to 80,000 and Very Large-more than 80,000. For each of the four-workload categories in each of the 11 Deanery regions, every third unit was selected resulting in a total of 106 (34.1%) departments.
A questionnaire was administered by post to the 106 departments (see Appendix 1) requesting details about the location of X-ray, location of Urology services plus current practice in the investigation and management of pain in acute renal colic patients. A covering letter was included indicating that the purpose of the survey was to collect information about practice when patients present to the A&E department and not their subsequent management. The most senior medical member of each department was invited to complete the questionnaire. Over a period of nine months, each of the 106 departments returned a completed questionnaire. A total of 35 departments did not respond initially and they were sent a second questionnaire by post as a reminder. Ten departments did not respond to the second questionnaire and these were followed up with a telephone call. Consultants completed the questionnaire in 74.5% (n = 79) of units. Middle grade doctors who had been in post for at least six months completed the remaining 27 (25.5%) questionnaires.
Statistical methods
The Chi-squared test (test statistic denoted by χ2) was used to investigate the following associations: a) the location of X-ray and intravenous urography; b) location of Urology services and total number of investigations performed and c) categorised departmental workload with the investigations performed and also the analgesics used. When the Chi-squared test was invalid, Fisher's Exact test (test statistic denoted by FI) was employed. The Chi-Squared was considered to be invalid if more than 20% of the cells had an expected value less than five or if one of the cells had an expected value less than one [2]. The categorised departmental workload groups were compared in the number of films used during an IVU procedure using the Kruskal-Wallis test (test statistic approximated to the Chi-Squared distribution and denoted by χ2) [3]. Degrees of freedom were abbreviated to df. The critical significance level was 0.05. All statistical analyses were performed using SPSS for Windows (version 11).
Results
On-site services
Of the 106 departments, a total of 94 (88.7%) had X-ray facilities located in the department. A greater proportion of those departments that have X-ray facilities within their premises used the Intra-Venous Urogram (IVU) option compared to those departments without these facilities [n = 59 (62.5%) versus n = 3 (25%); FI = 6.03, df = 1, P = 0.026].
Urology was located within the hospital for 64 (60.4%) departments. The total number of radiological investigations [IVU, Ultrasound Scan (USS) or Computed Tomogram (CT)] that were available to units was categorised as none, one and two or more. Those departments that had urology on-site had more radiological options available than those without (P = 0.001) (see Table 1). At least one radiological option was used by 85.9% (n = 55) of those units with on-site urology services compared to 52.3% (n = 22) of units without.
Table 1 Association between total number of radiological investigations performed and location of urology services.
Total Number of Investigations
Location of Urology services None One Two or three Total
Within Hospital 9 (14.1%) 36 (56.3%) 19 (29.7%) 64 (60.4%)
Outside Hospital 20 (47.6%) 13 (31.0%) 9 (21.4%) 42 (39.6%)
Percentages in brackets are those within the category of the location of urology services; those in the 'total' column are those for the whole sample (n = 106). There was a significant difference between hospitals as regards their location of services in the number of investigations performed (χ2 = 14.6, df = 2, P = 0.0007).
None of the departments in our study routinely used nuclear medicine to investigate renal colic.
Radiological investigations
Intra-Venous Urogram (IVU)
A significant relationship existed between department workload and if an IVU option was available (P = 0.001) (see Table 2). An IVU option was available to 28.6% of the small departments, compared to at least 62.5% of those units in the larger categories, namely the medium, large and very large departments.
Table 2 Tabulation of department workload by radiological investigations performed plus total number of investigations, and number of films used in IVU investigations.
Number of new patients per year
< 30,000 30,000 to 50,000 50,000 to 80,000 >80,000 All departments Test statistics
IVU performed
No 20(71.4%) 8 (22.9%) 13 (37.1%) 3 (37.5%) 44 (41.5%) FI = 15.54, df = 3, P = 0.001
Yes 8 (28.6%) 27 (77.1%) 22 (62.9%) 5 (62.5%) 62 (58.5%)
USS performed
No 21 (75.0%) 25 (71.4%) 17 (48.6%) 4 (50.0%) 67 (63.2%) χ2 = 6.52, df = 3, P = 0.089
Yes 7 (25.0%) 10 (28.6%) 18 (51.4%) 4 (50.0%) 39 (36.8%)
CT performed
No 27 (96.4%) 33 (94.3%) 32 (91.4%) 5 (62.5%) 97 (91.5%) FI = 6.87, df = 3, P = 0.056
Yes 1 (3.6%) 2 (5.7%) 3 (8.6%) 3 (37.5%) 9 (8.5%)
Total number of investigations
None 16 (57.1%) 6 (17.1%) 6 (17.1%) 1 (12.5%) 29 (27.4%) FI = 18.85, df = 6, P = 0.003
One 9 (32.2%) 21 (60.0%) 16 (45.7%) 3 (37.5%) 49 (46.2%)
Two or three 3 (10.7%) 8 (22.9%) 13 (37.2%) 4 (50.0%) 28 (26.4%)
If IVU, total Number of films
n 8 27 22 5 62 χ2 = 6.68, df = 3, P = 0.083
mean 3.0 2.9 2.6 1.6 2.7
standard deviation 1.14 0.97 1.05 0.89 1.10
median 3 3 3 1 3
lower quartile 2.0 3.0 2.0 1.0 2
upper quartile 4.5 3.0 3.0 2.5 3
Percentages in brackets are those of the grouped department workload; those in the "All departments" column are of the 106 units. The test statistics comparing the four groups of department size are displayed.
The relationship between department workload and the average number of films used when an IVU was performed is shown in Table 2. Of the 106 departments, 43 (40.6%) did not undertake an IVU leaving a total of 63 (59.4%) units for analysis. All of these 63 departments used between one and five films per IVU investigation except for the very large category where the greatest number of films used by a department was three. The very large departments used a median number of a single film whilst the other three categories of department size used a median number of three films. Although there was a tendency for fewer films to be used as departmental size increased, this just failed to reach statistical significance at the 5% level (P = 0.083).
Ultrasound Scan (USS)
There was no statistically significant relationship between department workload and if an USS option was available (P = 0.089). However, at least half of the large and the very large units used USS compared to less than 30% of the departments in the small and medium sized categories (see Table 2). Overall, 36.8% (n = 39) of departments were able to perform an USS.
Computerised Tomogram (CT) Scan-Helical CT
The relationship between department workload and if a CT scan was available just missed statistical significance (P = 0.056) (see Table 2). Of the very large departments, 37.5% (n = 3) could perform a CT scan compared to less than 10% of the units in each of the small, medium and large categories.
Total number of radiological investigations
A total of 29 units (27.4%) did not perform any radiological investigations. The relationship between the total number of investigations available and department workload was statistically significant (P = 0.003) (see Table 2). No radiological investigations were carried out by 16 (57.1%) of the small departments whilst at least 83% of the units in each of the other three departmental workload categories were able to perform at least one radiological investigation. Exactly half (n = 4) of the very large departments had at least two options available.
Choice of analgesia
There was a statistically significant relationship between department workload and the first choice analgesia: either NSAIDs (Diclofenac or Ketorolac) or parenteral opiates (P = 0.011) (see Table 3). Parenteral opiates were used by 21.4% (n = 6) of the small departments compared to at most 12.5% of units in the other workload categories. Neither NSAIDs nor parenteral opiate was used by four (14.3%) of the small departments and one large department; one of these small units plus the large department reported using codydramol (a combination of paracetamol with dihydrocodeine). Of the 106 departments, 91 (85.8%) used NSAIDs including 86 (81.1%) – diclofenac and five (4.7%)- ketorolac as the first choice analgesia. Of the 86 departments that used diclofenac, 68 (79.1%) routinely used the intra-muscular route, 17 (19.7%) the rectal route and one (1.2%) administered it orally.
Table 3 First choice analgesic (either NSAIDs, Parenteral opiates or neither) by department workload (n = 106).
First choice analgesia (NSAIDs or Parenteral opiates) Number of new patients per year All departments
< 30,000 30,000 to 50,000 50,000 to 80,000 > 80,000
None used 4 (14.3%) 0 1 (2.9%) 0 5 (4.7%)
NSAIDs 18 (64.3%) 34 (97.1%) 32 (91.4%) 7 (87.5%) 91 (85.8%)
Parenteral opiates 6 (21.4%) 1 (2.9%) 2 (5.7%) 1 (12.5%) 10 (9.4%)
Percentages in brackets are those of the grouped departmental workload; those in the "All departments" column are of the 106 units. There was significant difference between department workloads in first choice analgesia (FI = 13.49, df = 6, P = 0.011).
Discussion
This study reports the initial management of renal colic irrespective of which specialty team carried out the management. Traditionally, renal colic was confirmed by IVU alone [4] although the use of USS and helical CT scans has increased in current practice [5]. A study of a single department reported that up to 37% of patients with suspected renal colic were investigated with ultrasound, although this included mainly patients with allergy to the contrast used in IVU and those in early pregnancy when irradiation needs to be avoided [6]. There may to be an upward trend in the use of USS in A&E departments due to the current drive for USS by non-radiologists [7-9]. Our study found that only a quarter of UK units used USS although these included at least half of each of the large and very large departments.
Radiological investigations confirm or refute a diagnosis of renal colic. If the diagnosis is refuted, then the clinician is prompted to consider other diagnoses. We found that over a quarter of departments (27.4%; n = 29) did not perform any radiological investigation (see Table 2). This is of concern since it has been reported that renal colic is one of the most common misdiagnoses in catastrophic abdominal conditions including ectopic pregnancy and abdominal aortic aneurysm [10]. The concern is greatest for those departments in the small category; nearly 60% of them did not routinely perform any radiological investigations and they may be located in remote areas lacking specialist surgical facilities such as on-site vascular surgery. When departments are isolated with minimal specialist back up, an early diagnosis is crucial, so that patients with other abdominal conditions requiring urgent specialist management can be appropriately referred. An IVU can be easily done in the X-ray department; a negative IVU should prompt the clinician to consider an alternative diagnosis to renal colic and this, in our opinion should not be beyond the reach of any A&E department in the UK.
Intra-venous urography was performed by a significantly greater proportion of those departments with X-ray facilities within the unit compared to those with X-ray facilities located elsewhere. This would suggest that if all A&E departments had X-ray facilities located within the unit, the potential for misdiagnosis would be minimised since all units would then be more likely to perform at least an IVU.
Those hospitals that had on-site urology services performed more investigations than those sites without (see Table 1). In particular nearly half of those hospitals with urology services located outside the hospital did not perform any investigations at all compared to 14% of those hospitals with on-site services. This potentially means that patients with conditions other than renal colic are sent to a urology clinic with the consequence that their management is delayed.
We found that when an IVU was performed, the larger units used fewer films although this relationship just missed statistical significance (see Table 2). This finding suggests that adequate information to diagnose renal colic might be obtained by using only one film, in keeping with previous findings [11]. However, these suggestions need to be verified by further research.
We found that less than 10% of UK A&E departments use a CT scan in the assessment of renal colic and the association with departmental workload just missed statistical significance at the 5% level (see Table 2). A CT scan was available to 37.5% of the very large departments compared to less than 10% in the other sized categories. The main difficulty with performing a CT scan in an A&E setting is that interpretation of the images requires urologists or radiologists who are not always available [5]. When appropriate personnel are available, CT should be the favoured investigation as it has been shown not only to diagnose urinary tract calculi accurately but also provide other diagnoses. The choice of investigation in some of the units that reported using more than one type of radiological investigation may have been influenced by availability of the required personnel, as both USS and CT require some expertise. However, this study did not investigate this aspect of practice.
Previous research has shown the efficacy of NSAIDS in renal colic [12-17]. We found that 85.8% of UK A&E departments use NSAIDS. Intra-venous ketorolac reported to have the fastest onset of action and equal analgesic properties to other NSAIDS, was used by only 4.7% of units although its use may have been precluded by difficulty with venous access. Intra-muscular diclofenac was routinely used in 64% of departments despite the problems associated with this route including discomfort at the injection site and the potential for sterile abscess formation [16]. Whilst the rectal route is favoured over the intra-muscular route since it is equally effective and avoids possible injection site problems, only 16% of all departments in this study reported using it.
In spite of the proven efficacy of NSAIDS, we found that nearly 10% of all A&E departments used parenteral opiates as the analgesic of first choice. Given that opiate administration requires checking and crosschecking by at least two nurses, there will inevitably be a delay in relieving the patients' pain. Parenteral opiates would be better as second-choice analgesic in our opinion.
We found five departments, including four in the small and one in the large workload categories that did not use either NSAIDs or parenteral opiates in suspected renal colic. The large and one of the small departments prescribed Codydramol. The other three small departments referred renal colic patients directly to a urology team off-site without even offering analgesia. Why these departments adopted this approach was unclear. Nonetheless, these findings were of concern since an A&E department would be expected to at least consider offering analgesia to patients that present to them in pain irrespective of the patients' final destination.
There is no reason to suspect that the departments in this study were not representative of A&E units in the UK, since the sample was derived from each of the workload categories in the 11 UK Deanery regions. However, as with any questionnaire study it is difficult to assess the reliability of the answers provided. The most senior individual in the department was invited to complete the questionnaire. However, it is not possible to quantify the bias, if any, that may be introduced by the variation in grade of the respondents. Since nearly three quarters of the questionnaires were completed by consultants one might expect that that the results were reliable. However, it is possible that consultants may not be fully aware of routine practice and therefore the information provided could be inaccurate. Furthermore, as soon as a unit was asked about its current practice through the questionnaire, it may subsequently have adjusted it, particularly if it was sub-optimal. Therefore, the information provided on the questionnaire may reflect the altered, rather than original practice.
Conclusions
The management of acute renal colic differs between A&E departments in the UK. Local factors may contribute to these differences. The total number of radiological procedures that were available to a unit was positively associated with departmental workload. Of great concern was that a significant proportion of departments overall (27.4%) did not perform any radiological investigation. The concern is greatest for those departments in the small category with nearly 60% performing no radiological investigation. Location of X-ray facilities within the A&E premises is associated with whether an IVU is ever performed. Departments with on-site urology have a greater range of radiological investigations to choose from. Furthermore, the very large units tended to routinely use fewer films per IVU, with a median number of one compared to three in all other smaller units.
The first choice analgesic used by most units is NSAIDS in keeping with the literature; more departments, however, need to adopt the use of the rectal route for diclofenac in order to avoid the potential complication of the intra-muscular route. The low percentage of departments using parenteral opiates as first-choice analgesic is encouraging as parenteral opiates are better used as second choice in view of the unavoidable delay that occurs before their administration.
The practice in over a quarter of A&E departments in the UK is below standard. There is significant association with departmental workload and location of services such as radiology and Urology relative to A&E. We suggest that national guidelines be developed for the management of acute renal colic in A&E departments to ensure optimum care for all patients. Subsequent to the implementation of any guidelines, we suggest that UK practice is regularly reviewed.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TAL undertook the literature search, designed the questionnaire, participated in the collection and analysis of data, and wrote the paper. PMS undertook the statistical analysis and contributed to the writing of the paper. NN initiated the research, participated in data collection and contributed to the paper. CS performed the initial stratification and selection of units for the study. NP participated in data collection.
Appendix 1
Questionnaire on renal colic
1. Name of hospital: ____________________________________________
2. Where is X-ray located? Within or adjacent to A&E □
Distant □
3. Where are urology services located? Same Site as A&E □
Separate Site from A&E □
4. Are the following investigations performed on suspected cases of renal colic?
a) Urinalysis No □ Yes □
b) IVU No □ Yes □
If IVU is performed, please indicate how many films are used □
c) CT No □ Yes □
d) USS No □ Yes □
e) Nuclear Medicine No □ Yes □
5. Which of the following analgesics are given on presentation?
a) Codydramol Other Oral No □ Yes □
b) NSAIDS No □ Yes □
If NSAIDS used; which one: (indicate route below)
i. Intra-muscular □
ii. Oral □
iii. Rectal □
iv. Intra-venous □
c) Parenteral opiate No □ Yes □
If parenteral opiates are used then please indicate if first or second choice:
i. First choice □
ii. Second choice □
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
British Association for Accident and Emergency Handbook of Directory 1999/2000 1999 London BOE 105 224
Bland M An Introduction to Medical Statistics 2000 13 Third Oxford: Oxford University Press 230 255
Altman DG Practical Statistics for Medical Research 1993 9 First London: Chapman & Hall 179 233
Smith I Urograghy during renal colic Brit J Surg 1966 53 93 102 4952333
Greenwell TJ Woodhams S Denton ER Mackenzie A Rankin SC Popert R One year's experience with unenhanced spiral computed tomography for the assessment of acute loin pain suggestive of renal colic BJU 2000 85 632 636 10.1046/j.1464-410x.2000.00605.x
Shokeir AA Mahran MR Abdulmaaboud M Renal colic in pregnant women: role of renal resistive index Urology 2000 55 344 347 10699607 10.1016/S0090-4295(99)00475-6
Burnett HC Nicholson DA Current and future role of ultrasound in the Emergency department J Accid Emerg Med 1999 16 250 254 10417929
Robinson NA Clancy MJ Should UK Emergency physicians undertake diagnostic ultrasound examination? J Accid Emerg Med 1999 16 248 249 10417928
Thompson P Brown R Sidhu P Bedside emergency department ultrasound: how to introduce it J Accid Emerg Med 1998 15 413
Marston WA Ahlquist R Johnson G JrMeyer AA Misdiagnosis of ruptured abdominal aortic aneurysms J Vasc Surg 1992 16 17 22 1619721 10.1067/mva.1992.34344
Murali K Nayeem N Single exposure intravenous urograghy in the accident and emergency department-is it worthwhile? J Accid Emerg Med 1998 15 413
Heller MB Emergency management of acute pain: new option and strategies Postgrad Med 1992 3 39 46 1508781
Larson LS Miller A Allegra JR The use of intravenous ketorolac for the treatment of renal colic in the Emergency department Am J Emerg Med 1993 11 197 199 8489656 10.1016/0735-6757(93)90123-S
Di-Trolio RN Sing RF Bates GM Use of ketorolac in renal colic J Am Osteopath Assoc 1999 11 589 90 10612955
Cohen E Hafner R Retenberg Z Fadilla M Garty M Comparison of ketorolac and diclofenac in the treatment of renal colic Eur J Clin Pharmacol 1998 54 455 458 9776434 10.1007/s002280050492
Tramer MR Williams JE Carroll D Wiffen PJ Moore RA McQuay HJ Comparing the efficacy of non-steroidal anti-inflammatory drugs given by different routes in acute and chronic pain: a qualitative systematic review Acta Anaesthesiol Scand 1998 42 71 79 9527748
Holdgate A Pollock T Nonsteroidal anti-inflammatory drugs (NSAIDs) versus opiods for acute renal colic (Cochrane Review) The Cochrane Library 2004 Chichester, UK: John Wiley & Sons Ltd
| 15585056 | PMC539300 | CC BY | 2021-01-04 16:31:01 | no | BMC Emerg Med. 2004 Dec 7; 4:5 | utf-8 | BMC Emerg Med | 2,004 | 10.1186/1471-227X-4-5 | oa_comm |
==== Front
BMC Pregnancy ChildbirthBMC Pregnancy and Childbirth1471-2393BioMed Central London 1471-2393-4-241558830310.1186/1471-2393-4-24Study ProtocolProtocol for a randomised controlled trial of a decision aid for the management of pain in labour and childbirth [ISRCTN52287533] Roberts Christine L [email protected] Camille H [email protected] Natasha [email protected] Lyndal [email protected] Kirsten [email protected] Centre for Perinatal Health Services Research, QEII Building DO2, University of Sydney, NSW 2006, Australia2 School of Public Health, Edward Ford Building A27, University of Sydney NSW 2006, Australia2004 9 12 2004 4 24 24 10 11 2004 9 12 2004 Copyright © 2004 Roberts et al; licensee BioMed Central Ltd.2004Roberts 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
Women report fear of pain in childbirth and often lack complete information on analgesic options prior to labour. Preferences for pain relief should be discussed before labour begins. A woman's antepartum decision to use pain relief is likely influenced by her cultural background, friends, family, the media, literature and her antenatal caregivers. Pregnant women report that information about analgesia was most commonly derived from hearsay and least commonly from health professionals. Decision aids are emerging as a promising tool to assist practitioners and their patients in evidence-based decision making.
Decision aids are designed to assist patients and their doctors in making informed decisions using information that is unbiased and based on high quality research evidence. Decision aids are non-directive in the sense that they do not aim to steer the user towards any one option, but rather to support decision making which is informed and consistent with personal values.
Methods/design
We aim to evaluate the effectiveness of a Pain Relief for Labour decision aid, with and without an audio-component, compared to a pamphlet in a three-arm randomised controlled trial. Approximately 600 women expecting their first baby and planning a vaginal birth will be recruited for the trial.
The primary outcomes of the study are decisional conflict (uncertainty about a course of action), knowledge, anxiety and satisfaction with decision-making and will be assessed using self-administered questionnaires. The decision aid is not intended to influence the type of analgesia used during labour, however we will monitor health service utilisation rates and maternal and perinatal outcomes. This study is funded by a competitive peer-reviewed grant from the Australian National Health and Medical Research Council (No. 253635).
Discussion
The Pain Relief for Labour decision aid was developed using the Ottawa Decision Support Framework and systematic reviews of the evidence about the benefits and risks of the non-pharmacological and pharmacological methods of pain relief for labour. It comprises a workbook and worksheet and has been developed in two forms – with and without an audio-component (compact disc). The format allows women to take the decision aid home and discuss it with their partner.
==== Body
Background
Patient participation in clinical decision making
Making evidence-based decisions in clinical practice is not always straightforward: patients and their healthcare providers may need to weigh up the evidence between several comparable options, the evidence for some treatments may be inconclusive, and the information needs to be tailored to each patient's clinical context and personal preferences [1,2]. Good medical decision making should take into account the best available evidence, along with patients' preferences and values [3]. However, finding effective and efficient mechanisms for doing this in the clinical setting is a challenge.
To assist patients and their doctors in making informed decisions, information must be unbiased and based on current, high quality, quantitative research evidence. However, patient information materials are often outdated, inaccurate, omit relevant data, fail to give a balanced view and ignore uncertainties and scientific controversies [4,5]. It is increasingly evident that the provision of patient and provider information alone, even if evidence-based, is not sufficient to influence health outcomes and behaviour [6]. It is only when mechanisms are provided that tailor this information to the individual patient that health outcomes, related to treatment decisions, are positively effected [7]. With this in mind, decision aids are emerging as a promising tool to assist practitioners and their patients in evidence-based decision making [1].
Decision Aids
Decision aids are "interventions designed to help people make specific and deliberative choices among options by providing (at minimum) information on the options and outcomes relevant to the person's health status" [1]. Additional strategies may include providing: information on the condition; the probabilities of outcomes tailored to a person's health risk factors; an explicit values clarification exercise; examples of others' decisions; and guidance in the steps of decision making [1]. Decision aids are non-directive in the sense that they do not aim to steer the user towards any one option, but rather to support decision making which is informed, consistent with personal values and acted upon [1]. Decision aids have been found to improve patient knowledge and create more realistic expectations, to reduce decisional conflict (uncertainty about the course of action) and to stimulate patients to be more active in decision making without increasing anxiety [1].
Internationally decision aids have been evaluated in a variety of health and clinical settings. Although their use in pregnancy and birth has only just begun to be explored, this is an area in which consumers are known to want to participate actively in decision making [8]. A survey of 790 Australian women reported a tenfold increase in dissatisfaction among women who did not have an active say in decisions about pregnancy care [8]. Similarly in the UK, women rated the explanation of procedures, including the risks, before they are carried out and involvement in decision making as most important to satisfaction with care [9]. Significantly, neither obstetricians nor midwives appreciated the importance to women of "being told the major risks for each procedure" [9]. Our own survey of pregnant women attending an antenatal clinic found that overwhelmingly women wanted to be involved in decisions regarding their pregnancy care, and this was regardless of age, parity, education or delivery preferences [10].
Labour pain
The pain of labour is a central part of women's experience of childbirth and is a constant feature of antenatal discussion groups [11]. Most women giving birth use some methods of pain relief (pharmacologic and/or non-pharmacologic) during labour. In Australia 92% of primiparas and 71% of multiparas use some analgesic agents for labour analgesia [12]. Significantly, there have been more clinical trials of pharmacological pain relief during labour and childbirth than of any other intervention in the perinatal field [13].
However satisfaction with childbirth is not necessarily contingent upon the absence of pain [14]. Many women are willing to experience pain in childbirth but do not want pain to overwhelm them. The Royal College of Obstetrics and Gynaecology (RCOG) makes the following evidence-based recommendations [15]:
• Continuous caregiver support for a single individual should be available to women in labour
• Midwives must involve women in decisions about analgesia and recognise the value of promoting personal control
• Maternity services should ensure access to written and verbal information on pain relief and should support women in their choices for pain relief
• Maternity services should respect women's wishes to have some control over their pain relief
• Improved public information and data on pain and analgesia
In Australia over 250,000 women give birth annually and the increasing use of epidural analgesia means some 75,000 women have an epidural in labour each year [16]. Among primiparas in NSW, the epidural rate increased from 25% in 1990 to 42% in 2000, but was as high as 74% in hospitals with greater availability of epidurals [12]. Other pharmacologic methods of pain relief for primiparas include 36% opioids and 55% nitrous oxide [12].
Pharmacologic methods of pain relief in labour and childbirth
Randomised controlled trials have shown epidural analgesia provides the most efficacious pain relief for labour, but the adverse consequences include prolonged labour, restricted mobility, use of oxytocin augmentation and an increased incidence of instrumental delivery [17,18]. Consequences of instrumental delivery at 6 months postpartum include perineal pain 54%, urinary incontinence 18%, bowel problems 19%, haemorrhoids 36% and sexual problems 39% [19]. Further, the complications of epidurals can include unsatisfactory analgesia, dural-puncture headache, hypotension, nausea/vomiting, fever, localised backache, shivering, pruritis and urinary retention [18].
Although not as effective as epidural, randomised trials show inhalational analgesia (e.g. 50% nitrous oxide in oxygen) and systemic opioid analgesics (e.g. pethidine) can provide modest benefit to some patients during labour or supplement an unsatisfactory epidural [13]. Both these methods can cause nausea, vomiting and dizziness, and additionally opioid side-effects may include orthostatic hypotension, delayed stomach emptying and respiratory depression in the baby [13].
Non-pharmacologic methods of pain relief in labour and childbirth
A number of women prefer to avoid pharmacological analgesia if possible [20]. The wish to maintain personal control during labour and birth, the desire to participate fully in the experience, and concerns about untoward effects of medications during labour, are among the factors that influence their attitude [20]. Non-pharmacological methods of pain relief include maternal movement and position changes, superficial heat and cold, immersion in water*, massage, acupuncture/acupressure, transcutaneous electrical nerve stimulation (TENS)*, aromatherapy, attention focussing, hypnosis*, music/audioanalgesia* and continuous caregiver support*. Only a few of these methods (marked*) have been assessed in randomised trials [20-22]. Only continuous caregiver support resulted in reduced analgesia requirements (and length of labour and the incidence of operative delivery). Although the other interventions trialled did not reduce the use of pharmacologic analgesia, they were well liked by women and had few side effects.
Decision making and pain in labour
Women report fear of pain in childbirth and often lack complete information on analgesic options prior to labour [11]. For example a Royal Australian and New Zealand College of Obstetrics and Gynaecology brochure on 'Epidural and Spinal Anaesthesia' reports the advantages of epidurals but does not mention any possible adverse outcomes or complications [23]. While written informed consent is required for epidural analgesia, it is not required for other analgesic options. Further, the consent for epidural (covering only the procedure and complications) is obtained by the anaesthetist at the time of the procedure – by which time most women are already distressed [24].
Dickerson stresses the importance of discussing preferences for pain relief before labour begins [13]. A woman's antepartum decision to use pain relief is likely influenced by her cultural background, friends, family, the media, literature and her antenatal caregivers [25]. A survey of Australian women found that antepartum information about analgesia was most commonly derived from hearsay and least commonly from health professionals [26]. Antenatally 82% of women wish to see how labour progresses and only want analgesia when pain becomes severe or intolerable [14]. Antenatal plans for analgesia are strongly associated with use: 96% of women who definitely planned to have an epidural, received one [25].
The management of pain in labour is a clinical decision that fulfils Eddy's criteria for a decision in which patients' values and preferences should be included [2]. The outcomes for analgesia options and, women's preferences for the relative value of benefits compared to risks are variable and could result in decisional conflict. For such a clinical decision, a decision aid would be expected to improve patient knowledge and create realistic expectations, to reduce decisional conflict and to stimulate patients to be more active in decision making without increasing anxiety [1]. Leap has suggested a 'working with pain' framework for managing labour and childbirth in a positive context [11]. This framework which aims to develop an understanding of 'normal pain' as part of the process of labour, rather than the absolute amelioration of pain, has been recommended by the Royal College of Obstetrics and Gynaecology.
Development of a decision aid on the management of pain during labour
During 2003 and 2004, we developed an evidence-based decision aid about the management of pain in labour for women having their first baby. This followed a needs assessment that collected data on the attitudes, preferences and knowledge of nulliparous women who were making plans about pain relief for labour and childbirth. The needs assessment found that women's knowledge of pain relief options was limited and these women would benefit from a decision aid for labour analgesia.
In developing the decision aid we utilised the NHMRC guideline "How to prepare and present information for consumers of health services" [27] and the Ottawa framework established and rigorously tested by the Ottawa Health Decision Center [28]. The decision aid was developed to incorporate a workbook (with and without a complementary audio-component as a compact disc) and worksheet. The workbook highlights key points (similar to a slide presentation) and the audio component connects these points in a narrative format, providing more detail than the workbook. The worksheet is a one-page sheet to be completed by the woman to record her decision making steps, to list any questions she needs answered before deciding, and to encourage her to discuss he plans with her labour care providers. Most importantly, the decision aid is intended to be non-directive in that it does not aim to steer the user towards any one option or increase or decrease intervention rates but rather act as an adjunct to care
The decision aid was designed for women to use at home or in the clinical setting, and takes about 30 minutes to complete. After working through the decision aid, women should take the completed worksheet to their next antenatal appointment to discuss their preferences with their health care provider. The worksheet is also useful for the practitioner, who can see rapidly from it what evidence the patient has considered, what her values and preferences are and which way she is leaning in her preferences for analgesia during labour.
The decision aid was developed, pilot tested and revised with extensive consumer involvement, as outlined in the NHMRC guideline on preparing information for consumers [27]. The content of the decision aid was largely driven by consumers' questions and information needs as determined from the focus groups and from the process of drafting, pilot testing and re-drafting.
A number of draft decision aids (including workbook, audio transcript, and worksheet), were developed and each subjected to pilot testing and revision as we obtained feedback. The process of testing and revising started with the study project group. The next phase included a review by a group of national and international content experts, including decision aid experts, obstetricians, midwives, perinatal epidemiologists, parent educators and psychologists. Once we were convinced that the content was accurate the decision aid was pilot-tested amongst consumers. There were several rounds of consumer review and refinement.
Initially we aimed to compare the Decision Aid (workbook and audio-component) with usual care and counselling however preliminary work led us to alter our original study design. We could find no studies that compared Decision Aids with and without an audio-component. As the audio-component adds considerable complexity to the development and cost of the Decision Aid we decided to have two intervention arms: a Decision Aid with an audio-component and a Decision Aid without an audio-component. Further in pilot testing we found that women in the usual care arm were disappointed to not receive any information. Thus, to minimise refusals and losses to follow-up we decided to issue the women in the control group with a pamphlet called "Pain relief during childbirth – A guide for women" This pamphlet is published by the Royal Australian and New Zealand College of Obstetricians and Gynaecologists, is publicly available and includes information about methods of pain relief during labour [29]. These changes to the study protocol were approved by the institutional ethics committee prior to commencement of the trial.
Methods/design
1. Specific Aim
To compare the relative effectiveness of the Pain Relief for Labour Decision Aid with a pamphlet on women's decisional conflict, knowledge, expectations, satisfaction with decision making and anxiety, and examine its impact on service utilisation and perinatal outcomes (as secondary outcomes).
2. Hypotheses
The primary study hypotheses are:
Use of the Pain Relief for Labour Decision Aid by women expecting their first baby:
1. Reduces decisional conflict (uncertainty about the course of action)
2. Increases knowledge of labour analgesia
3. Increases satisfaction with their decision making
4. Reduces anxiety.
The secondary hypotheses of the study are:
Use of the Pain Relief for Labour Decision Aid by women expecting their first baby will not influence:
1. The type of analgesia women use for labour
2. Maternal and infant outcomes.
3. Study design
We will conduct a randomised trial with the following study groups to assess the impact of the decision aid:
Group 1: The pamphlet, "Pain relief during childbirth – A guide for women" [29]
Group 2: Decision aid with an audio-component
Group 3: Decision aid without an audio-component
4. Setting
An Australian tertiary obstetric hospital with a full range of non-drug and anaesthetic options for pain relief in labour. Epidurals are available 24 hours a day from anaesthetic staff designated to labour ward. All forms of antenatal care (clinic, birth centre, private, shared care with a family physician) will be included in the study.
5. Participants/eligibility criteria
Primiparous women in late pregnancy (≥36 weeks gestation) who are expecting to have a vaginal birth of a single infant will be eligible for the study. Primiparous women were selected because previous pregnancies have a strong impact on decision making and analgesia use in labour [14,16]. Exclusions include women who will not have any choice about analgesia, for example planned caesarean section (eg breech, placenta praevia, HIV), planned epidural (eg symptomatic heart disease), contraindications to analgesia (e.g drug sensitivities, anticoagulants, thrombocytopaenia). The decision aid was produced in English and designed to be simple and accessible for women with low levels of literacy.
6. Procedures, recruitment, randomisation and collection of baseline data
The study procedure draws on the usual schedule of weekly antenatal visits in late pregnancy (Figure 1). We plan a pragmatic approach to assess the decision aid under the conditions most likely to be applied in practice. A research nurse will ask eligible women to participate, explain the trial and obtain informed consent, collect baseline data and randomly allocate women (using telephone randomisation) to one of the study groups. This is only a minor deviation from current practice. As women of child-bearing age are known to be very mobile, participants will be asked to provide alternate contact details (eg friend or relative) to enhance subsequent follow-up. Private obstetricians will be asked to offer participation in the study to their patients. Those interested will be requested to come to the antenatal clinic for recruitment and randomisation. The private obstetrician will provide standard care. Flyers and posters will be prepared to inform women of the study and will be distributed through family physicians and obstetricians as well as the clinics.
Figure 1 Schema of Pain Relief for Labour Decision Aid trial
Brief baseline data will be collected to assess comparability of the study groups. The baseline assessment will include age, brief socio-demographic data, highest level of education achieved, anxiety as assessed by the state component of the short Spielberger anxiety scale [30], and information sources about labour analgesia.
7 Intervention
The aim of the decision aid is to assist preference elicitation, and not to influence the direction of the decision taken. Women in each study group will be given the opportunity to review the intervention they are allocated (decision aid or pamphlet) while in the antenatal clinic and/or to take home, which ever is most convenient. Many women will also want to discuss their preferences with their partner. At the next antenatal visit, women will be contacted by the research nurse to discuss the information materials and any questions they may have had.
8 Follow-up
i) First follow-up questionnaire
All participants will be given a follow-up questionnaire prior to their next antenatal consultation. (See Outcome Measure details below).
ii) Midwife questionnaire
After a study participant delivers, the midwife who provided the labour care will complete a brief questionnaire to assess the impact of the decision aid on the management of labour analgesia. Information will also be collected on caregiver support in labour, birthplace (delivery suite or birth centre), use of non-drug analgesic options and stage of labour at admission.
iii) Second follow-up questionnaire
At 12–16 weeks postpartum all participants will be mailed a second follow-up questionnaire. This will assess women's satisfaction with the decisions made and the decision-making processes. (See Outcome Measures below). Questionnaires will be mailed with reply paid envelopes, with up to two reminder telephone prompts to non-responders.
iv) Qualitative follow-up
We will conduct in-depth interviews to explore the impact of the decision aid on women's experiences in labour and childbirth. A sub-sample of 30 women will be purposively selected, to reflect heterogeneity of experience of labour. The interviews will provide an understanding of the complexities of analgesic preferences, management, expectations, satisfaction, and psychological health following delivery. This data will enable examination of unpredicted and subtle effects of the decision aid on psychosocial outcomes that may not be captured using quantitative methods. Interviews will be face-to-face and conducted in women's homes or at a clinic, according to participants' preferences. Interviews will be recorded and transcribed. Data will be analysed using thematic analysis.
9. Blinding and contamination
As with many obstetric interventions blinding is virtually impossible. The main outcomes of this study are self-reported and the women are clearly not blinded to their treatment allocation. However, we will institute a number of measures aimed at keeping antenatal staff blind to the treatment allocation and preventing contamination of the control group:
• Women will review the decision aid with the research nurse and complete the first questionnaire (primary outcome measures) prior to their next antenatal consultation
• Usual antenatal care providers will be blinded to the exact content and format of the decision aid
• Regular in-service (educational training) for the antenatal care providers to explain the trial protocol and to make clear the potential effect of unmasking or contamination.
• Monitoring decision aid distribution and keeping them locked up and only accessible by the research nurse
• Asking participants not to reveal their treatment allocation, or share their decision aid material with antenatal staff or other women. If participants do not want to keep their decision aid they will be asked to return it.
10 outcome measures
Primary outcomes
The primary outcomes of this study will be:
Decisional conflict (uncertainty about which preference to choose) will be assessed by the Decisional Conflict Scale which has established reliability, good psychometric properties and is short (16 items) [31]. It has been used to evaluate a range of decision aids [1].
Measures of knowledge and realistic expectations about labour analgesia options and the benefits and risks of these options will be specific to this project. Thus we will need to develop, and test these measures as part of the project.
Anxiety will be measured by the state component of the short Spielberger anxiety scale which has been extensively used and validated [30,32]. We do not anticipate the decision aid will increase women's anxiety but it is important to document any changes in anxiety associated with the decision aid.
Satisfaction with analgesia decisions will be assessed using the Satisfaction with Decision Scale – a very brief six item scale with high reliability was developed specifically to assess satisfaction with health care decisions [33].
Satisfaction with the decision and anxiety will be measured again at 12–16 weeks postpartum. This interval was chosen to avoid the potential bias arising from questioning women still in the hospital who may feel a disloyalty to their caregivers by a critical appraisal and whose opinions have been shown to be more positive and short-lived than those obtained further out from the birth itself [34]. At that time we will also ask about exposure to the decision aid (to assess contamination), support during labour and use of pain relief methods prior to hospital admission. These issues will be further explored in the sample selected for in-depth interview.
Secondary outcomes
Service utilisation outcomes
The aim of the decision aid is to assist preference elicitation, and not to influence the direction of the decisions taken. Nevertheless, it is important to collect service utilisation and pregnancy outcome data so we will record and compare the pain relief methods used by women in all arms of the study, as well as recording and comparing rates of pregnancy complications and perinatal outcomes. The latter will be obtained (with informed consent) from the existing computerised obstetric database and include: medical or obstetric complications, induction or augmentation of labour, mode of delivery (vaginal, emergency or planned CS), enrolment to delivery interval, gestational age, birthweight, Apgar scores, perinatal deaths, Neonatal Intensive Care Unit admission and length of stay.
11 statistical issues
Sample size
The planned sample size is 600 women, with approximately 200 women to be recruited to each arm of the trial. Based on data for 2001 from the tertiary obstetric hospital where the study will be conducted, about 1500 primiparous women give birth to singleton infants after 36 weeks gestation and 92% use some form of analgesia. We anticipate that at least 50% of women will be both eligible and willing to participate.
The sample size calculations for the trial (significance 0.05, power 0.8) are based on the mean difference in the decisional conflict scale between any two arms of the trial. The effect of decision aids on this scale is documented and effect size data are available [1]. Meta-analysis of four randomised controlled trials comparing a decision aid to a pamphlet and that report a mean difference in decisional conflict gives a pooled mean difference of -4.35, 95%CI -6.8, -1.9 (on a scale ranging from 0 lowest to 100 highest decisional conflict; median standard deviation 13.0) [35-38]. Assuming a mean difference of -4.35 and standard deviation 13.0, we will need about 141 women in each arm of the trial to demonstrate a difference in decisional conflict.
Approximately 20% of primiparous women have a caesarean section (6% before labour and 14% after labour has commenced) [12]. Some of these women will lose their options for analgesia, although some may have extensive use of analgesic agents prior to caesarean section (CS). We plan to conduct an a priori sub-group analysis that excludes women who lose their options for analgesia (defined as a CS planned after randomisation, an emergency CS within 1 hour of arriving in labour or those who receive a therapeutic epidural) as these women may have different satisfaction, anxiety and decisional conflict outcomes. We will inflate the sample size estimate by 20% (from 141 to 169) to ensure sufficient power in the sub-group analyses. A further inflation of 15% for loss to follow-up, gives the final sample size of at least 195 women in each arm of the trial.
If there are no significant differences in outcome for the two decision aid groups (with or without the audio-component), the decision aid groups will be pooled giving two women with the intervention for each woman in the pamphlet group thereby increasing the power to detect differences between the decision aid and the pamphlet.
Data analysis
Analyses will be by intention to treat, including withdrawals and losses to follow-up firstly of all women randomised and then excluding women who lose their options for analgesia. Study groups will be compared in terms of baseline characteristics. As this is a randomised trial, we would anticipate minimal differences in baseline characteristics. If however, important differences are found, these potential confounders will be adjusted for in the analysis of outcomes. For the primary outcomes, the mean score for each measure for each group will be compared using t-tests. If adjustment for confounders is needed a multiple linear regression model will be used. The secondary outcomes will be compared using chi-squared tests of significance for categorical data and t-tests for continuous data. If adjustment for confounding is necessary logistic regression and multiple linear regression will be used respectively.
12 Ethical considerations
This work involves the development of a decision aid for the management of pain in labour and childbirth. Women must decide between a range of non-pharmacological and pharmacologic methods of pain relief. However this decision must be made in the context of the likely analgesic effects of each option, the risk of complications and adverse obstetric effects, and maternal preference for relief of pain. There are currently no evidenced based materials available. We therefore expect this project to be beneficial for participating women. A systematic review of decision aids found they improved knowledge without increasing anxiety. Nevertheless we will measure anxiety levels at baseline and follow-up to document any adverse effects. A trained research nurse will interview all women and obtain written informed consent. Women will be encouraged to discuss any concerns/anxiety with the research nurse and/or with their usual antenatal care provider. Women will be reassured that they are able to withdraw from the study at any time with no adverse effects on their pregnancy management. Participation will require women to complete self-report questionnaires during and after pregnancy. Working through the decision aid will take approximately 30 minutes and review of their preferences or outstanding questions will be at a routine antenatal visit. Therefore we do not consider this to be an excessive burden on their time.
The study has been approved by the Central Sydney Area Health Service Ethics Review Committee (Protocol no. X02-0247) and the University of Sydney Human Ethics Committee (Ref No. 3419). This project is funded by a nationally competitive peer-reviewed grant from the Australian National Health and Medical Research Council (No. 253635).
13 Confidentiality and data security
Participants in the trial will be identified by a study number only, with a master code sheet linking names with numbers being held securely and separately from the study data. To ensure that all information is secure, data records will be kept in a secure location at the University of Sydney and accessible only to research staff. As soon as all follow-up is completed the data records will be de-identified. De-identified data will be used for the statistical analysis and all publications will include only aggregated data. The electronic version of the data will be maintained on a computer protected by password. All hard copy patient identifiable data and electronic backup files will be kept in locked cabinets, which are held in a locked room accessed only by security code and limited staff. Data files will be stored for seven years after completion of the project as recommended by the NHMRC. Disposal of identifiable information will be done through the use of designated bags and/or a shredding machine.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CR, CRG, LT and KM were involved in the conception and design of the study. CR, NN and CRG were responsible for the drafting of the protocol. All authors have read and given final approval of the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study is funded by an Australian National Health and Medical Research Council project grant (253635).
Christine Roberts is funded by an Australian National Health and Medical Research Council Public Health Practitioner Fellowship (245501).
Camille Raynes-Greenow and Natasha Nassar are both funded by Australian National Health and Medical Research Council Public Health Postgraduate Research Scholarships.
==== Refs
O'Connor AM Stacey D Entwistle V Llewellyn-Thomas H Rovner D Holmes-Rovner M Tait V Tetroe J Fiset V Barry M Jones J Decision aids for people facing health treatment or screening decisions Cochrane Library, Issue 4 2002
Eddy DM A manual for assessing health practices and designing practice policies: the explicit approach 1992 Philadelphia, American College of Physicians
National Health and Medical Research Council A Guide to the Development, Implementation and Evaluation of Clinical Practice Guidelines. 1999
Coulter A Evidence based patient information. is important, so there needs to be a national strategy to ensure it BMJ 1998 317 225 226 9677206
Coulter A Entwistle V Gilbert D Sharing decisions with patients: is the information good enough? BMJ 1999 318 318 322 9924064
Bekker H Thornton JG Airey CM Connelly JB Hewison J Robinson MB Lilleyman J MacIntosh M Maule AJ Michie S Pearman AD Informed decision making: an annotated bibliography and systematic review Health Technology Assessment 1999 3 1 156 10350446
Edwards A Hood K Matthews E Russell D Russell I Barker J Bloor M Burnard P Covey J Pill R Wilkinson C Stott N The effectiveness of one-to-one risk communication interventions in health care: a systematic review Medical Decision Making 2000 20 290 297 10929851
Brown S Lumley J Satisfaction with care in labor and birth: a survey of 790 Australian women Birth 1994 21 4 13 8155224
Drew NC Salmon P Webb L Mothers', midwives' and obstetricians' views on the features of obstetric care which influence satisfaction with childbirth. Br J Obstet Gynaecol 1989 96 1084 1088 2804012
Raynes-Greenow CH Roberts CL Barratt A Brodrick B Peat B Pregnant women's preferences and knowledge of term breech management, in an Australian setting Midwifery 2004 20 181 187 15177862 10.1016/j.midw.2003.10.002
Leap N Pain in labour: towards a midwifery perspective MIDIRS Midwifery Digest 2000 10 49 53
NSW Department of Health NSW Mothers and Babies 1998 NSW Public Health Bulletin 2000 9 Sydney, NSW Department of Health
Dickersin K Chalmers I, Enkin M and Keirse MJN Pharmacological control of pain during labour Effective care in pregnancy and childbirth 1989 Oxford, Oxford University Press 913 945
Ranta P Spalding M Kangas-Saarela T Jokela R Hollmen A Jouppila P Jouppila R Maternal expectations and experiences of labour pain--options of 1091 Finnish parturients Acta Anaesthesiologica Scandinavica 1995 39 60 66 7725885
Royal College of Obstetricians and Gynaecologists Recommendations arising from the 41st Study Group on Pain in Obstetrics and Gynaecology
Roberts CL Algert CS Douglas ID Tracy SK Peat B Trends in labour and birth interventions among low-risk women in NSW Australian and New Zealand Journal of Obstetrics and Gynaecology 2002 42 165 168
Halpern SH Leighton BL Ohlsson A Barrett JF Rice A Effect of epidural vs parenteral opioid analgesia on the progress of labor: a meta-analysis Jama 1998 280 2105 2110 9875879 10.1001/jama.280.24.2105
Howell CJ Epidural versus non-epidural analgesia for pain relief in labour (Cochrane review) The Cochrane Library, Issue 1 Oxford: Update Software 2004
Brown S Lumley J Maternal health after childbirth: results of an Australian population based survey British Journal of Obstetrics and Gynaecology 1998 105 156 161 9501779
Simkin P Chalmers I, Enkin M and Keirse MJN Non-pharmalogical methods of pain relief during labour Effective care in pregnancy and childbirth 1989 Oxford, Oxford University Press 1182 1195
Nikodem VC Immersion in water in pregnancy, labour and birth. Cochrane Library, Issue 4 2001
Hodnett ED Caregiver support for women during childbirth Cochrane Library, Issue 4 2001
Royal Australian and New Zealand College of Obstetricians and Gynaecologists Epidural and spinal anasthesia. Patient information pamphlet
Swan HD Borshoff DC Informed consent--recall of risk information following epidural analgesia in labour Anaesthesia and Intensive Care 1994 22 139 141 8210014
Goldberg AB Cohen A Lieberman E Nulliparas' preferences for epidural analgesia: their effects on actual use in labor Birth 1999 26 139 143 10655812 10.1046/j.1523-536x.1999.00139.x
Paech MJ Gurrin LC A survey of parturients using epidural analgesia during labour. Considerations relevant to antenatal educators Australian and New Zealand Journal of Obstetrics and Gynaecology 1999 39 21 25 10099742
National Health and Medical Research Council. How to prepare and present information for Consumers of Health Services. 2000
O'Connor AM Fiset V DeGrasse C Graham ID Evans W Stacey D Laupacis A Tugwell P Decision aids for patients considering options affecting cancer outcomes: evidence of efficacy and policy implications Journal of the National Cancer Institute Monographs 1999 67 80 10854460
Royal Australian and New Zealand College of Obstetricians and Gynaecologists Pain relief during childbirth. A guide for women. 2004 Melbourne, Mi-tec Medical Publishing www.mitec.com.au
Spielberger CD State-Trait Anxiety Inventory for Adults (Form Y). 1983 , Palo Alto; Mind Garden
O'Connor AM Validation of a decisional conflict scale. Med Decis Making 1995 15 25 30 7898294
Marteau TM Bekker H The development of a six-item short-form of the state scale of the Spielberger State-Trait Anxiety Inventory (STAI) Br J Clin Psych 1992 31 301 306
Holmes-Rovner M Kroll J Schmitt N Rovner DR Breer ML Rothert ML Padonu G Talarczyk G Patient satisfaction with health care decisions: the satisfaction with decision scale Medical Decision Making 1996 16 58 64 8717600
Lumley J Assessing satisfaction with childbirth Birth 1985 12 141 145 3850745
O'Connor AM Tugwell P Wells GA Elmslie T Jolly E Hollingworth G McPherson R Drake E Hopman W Mackenzie T Randomized trial of a portable, self-administered decision aid for postmenopausal women considering long-term preventive hormone therapy Medical Decision Making 1998 18 295 303 9679994
Goel V Sawka CA Thiel EC Gort EH O'Connor AM Randomized trial of a patient decision aid for choice of surgical treatment for breast cancer Medical Decision Making 2001 21 1 6 11206942
Rothert ML Holmes-Rovner M Rovner D Kroll J Breer L Talarczyk G Schmitt N Padonu G Wills C An educational intervention as decision support for menopausal women Research in Nursing & Health 1997 20 377 387 9334792 10.1002/(SICI)1098-240X(199710)20:5<377::AID-NUR2>3.0.CO;2-L
Tiller K Meiser B Tucker K Andrews L Gaff C Kirk J A randomised trial of a decision aid for women at risk of ovarian cancer 6th World Congress of Psycho-Oncology, Banff, Canada 2003 http://capo.ca/preliminaryprogram2.cfm
| 15588303 | PMC539301 | CC BY | 2021-01-04 16:32:03 | no | BMC Pregnancy Childbirth. 2004 Dec 9; 4:24 | utf-8 | BMC Pregnancy Childbirth | 2,004 | 10.1186/1471-2393-4-24 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-4-901557596210.1186/1471-2407-4-90Case ReportPenile metastasis from primary transitional cell carcinoma of the renal pelvis: first manifestation of systemic spread Pomara Giorgio [email protected] Ilaria [email protected] Maurizio [email protected] Paolo [email protected] Gabriella [email protected] Francesco [email protected] Department of Surgery, Urology Unit, S. Chiara Hospital, Pisa, Italy2 Department of Oncology, S. Chiara Hospital, Pisa, Italy3 Department of Pathological Anatomy, S. Chiara Hospital, Pisa, Italy2004 3 12 2004 4 90 90 29 6 2004 3 12 2004 Copyright © 2004 Pomara 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
Almost one-third of all penile metastases are detected at the same time as a primary tumor, whereas the remaining two-thirds are detected a mean of 18 months after the discovery of the primary tumor. Cutaneous metastasis of transitional cell carcinoma (TCC) is extremely rare and generally accepted as the late manifestation of a systemic spread.
Case presentation
We report the first case of simultaneous penile and lung metastases from a primary TCC of the renal pelvis in a 76-year-old man, that occurred 8 years after a left nephroureterectomy.
Conclusions
This case report underscores the importance of physical examinations of the skin of patients who undergo surgical procedures for TCC from bladder as well as from the upper urinary tract, including those seemingly without metastatic disease, because of the possibility of skin and penile metastatic spread.
==== Body
Background
Approximately 300 cases of penile metastases are reported worldwide [1]. Seventy percent of these metastases have their origin in the genitourinary tract, and the primary tumor is most frequently located in the bladder [2]. We present the first case of penile metastasis from primary transitional cell carcinoma (TCC) of the renal pelvis, and discuss the striking feature of discovering lung metastasis almost simultaneously.
Case presentation
A 76-year-old man with a painful, ulcerative swelling of the glans that had appeared less than 2 months previously was brought to our attention. Physical examination revealed a 3.5-cm fungating lesion involving the upper half of the glans [Fig. 1]. The lesion was malodorous with varying amounts of seropurulent discharge, partially ulcerated, erythematous, and slightly itchy. The superficial inguinal lymph nodes were not palpable. No similar lesions were evident on the head, neck, trunk, or arms. The patient had been treated in another hospital 8 years previously for TCC of the left renal pelvis by nephroureterectomy (pT2 Nx M0 G2). Subsequent postoperative follow-ups (every 6 months for the first 3 years after surgery and then once every year) consisting of a clinical examination, total body CT scan, and regular endoscopic examination showed no evidence of recurrence and the patient did well until he noticed the painful penile nodule. Pathological examination of incisional biopsies confirmed the lesion to be urothelial carcinoma [Fig. 2]. A subsequent CT workup revealed a 4-cm lesion localized in the upper-right pulmonary lobe and enlarged pelvic lymph nodes suggestive of multiple metastases.
The patient began combination chemotherapy (gemcitabine at 1250 mg/m2 on days 1, 8, and 15, and then every 28 days for six courses) and external beam radiotherapy to the mass, which promptly relieved the penile pain. At an 8-month follow-up the patient was still alive with no remarkable changes to the pulmonary and lymph node metastases, or the penile swelling.
Conclusion
Cutaneous metastases from primary TCC of the urinary system are extremely rare and are generally accepted as late manifestations of systemic spread [1]. Spector et al. (1987) suggested that skin metastases in TCC were the result of increased longevity in successfully treated patients [3]. Various mechanisms by which lesions may metastasize to the penis have been reported. Recently Bordeau and Lynch and Berger et al. reported three cases of penile metastases from TCC of the bladder in which they assumed that metastatic spread from primary bladder cancer to the penis occurred mainly via the retrograde venous route [2,4]. There are also a few reports of TCC seeding outside the urinary tract after iatrogenic procedures (i.e., partial cystectomy, suprapubic cystostomy, pyelotomy, and laparoscopy) as the cause of cutaneous metastasis [1,5,6]. Between 5% and 20% of patients with superficial bladder cancer have vascular or lymphatic spread [5], and the reported rate of skin metastasis from TCC of the bladder is between 0.2% and 2% [7]. The rate of skin metastasis from TCC of the renal pelvis is currently unknown, and to the best of our knowledge ours is the first reported case of penile metastasis from TCC of the upper urinary tract.
Almost one-third of all penile metastases are generally detected at the same time as a primary tumor, whereas the remaining two-thirds are detected a mean of 18 months after the discovery of the primary tumor [8]. Because the disease-free survival of this patient was 8 years following the diagnosis of the primary tumor, this case report represents a description of the biology of an unusual tumor. Moreover, the penile metastasis in this case was not the late manifestation of systemic spread, as is usually reported in the literature [1]; in fact, the patient was accurately evaluated with a standard follow-up scheme. Therefore, it is plausible that the lung metastasis appears almost simultaneously with the penile metastasis.
The optimal treatment of penile metastasis requires a multidisciplinary approach that is correlated with the disease extent. The average survival in patients with penile metastasis is 3.9 months from diagnosis and, with extensive surgery and chemotherapy, a survival of 9.2 months has been reported [4].
This patient was not eligible for platinum-based regimens due to impaired renal function (prior nephrectomy) and advanced age. As a result, he was treated with gemcitabine as a single agent. This treatment has demonstrated objective response rates of 25–29% with minimal toxicity as compared to standard regimens [9-11].
In conclusion, this case report suggests that particular attention should be paid to physical examination of the skin of patients who undergo surgical procedures for TCC from bladder as well as from the upper urinary tract, including those seemingly without metastatic disease, because of the possibility of skin and penile metastatic spread.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GP drafted the manuscript and coordinated the co-authors. IP, MS, PC participated in the sequences alignment. GM carried out the histological features of the lesion. FF participated in the sequences alignment and coordinated the co-authors. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgement
The patient gave his approval for publication.
Figures and Tables
Figure 1 Appearance of the metastatic swelling at the corona glandis with erythema and erosion of the glans.
Figure 2 Microscopic appearance of the penile lesion demonstrating a nest of atypical malignant cells (hematoxylin-eosin stain).
==== Refs
Akman Y Cam K Kavak A Alper M Extensive cutaneous metastasis of transitional cell carcinoma of the bladder Int J Urol 2003 10 103 104 12588608 10.1046/j.1442-2042.2003.00571.x
Bordeau KJ Lynch DF Transitional cell carcinoma of the bladder metastatic to the penis Urology 2004 63 981 983 15134996 10.1016/j.urology.2003.12.033
Spector JI Zimbler H De Leo M Ross JS Skin metastasis from transitional cell carcinoma Urology 1987 29 215 217 3811102 10.1016/0090-4295(87)90159-2
Berger AP Rogatsch H Hoeltl L Steiner H Bartsch G Hobisch A Late penile metastasis from primary bladder carcinoma Urology 2003 62 145 12837452 10.1016/S0090-4295(03)00239-5
Miyamoto T Ikehara A Araki M Akaeda T Mihara M Cutaneous metastatic carcinoma of the penis:suspected metastasis implantation from a bladder tumor J Urol 2000 163 1519 10751873 10.1097/00005392-200005000-00030
Andersen JR Steven K Implantation metastasis after laparoscopic biopsy of bladder cancer J Urol 1995 153 1047 7853557 10.1097/00005392-199503001-00038
Zirwas MJ Hunt S Logan TF Abernethy JL Seraly MP A painful cutaneous nodule as the presentation of metastatic transitional cell carcinoma of the renal pelvis J Am Acad Dermatol 2000 42 867 868 10767689 10.1016/S0190-9622(00)90256-1
Maier U Grimm M Transitional cell carcinoma of the bladder with solitary metastasis to the penis 4 years after successful heart transplantation Transplantation 1994 58 861 863 7940726
Stadler WM Kuzel T Roth B Raghavan D Dorr FA Phase II studyof single-agent gemcitabine in previously untreted patients with metastatic urothelial cancer J Clin Oncol 1997 15 3394 3399 9363871
Lorusso V Pollera CF Antimi M Luporini G Gridelli C Frassineti GL Oliva C Pacini M De Lena M A phaseII study of gemcitabine in patients with transitional cell carcinoma of the urinary tract previously treated with platinum Eur J Cancer 1998 34 1208 1212 9849481 10.1016/S0959-8049(98)00030-6
Moore MJ Tannock IF Ernest DS Huan S Murray N Gemcitabine a promising new agent in the treatment of advanced urothelial cancer J Clin Oncol 1997 15 3441 3445 9396395
| 15575962 | PMC539302 | CC BY | 2021-01-04 16:03:02 | no | BMC Cancer. 2004 Dec 3; 4:90 | utf-8 | BMC Cancer | 2,004 | 10.1186/1471-2407-4-90 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-4-931560358910.1186/1471-2407-4-93Research ArticleExpression of inwardly rectifying potassium channels (GIRKs) and beta-adrenergic regulation of breast cancer cell lines Plummer Howard K [email protected] Qiang [email protected] Yavuz [email protected] Hildegard M [email protected] Molecular Cancer Analysis Laboratory, Department of Pathobiology, College of Veterinary Medicine, University of Tennessee, Knoxville, TN 37996-4542, USA2 Experimental Oncology Laboratory, Department of Pathobiology, College of Veterinary Medicine, University of Tennessee, Knoxville, TN 37996-4542, USA3 Current address: Department of Pathology, University of Alabama at Birmingham, Birmingham, AL 35294-0019, USA2004 16 12 2004 4 93 93 21 9 2004 16 12 2004 Copyright © 2004 Plummer 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
Previous research has indicated that at various organ sites there is a subset of adenocarcinomas that is regulated by beta-adrenergic and arachidonic acid-mediated signal transduction pathways. We wished to determine if this regulation exists in breast adenocarcinomas. Expression of mRNA that encodes a G-protein coupled inwardly rectifying potassium channel (GIRK1) has been shown in tissue samples from approximately 40% of primary human breast cancers. Previously, GIRK channels have been associated with beta-adrenergic signaling.
Methods
Breast cancer cell lines were screened for GIRK channels by RT-PCR. Cell cultures of breast cancer cells were treated with beta-adrenergic agonists and antagonists, and changes in gene expression were determined by both relative competitive and real time PCR. Potassium flux was determined by flow cytometry and cell signaling was determined by western blotting.
Results
Breast cancer cell lines MCF-7, MDA-MB-361 MDA-MB 453, and ZR-75-1 expressed mRNA for the GIRK1 channel, while MDA-MB-468 and MDA-MB-435S did not. GIRK4 was expressed in all six breast cancer cell lines, and GIRK2 was expressed in all but ZR-75-1 and MDA-MB-435. Exposure of MDA-MB-453 cells for 6 days to the beta-blocker propranolol (1 μM) increased the GIRK1 mRNA levels and decreased beta2-adrenergic mRNA levels, while treatment for 30 minutes daily for 7 days had no effect. Exposure to a beta-adrenergic agonist and antagonist for 24 hours had no effect on gene expression. The beta adrenergic agonist, formoterol hemifumarate, led to increases in K+ flux into MDA-MB-453 cells, and this increase was inhibited by the GIRK channel inhibitor clozapine. The tobacco carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), a high affinity agonist for beta-adrenergic receptors stimulated activation of Erk 1/2 in MDA-MB-453 cells.
Conclusions
Our data suggests β-adrenergic receptors and GIRK channels may play a role in breast cancer.
==== Body
Background
Breast cancer is the leading cancer in women [1] and estrogen receptor (ER)(-) breast cancers have a poorer prognosis than ER(+) cancers [2,3]. Smoking is a controversial risk factor for the development of these malignancies [4-7]. However, increases in pulmonary metastatic disease and lung cancer have been seen in smokers with breast cancer [8,9]. The tobacco-specific nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) causes cancer of the oral cavity, esophagus, respiratory tract and pancreas, but no breast cancer in laboratory animals [10] and has not been implicated in breast carcinogenesis to date.
Recent studies in human cancer cell lines or in animal models have shown that the growth of adenocarcinomas of the lungs, pancreas and colon are under β-adrenergic control [11-15]. Studies in a cohort of 2442 men found an inverse association between risk of incident adenocarcinomas of the prostate and use of antihypertensive medication, including beta-blockers [16]. The tobacco-specific carcinogenic nitrosamine NNK has recently been identified as a high affinity β-adrenergic agonist that stimulated the growth of pulmonary and pancreatic adenocarcinomas in vitro and in animal models [11,13,15]. The expression of β-adrenergic receptors has been correlated with the over-expression of the arachidonic acid-metabolizing enzymes cyclooxygenase-2 (COX-2) and lipoxygenases (LOX) in adenocarcinomas of lungs [17], colon [18], prostate [19], and pancreas [15]. Inhibitors of these enzymes have been identified as cancer preventive agents in animal models of these cancers [13,20-22]. Collectively, these findings suggest that among the superfamily of adenocarcinomas at various organ sites, there is a subset of malignancies that is regulated by β-adrenergic and arachidonic acid-mediated signal transduction pathways.
The majority of breast cancers are also adenocarcinomas and many of them over express COX-2 and/or LOX [23]. This raises the possibility that comparable to findings in adenocarcinomas of the lungs, pancreas, colon and prostate, a subset of breast cancers may also be under beta-adrenergic control. In support of this hypothesis, studies have demonstrated that three estrogen-responsive and three non-estrogen responsive human cell lines derived from breast adenocarcinomas demonstrated a significant reduction in DNA synthesis in response to beta-blockers or inhibitors of the arachidonic acid-metabolizing enzymes COX-2 and 5-LOX [24]. In addition, analysis by reverse transcription polymerase chain reaction (RT-PCR) revealed expression of β2-adrenergic receptors in all six breast cancer cell lines tested (MDA-MB-361, ZR-75-1, MCF-7, MDA-MB-453, MDA-MB-468, MDA-MB-435S), whereas β1 receptors were not found in two estrogen non-responsive cell lines (MDA-MB-435S, MDA-MB-453) [24].
Expression of mRNA that encodes a G-protein coupled inwardly rectifying potassium channel (GIRK1) has been shown in tissue samples from approximately 40% of primary human breast cancers tested [25], and this expression of GIRK1 was associated with a more aggressive clinical behavior. Increases in GIRK currents by beta-adrenergic stimulation have been reported in adult rat cardiomyocytes and in Xenopus laevis oocytes coexpressing β2-adrenergic receptors and GIRK1/GIRK4 subunits [26]. In addition, in rat atrial myocytes transiently transfected with β1 or β2 adrenergic receptors, the beta-adrenergic agonist isoproterenol stimulated GIRK currents, whereas this stimulation was not seen in non-transfected cells [27]. The current investigations test the hypothesis that GIRK1 channels in human breast cancers are correlated with beta-adrenergic control.
Methods
Cell culture
The ER(+) human breast cancer cell lines MDA-MB-361, ZR-75-1, and MCF-7 and the ER(-) cell lines MDA-MB-453, MDA-MB-468 and MDA-MB-435S were purchased from the American Type Culture Collection (Rockville, MD). Cells were maintained in RPMI 1640 medium supplemented with fetal bovine serum (10%, v/v), L-glutamine (2 mM), 100 U/ml of penicillin and 100 μg/ml streptomycin (Invitrogen-Life Technologies, Grand Island, NY) in an environment of 5% CO2. Exposure of cells to propranolol, isoproterenol, or clozapine (Sigma, St. Louis, MO), NNK (Chemsyn, Lexena, KS), or formoterol hemifumarate (Tocris, Ballwin, MO) for experiments was as detailed in the Figure Legends.
RT-PCR
RNA was isolated by Trizol reagent (Invitrogen-Life Technologies) or by an Absolutely RNA kit (Stratagene, La Jolla, CA). RT-PCR was done as previously described [28]. The GIRK1 primers are forward 5'-ctatggctaccgatacatcacag-3' and reverse 5'-ctgttcagtttgcatgcttcgc-3' which span exon 1 and 2 [29] and amplifies a 441 bp fragment (bases 631–1072, Genbank Acession # NM_002239). The GIRK2 primers are forward 5'-atggatcaggacgtcgaaag-3' and reverse 5'-atctgtgatgacccggtagc-3' amplifies a 438 bp fragment (bases 700–1137, Genbank Acession #U52153). The GIRK4 primers are forward 5'-aaccaggacatggagattgg-3' and reverse 5'-gagaacaggaaagcggacac-3' which amplifies a 401 bp fragment (bases 117–517, Genbank Acession # L47208). PCR conditions are 94°C, 30 sec; 55°C, 30 sec; 72°C, 45 sec for 40 cycles. Cyclophylin primers were used as an internal control (Ambion, Austin, TX).
Relative competitive RT-PCR
Preliminary experiments were done with MDA-MB-453 cells to determine a cycle number of PCR amplification that is within the linear range, which is critical for meaningful results to compare expression levels between samples and to determine the mixture of 18S primers/18S competimers (Ambion-Classic II). The 18S ribosomal RNA primers/competimers are used as an invariant internal control, which allows correction for sample variation. Results indicated this was 31 cycles of PCR and a 1:9 18S primer/competimers ratio. For experimental treatments, as described before [33], cDNA was made and PCR performed except reactions were spiked with 5 μCi [α-32P]-dCTP (3000 Ci/mmole, Dupont-NEN, Boston, MA). Reactions were run with the following conditions: 1 cycle of 2 min. at 94°C, then 31 cycles of 94°C, 30 sec; 55°C, 30 sec; 72°C, 45 sec. A 10 μl sample of each PCR reaction was heated at 95°C for 3 min., then loaded into a 5% TBE-urea Ready Gel (Bio-Rad, Hercules, CA). This underwent electrophoresis at 200 V in TBE buffer until the xylene cyanol dye front reached the bottom of the gel. The gel was transferred to filter paper, dried and exposed to film or imaged on a Molecular Dynamics 445 SI phosphoimager (Sunnyvale, CA). A 100 bp DNA ladder (Invitrogen-Life Technologies) was exchange labeled with T4 polynucleotide kinase and 30 μCi [γ-32P] ATP (3000 Ci/mM, Dupont-NEN).
Real-time PCR
The GIRK-1 primers for real time PCR are forward 5'-ctctcggacctcttcaccac-3' and reverse 5'-gccacggtgtaggtgagaat-3' (bases 398–477, Genbank Acession # NM002239). and the internal TaqMan probe is 6-FAM-tcaagtggcgctggaacctc-TAMRA (bases 429–449, Sigma-Genosys, The Woodlands, TX), annealing temperature 62°. GIRK2 primers-forward 5'-gacctgccaagacacatcag-3' and reverse 5'-cggtcaggtagcgataggtc-3' (bases 766–886, Genbank Acession # U52153) and the internal TaqMan probe is 6-FAM-gtgcaatgttcatcacggcaac-TAMRA (bases 837–859), annealing temperature 56°. GIRK4 primers-forward 5'-agcgctacatggagaagagc-3' and reverse 5'-aagttgaagcgccacttgag-3' (bases 241–358, Genbank Acession # L47208) and the internal TaqMan probe is 6-FAM-accggtacctgagtgacctcttca-TAMRA (bases 301–324), annealing temperature 62°. Reactions were run on a Cepheid SmartCycler (Sunnyvale, CA). Reaction conditions are 200 μM dNTPs, 0.3 μM gene specific primers, 0.2 μM TaqMan probe, 4 mM (GIRK1) or 6 mM (GIRK2or4) magnesium acetate, 2 μl cDNA and 1.5 U MasterTaq (Eppendorf, Westbury, NY) and MasterTaq buffer in a final volume of 25 μl. TaqMan beta-actin detection reagents (Applied Biosystems) were used with the same reaction conditions as above except a 5 mM magnesium concentration was used and this was run at 95° for 120 seconds, followed by 45 cycles of 95°, 15 seconds; 68°, 30 seconds.
Measurement of potassium flux
We determined inward potassium flux in these cells by flow cytometry via the method of Krjukova et al. [30]. The negatively charged fluorescent dye bis-(1,3-dibutylbarbituric acid)trimethine oxonol (DiBaC4(3)) (Molecular Probes, Eugene, OR) was added to MDA-MB-453 breast cancer cell line suspensions of 1 × 106 cells at a final concentration of 150 × 10-9 M. Fluorescence intensity measurement after treatment of the cells was obtained from a FACS Vantage/SE Cell Sorter (San Jose, CA).
Analysis of protein expression by western blots
Following incubation with agents as detailed in the Figure legends, cells were washed twice with phosphate buffered saline and lysed with cold RIPA lysis buffer containing protease inhibitors (50 mM Tris pH 7.4, 150 mM NaCl, 1% NP-40, 1% Triton × 100, 0.1% SDS, 1% sodium deoxycholate, 1 mM EDTA, 50 mM NaF, 10 mM sodium pyrophosphate, 0.5 mM DTT). Cell lysates were collected from culture plates using a rubber policeman, and protein collected by centrifugation. Protein concentrations were determined by BCA protein assay (Pierce, Rockford, IL). Aliquots of 20 μg protein were boiled in 2x loading buffer (0.1 M Tris-Cl, pH 6.8, 4% SDS, 0.2% Bromophenyl blue, 20% glycerol) for 4 minutes, then loaded onto 10% Tris-HCl-Polyacrylamide gels (Biorad, Hercules, CA), and transferred electrophoretically to nictrocellulose membranes. Membranes were incubated with primary antibodies (phospho-Erk; Cell Signaling, Beverly, MA) and appropriate secondary antibodies (Cell Signaling or Rockland, Gilbertsville, PA or Molecular Probes, Eugene OR). In all western blots, membranes were additionally probed with an antibody for actin (Sigma) to ensure equal loading of protein between samples. The antibody-protein complexes were detected as previous described [28] or by the LiCor Odyssey infrared imaging system (Lincoln, NE).
Results
The estrogen-responsive (MCF-7, ZR-75-1, MDA-MB-361) and estrogen non-responsive (MDA-MB-453, MDA-MB-435S, MDA-MB-468) human breast cancer cell lines were screened for the presence of the GIRK1 potassium channel by RT-PCR analysis. The ER(+) cell lines MCF-7, MDA-MB-361 and ZR-75-1 and the ER(-) cell line MDA-MB-453 expressed mRNA for the GIRK1 channel (Figure 1). The ER(-) cell lines MDA-MB-468 and MDA-MB-435S did not express GIRK1 (Figure 1). GIRK1 is also not expressed in the normal breast epithelial cell line MCF 10A (data not shown). The PCR product from the MDA-MB-453 cell line was sequenced to verify the integrity of the PCR process and found to be homologous to the published sequence (data not shown). The PCR primers were designed to span exon 1 and 2 of GIRK1 [29]. In addition PCR amplification of negative control reactions (without the reverse transcriptase enzyme, data not shown) indicated that this was actually representative of mRNA expression and there was no contaminating genomic DNA. Since the GIRK1 potassium channels work as heterotetramers, we needed to determine which other GIRK channels were expressed in these breast cancer cell lines. As determined by RT-PCR, GIRK4 was expressed in all six breast cancer cell lines (Figure 2), and GIRK2 was expressed in four of the six cell lines. GIRK2 was not expressed in ZR-75-1 or MDA-MB-435S cell lines (Figure 2).
To determine if GIRK channels are functionally linked with β-adrenergic receptors in breast cancer cells expressing this ion channel, we decided to investigate the ER(-) cell line MDA-MB-453. This ER (-) cell line, which was the only ER(-) cell line tested that expressed GIRK1, was used for further experiments due to the fact that ER(-) breast cancers have a poorer prognosis than ER(+) cancers [2,3]. In addition, previous research in our laboratories indicated that this cell line expressed the β2 adrenergic receptor but not the β1 receptor [24]. MDA-MB-453 were continuously exposed to the beta-blocker propranolol (1 μM) for 6 days. Previous results from our laboratories indicated that maximal inhibition of breast cancer cell proliferation was at 1 μM propranolol [24]. Using relative RT-PCR, we saw a significant increase in GIRK1 channel mRNA expression (1.6 fold, Figure 3) after 6 days of continuous exposure to propranolol (p < 0.0001 by t-test). In these experiments, propranolol was added fresh each day. We also saw a significant decrease (1.5 fold) in β2-adrenergic receptor mRNA (p < 0.0079 by t-test) (data not shown).
Using the same cDNA samples, we performed a real-time RT-PCR assay using GIRK1 primers designed for real-time PCR and a TaqMan probe. We also saw a significant increase of GIRK1 mRNA using this method (Figure 4) and no change in the control, beta-actin values. Threshold values (CT) were calculated for each sample, which will be lower for samples with more mRNA expression. CT values for GIRK 1 expression were significantly (p < 0.001 by t-test) lower for propranolol treated cells (27.808 ± 0.107) (SD) as compared to control MDA-MB-453 cells (28.964 ± 0.338) (SD). Actin CT values were unchanged between control (13.666 ± 0.286) (SD) and propranolol treated cells (13.404 ± 0.427) (SD). The exposure to propranolol caused a slight decrease in GIRK2 mRNA expression (p < 0.04 by t-test) in the treated cells, opposite the result we found for GIRK1. Control (CT cycle values-31.35 ± 0.73) and propranolol (CT cycle values-32.24 ± 0.38). GIRK4 expression levels were unchanged between control and propranolol treated cells, indicated by real time PCR. Control (CT cycle values-33.0 ± 2.3) and propranolol treated (CT cycle values-31.7 ± 0.38). By contrast, MDA-MB-453 cells treated for 30 minutes daily for 7 days with 1 μM propranolol did not show changes in GIRK1 mRNA expression levels (Figure 5). No significant differences in GIRK1 mRNA expression were seen when MDA-MB-453 cells were exposed to 1 μM of either propranolol or the broad spectrum β-adrenergic agonist isoproterenol for 24 hours (data not shown).
Although the gene expression studies showed no effects at shorter time periods or when it was not in the media constantly, we wished to determine if other cellular function are affected at shorter time periods in MDA-MB-453 cells. Potassium flux into cells would be an important part of any cellular response involving GIRK channels. We determined inward potassium flux in MDA-MB-453 cells by flow cytometry. The negatively charged fluorescent dye bis-(1,3-dibutylbarbituric acid)trimethine oxonol (DiBaC4(3)) was added to MDA-MB-453 breast cancer cell line suspensions of 1 × 106 cells at a final concentration of 150 × 10-9 M. Fluorescence intensity measurement after treatment of the cells was obtained from a FACS Vantage/SE Cell Sorter. An increase of dye fluorescence corresponds to membrane potential depolarization and K+ flux. The β2 selective agonist, formoterol hemifumarate (1 μM), added to MDA-MB-453 cell suspensions at the same time as the fluorescent dye lead to a 2X increase of fluorescence inside the cells, indicating inward potassium movement (Figure 6A &6B). Dye alone added to cells had no effect (data not shown). The GIRK inhibitor clozapine [31] (50 μM) added just prior to dye and formoterol addition completely blocked the effect of the beta-adrenergic agonist formoterol, (Figure 6C) indicating that blockage of the GIRK channel inhibited potassium flux, and that effects of beta-adrenergic agents on this breast cancer cell line are indeed mediated by GIRK channels.
We also determined signaling events in MDA-MB-453 cells that are affected by beta-adrenergic agents. Increased activation of Erk 1/2 was seen in MDA-MB-453 cells after treatment with 100 pM NNK at times ranging from 15–150 minutes (Figure 7). The concentration of NNK used by us is within the range of systemic NNK concentrations found in smokers. In addition, an experiment in Patas monkeys [32] has shown blood levels of 1.6 pg/ml (7.72 × 10-12 M) after exposure to a dose of tritiated NNK equivalent to the amount of NNK found in two packs of cigarettes. Stimulation of Erk 1/2 was also seen using 1 μM of the beta-adrenergic agonist formoterol, but only at 150 minutes (data not shown).
Discussion
Our data demonstrate expression of the G-protein inwardly rectifying potassium channel 1 (GIRK1) in 67% of the breast cancer cell lines tested, with higher levels in ER(+) cell lines. Approximately 40% of primary human breast cancers were found to express GIRK1 and expression of GIRK1 was not found to be correlated with ER status [25]. These differences in our studies may be due to the subset of breast cancer cell lines tested. We also found that the normal breast epithelial cell line MCF 10A lacked GIRK1 expression (data not shown). GIRK1 cannot form functional channels by itself, other GIRK channels are needed [33]. All six breast cancer cell lines tested express either GIRK2 or GIRK4 indicating that functional GIRK potassium channels are possible in these breast cancer cell lines.
The majority of experiments in the present study were done with the ER(-) cell line MDA-MB-453 since it was the only ER(-) cell line tested that expressed GIRK1, and because ER(-) breast cancers have a poorer prognosis than ER(+) cancers [2,3]. We saw a significant increase in GIRK1 channel mRNA expression after 6 days of continuous exposure to propranolol in MDA-MB-453 cells. It is clear that at least six days of continuous exposure to the beta-blocker propranolol is necessary to effect gene expression. Gene expression of β2-adrenergic mRNA was decreased by the same treatment (data not shown). Addition of propranolol for 7 days for only 30 minutes daily had no effect on GIRK1 gene expression. Treatment for a shorter period of time (24 hours) also had no effect on GIRK1gene expression in our studies. The 6 day continuous exposure to propranolol caused a barely detectable decrease in GIRK2 mRNA expression and no change in GIRK4 mRNA expression levels. Longer treatment times may be necessary for gene expression changes in GIRK2 or GIRK4 similar to gene expression changes that are seen in GIRK1.
Although there were no short-term effects of beta-adrenergic agents on GIRK gene expression, we detected other cellular effects. The beta-adrenergic agonist formoterol hemifumarate stimulated potassium influx in MDA-MB-453 cells, and this influx was prevented by the GIRK channel inhibitor clozapine. NNK, a high affinity agonist for beta-adrenergic receptors [11] increased activation of Erk 1/2 in MDA-MB-453 breast cancer cells. Formoterol also increased activation of Erk 1/2, but to a lesser degree (data not shown). Previous studies indicated that the beta-adrenergic agonist isoproterenol stimulates growth [24]. GIRK currents have been shown to be increased in cells stimulated with the beta-adrenergic agonist isoproterenol in rat atrial myocytes transfected with β1or β2 receptors [27]. Heterologous facilitation of GIRK currents by β-adrenergic stimulation was also seen in rat cardiomyocytes [26]. Two polymorphisms in the β2 and β3 adrenergic receptors were found to be correlated with a decreased risk for breast cancer [34], suggesting an important role of this receptor family in the genesis of breast cancer. In previous work, we demonstrated mRNA expression by RT-PCR of the β2 adrenergic receptor in the six breast cancer cell lines used in this study, but expression of β1 in all the estrogen responsive cell lines but not in two ER(-) cell lines (MDA-MB-435S and MDA-MB-453) [24]. Further studies are needed to determine how GIRK1(+) and ER(-) breast cancers are regulated and if GIRK channel agonists and antagonists have effect on proliferation in breast cancer. It also remains to be determined if this same regulation is present in GIRK1(+) and ER(+) breast cancer malignancies. This is of particular importance since a recent report indicated that 17-β-estradiol can modulate GIRK channel activation in the brain [35]. Future studies are also needed to determine if GIRK3 is involved in breast cancer. However, we think this unlikely because one of the functions of GIRK3 is to inhibit plasma membrane expression of other GIRK subunits [36].
Conclusions
All six breast cancer cell lines tested express either GIRK2 or GIRK4 indicating that functional GIRK potassium channels are possible in these breast cancer cell lines. This is the first report that implicates β-adrenergic receptors and G-protein inwardly rectifying potassium channels 1 (GIRK1) in the regulation of human breast cancer cells and suggests a potential role of the tobacco nitrosamine NNK in breast cancers expressing these regulatory pathways. Beta-adrenergic antagonists have both long term effects on gene expression and beta-adrenergic agonists have short term effect on potassium flux and cellular signaling pathways.
Competing interests
The author(s) declare they have no competing interests.
Authors' contributions
HP carried out the majority of experiments and participated in the design of the study, and helped draft the manuscript. QY carried out the western blots. YC was involved in relative competitive RT-PCR studies. HS conceived of the study and helped draft the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgments
We gratefully acknowledge Dr. Neil Quigley (University of Tennessee Sequencing Laboratory) for his assistance with the sequencing and Kindra Walker for her assistance with cell culture, and we also acknowledge Nancy Neilsen for operation of the FACS. We also thank Alysyn Wallace-Gardner for helpful comments on the manuscript and Tommy Jordan for help with the final figures. Supported by the State of Tennessee Center of Excellence Program.
Figures and Tables
Figure 1 Agarose gel showing expression of mRNA for GIRK1 in human breast cancer cell lines by RT-PCR. The GIRK1 primers amplified a 441-bp fragment whereas the cyclophylin primers amplified a 216 bp fragment. For each cell line, a negative control reaction without M-MLV reverse transcriptase was performed and found to be negative. Lanes 1 & 7, ZR-75-1; Lanes 2 & 8, MCF-7; Lanes 3 & 9, MDA-MB-361; Lanes 4 & 10, MDA-MB-435S; Lanes 5 & 11, MDA-MB-453; Lanes 6 & 12, MDA-MB-468, Lane M, a 100 bp marker. PCR reactions resolved on this gel were in the plateau phase of PCR, therefore concentrations of PCR amplified cDNA samples cannot be compared.
Figure 2 Agarose gel showing expression of mRNA for GIRK2 and GIRK4 in human breast cancer cell lines by RT-PCR. The GIRK2 and 4 primers amplified 438 & 401-bp fragments respectively, whereas the cyclophylin primers amplified a 216 bp fragment. Cyclophylin was used as a positive control for both GIRK2 and 4. For each cell line, a negative control reaction without M-MLV reverse transcriptase was performed and found to be negative. Lanes 1 & 7, ZR-75-1; Lanes 2 & 8, MCF-7; Lanes 3 & 9, MDA-MB-361; Lanes 4 & 10, MDA-MB-435S; Lanes 5 & 11, MDA-MB-453; Lanes 6 & 12, MDA-MB-468, Lane M, a 100 bp marker. PCR reactions resolved on this gel were in the plateau phase of PCR, therefore concentrations of PCR amplified cDNA samples cannot be compared.
Figure 3 Comparison of GIRK1 mRNA expression levels by relative competitive RT-PCR in MDA-MB-453 cells treated with propranolol constantly for 6 days. Propranolol (1 μM) was added daily for six days. cDNA was amplified by PCR using GIRK1 primers and 18S primers/competimers. Lanes 1–5) untreated control; Lanes 6–10) propranolol treated cells; Lane 11) untreated control, RT reaction without MMLV; Lane 12, MDA-MB453 treated with propranolol, RT reaction without MMLV. Densitometry values were determined using the phosphoimager. Densitometry values of the bands for Girk 1 were normalized by the densitometry values of the bands for the 18S primers/competimers. Normalized values for control were 0.5494 ± 0.0285 (SD); and normalized values for propranolol treated were 0.9028 ± 0.0348 (SD), p < 0.0001. The bands were consistent with the expected sizes, 441 bp for the GIRK 1 primers and 324 bp for the 18S primers/competimers.
Figure 4 Comparison of GIRK1 mRNA expression levels by real time RT-PCR in MDA-MB-453 cells treated with propranolol constantly for 6 days. Propranolol (1 μM) was added daily for six days. Real time RT-PCR graphs of mRNA expression levels of GIRK1 and beta-actin. The graphs are from a Cepheid Smart Cycler using the same cDNA samples as used in Figure 2. N = 5 for each.
Figure 5 Comparison of GIRK1 mRNA expression levels by relative competitive RT-PCR in MDA-MB-453 cells treated with propranolol for 30 minutes daily for 7 days. Propranolol (1 μM) was added daily for seven days and then removed after 30 minutes. cDNA was amplified by PCR using GIRK1 primers and 18S primers/competimers. Lanes 1–5) untreated control; Lanes 6–10) propranolol treated cells; Lane 11) untreated control, RT reaction without MMLV; Lane 12, MDA-MB453 treated with propranolol, RT reaction without MMLV. Densitometry values were determined using the phosphoimager. Densitometry Girk 1/ Densitometry 18S-control-0.6055 ± 0.0685 (SD); propranolol treated 0.5636 ± 0.0611 (SD). The bands were consistent with the expected sizes, 441 bp for the GIRK 1 primers and 324 bp for the 18S primers/competimers.
Figure 6 Flow cytometry graphs showing potassium flux in MDA-MB-453 cells by the β2 agonist formoterol hemifumarate. A) Fluorescence in R2 (inside the cell) increased from 9.37% to 18.82% after 1 μM formoterol treatment. B) Fluorescence in R2 increased from 8.96% to 21.14% after formoterol treatment C) Fluorescence levels in R2 remained at control levels at 10.63% after addition of 50 μM clozapine along with formoterol.
Figure 7 Levels of ERK activation in MDA-MB-453 cells as assessed by western blot analysis. Activation was determined after indicated times after exposure to 100 pM NNK using a phospho-specific antibody.
==== Refs
Greenlee RT Hill-Harmon MB Murray T Thun M Cancer Statistics CA Cancer J Clin 2001 51 15 36 11577478
Nagai MA Marques LA Yamamoto L Fujiyama CT Brentani MM Estrogen and progesterone receptor mRNA levels in primary breast cancer: association with patient survival and other clinical and tumor features Int J Cancer 1994 59 351 356 7927940
Lemieux P Fuqua S The role of the estrogen receptor in tumor progression J Steroid Biochem Mol Biol 1996 56 87 91 8603051 10.1016/0960-0760(95)00269-3
Morabia A Smoking (active and passive) and breast cancer: epidemiologic evidence up to June 2001 Environ Molec Mutagen 2002 39 89 95 11921174 10.1002/em.10046
Egan KM Stampfer MJ Hunter D Hankinson S Rosner BA Holmes M Willett WC Colditz GA Active and passive smoking in breast cancer: prospective results from the Nurses' Health Study Epidemiology 2002 13 138 143 11880753 10.1097/00001648-200203000-00007
Khuder SA Mutgi AB Nugent S Smoking and breast cancer: a meta-analysis Rev Environ Health 2001 16 253 261 12041881
Claus EB Stowe M Carter D Breast carcinoma in situ: risk factors and screening patterns J Natl Cancer Inst 2001 93 1811 1817 11734598 10.1093/jnci/93.23.1811
Murin S Cigarette smoking and the risk of pulmonary metastasis from breast cancer Chest 2001 119 1635 1640 11399684 10.1378/chest.119.6.1635
Prochazka M Granath F Ekbom A Shields PG Hall P Lung cancer risks in women with previous breast cancer Eur J Cancer 2002 38 1520 1525 12110499 10.1016/S0959-8049(02)00089-8
Hecht SS Hoffmann D N-nitroso compounds and tobacco-induced cancers in man IARC Sci Publ 1991 105 54 61 1855913
Schuller HM Tithof PK Williams M Plummer HK III The tobacco-specific carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone is a β-adrenergic agonist and stimulates DNA synthesis in lung adenocarcinoma via β-adrenergic receptor-mediated release of arachidonic acid Cancer Res 1999 59 4510 4515 10493497
Schuller HM Plummer HK IIIBoschler PN Dudrick P Bell JL Harris RE Co-expression of β-adrenergic receptors and cyclooxygenase-2 in pulmonary adenocarcinoma Int J Oncology 2001 19 445 449
Schuller HM Porter B Riechert A Beta-adrenergic modulation of NNK-induced lung carcinogenesis in hamsters J Cancer Res Clin Oncol 2000 126 624 630 11079726
Masur K Niggerman B Zanker KS Entschladen F Norepinphrine-induced migration of SW 480 colon carcinoma cells is inhibited by beta-blockers Cancer Res 2001 61 2866 2869 11306460
Weddle DL Tithoff PK Williams M Schuller HM Beta adrenergic growth regulation of human cancer cell lines derived from pancreatic ductal carcinomas Carcinogenesis 2001 22 473 479 11238189 10.1093/carcin/22.3.473
Fitzpatrick AL Daling JR Furberg CD Kronmal RA Weissfeld JL Hypertension, heart rate, use of antihypertensives, and incident prostate cancer Ann Epidemiol 2001 11 534 542 11709272 10.1016/S1047-2797(01)00246-0
Hida T Yatabe Y Achiwa H Muramatsu H Kozaki K Nakamura S Ogawa M Mitsudomi T Sugiura T Takahashi T Increased expression of cyclooxygenase 2 occurs frequently in human lung cancers, specifically in adenocarcinomas Cancer Res 1998 58 3761 3764 9731479
Eberhart CE Coffey RJ Radhika A Giardiello FM Ferrenbach S Dubois RN Up-regulation of cyclooxygenase-2 gene expression in human colorectal adenomas and adenocarcinomas Gastroenterology 1994 107 1183 1188 7926468
Uotila P Valve E Martikainen P Nevalainen M Nurmi M Harkonen P Increased expression of cyclooxygenase-2 and nitric oxide synthase-2 in human prostrate cancer Urol Res 2001 29 23 28 11310211 10.1007/s002400000148
Castonguay A Rioux N Inhibition of lung tumourigenesis by sulindac: comparison of two experimental protocols Carcinogenesis 1997 18 491 496 9067547 10.1093/carcin/18.3.491
Rioux N Castonguay A Inhibitors of lipoxygenase: a new class of cancer chemopreventive agents Carcinogenesis 1998 19 1393 1400 9744535 10.1093/carcin/19.8.1393
Schuller HM Zhang L Weddle DL Castonguay A Walker K Miller MS The cyclooxygenase inhibitor ibuprofen and the FLAP inhibitor MK886 inhibit pancreatic carcinogenesis induced in hamsters by transplacental exposure to ethanol and the tobacco carcinogen NNK J Cancer Res Clin Oncol 2002 128 525 532 12384795 10.1007/s00432-002-0365-y
Parrett ML Harris RE Joarder FS Ross MS Clausen KP Robertson FM Cyclooxygenase-2 gene expression in human breast cancer Int J Oncology 1997 10 503 507
Cakir Y Plummer HK IIISchuller HM Beta-adrenergic and arachidonic acid-mediated growth regulation of human breast cancer cell lines Int J Oncology 2002 21 153 157
Stringer BK Cooper AG Shepard SB Overexpression of the G-protein inwardly rectifying potassium channel (GIRK1) in primary breast carcinomas correlates with axillary lymph node metastasis Cancer Res 2001 61 582 588 11212253
Mullner C Vorobiov D Bera AK Uezono Y Yakubovich D Frohnwieser-Steinecker B Dascal N Schreibmayer W Heterologous facilitation of G protein-activated K+ channels by β-adrenergic stimulation via cAMP-dependent protein kinase J Gen Physiol 2000 115 547 557 10779313 10.1085/jgp.115.5.547
Wellner-Kienitz MC Bender K Pott L Overexpression of β1 and β2 adrenergic receptors in rat atrial myocytes. Differential coupling to G protein inward rectifier K+ channel via Gs and Gi/o J Biol Chem 2001 276 37347 37354 11495921 10.1074/jbc.M106234200
Jull BA Plummer HK IIISchuller HM Nicotinic receptor-mediated activation by the tobacco-specific nitrosamine NNK of a Raf-1/MAP kinase pathway, resulting in phosphorylation of c-myc in human small cell lung carcinoma cells and pulmonary neuroendocrine cells J Cancer Res Clin Oncol 2001 127 707 717 11768610
Schoots O Voskoglou T Van Tol HM Genomic organization and promoter analysis of the human G-protein-coupled K+ channel Kir 3.1 (KCNJ3/HGIRK1) Genomics 1997 39 279 288 9119365 10.1006/geno.1996.4495
Krjukova J Osna N Pilmane M Investigation of K+ channel expression in human peripheral lymphocytes of healthy donors by means of flow cytometry Scand J Clin Lab Invest 2000 60 419 428 11003262 10.1080/003655100750019332
Kobayashi T Ikeda K Kumanishi T Effects of clozapine on the δ- and κ-opioid receptors and the G-protein-activated K+ (GIRK) channel expressed in Xenopus oocytes Br J Pharmacol 1998 123 421 426 9504382
Hecht SS Trushin N Reid-Quinn CA Burak ES Jones AB Southers JL Gombar CT Carmella SG Anderson LM Rice JM Metabolism of the tobacco-specific nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone in the patas monkey; Pharmacokinetics and characterization of glucuronide metabolites Carcinogenesis 1993 14 229 236 8435864
Mark MD Herlitze S G-protein mediated gating of inward-rectifier K+ channels Eur J Biochem 2000 267 5830 5836 10998041 10.1046/j.1432-1327.2000.01670.x
Haung XE Hamajima N Saito T Matsuo K Mizutani M Iwata H Iwase T Miura S Mizuno T Tokudome S Tajima K Possible association of β2- and β3-adrenergic receptor gene polymorphisms with susceptibility to breast cancer Breast Cancer Res 2001 3 264 269 11434877 10.1186/bcr304
Kelly MJ Qiu J Ronnekleiv OK Estrogen modulation of G-protein-coupled receptor activation of potassium channels in the central nervous system Ann N Y Acad Sci 2003 1007 6 16 14993035 10.1196/annals.1286.001
Ma D Zerangue N Raab-Graham K Fried SR Jan YN Jan LY Diverse trafficking patterns due to multiple traffic motifs in G protein-activated inwardly rectifying potassium channels from brain and heart Neuron 2002 33 715 729 11879649 10.1016/S0896-6273(02)00614-1
| 15603589 | PMC539303 | CC BY | 2021-01-04 16:03:02 | no | BMC Cancer. 2004 Dec 16; 4:93 | utf-8 | BMC Cancer | 2,004 | 10.1186/1471-2407-4-93 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-4-601559601510.1186/1471-2458-4-60Research ArticleValidity of self reported male balding patterns in epidemiological studies Taylor Rosalind [email protected] Julia [email protected] Justine E [email protected] Lin [email protected] School of Population Health, University of Western Australia, Perth, Australia2004 13 12 2004 4 60 60 17 8 2004 13 12 2004 Copyright © 2004 Taylor et al; licensee BioMed Central Ltd.2004Taylor et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Several studies have investigated the association between male pattern baldness and disease such as prostate cancer and cardiovascular disease. Limitations in the lack of standardized instruments to measure male pattern baldness have resulted in researchers measuring balding patterns in a variety of ways. This paper examines the accuracy and reliability of assessment of balding patterns by both trained observers and men themselves, using the Hamilton-Norwood classification system.
Methods
An observational study was carried out in Western Australia with 105 male volunteers aged between 30 and 70 years. Participants completed a short questionnaire and selected a picture that best represented their balding pattern. Two trained data collectors also independently assessed the participant's balding pattern using the same system and the men's self assessment was compared with the trained observer's assessment. In a substudy, observers assessed the balding pattern in a photo of the man aged 35 years while the man independently rated his balding at that age.
Results
Observers were very reliable in their assessment of balding pattern (85% exact agreement, κ = 0.83). Compared to trained observers, men were moderately accurate in their self-assessment of their balding status (48–55% exact agreement, κ = 0.39–0.46). For the substudy the exact agreement between the men and the observers was 67% and the agreement within balding groups was 87%.
Conclusions
We recommend that male balding patterns be assessed by trained personnel using the Hamilton-Norwood classification system. Where the use of trained personnel is not feasible, men's self assessment both currently and retrospectively has been shown to be adequate.
==== Body
Background
Male pattern baldness is the most common form of baldness observed in human beings[1]. The two main types of balding are: frontal balding in which the hair recedes bilaterally from the forehead region backwards; and vertex balding in which a bald spot appears on the top back of the head. Total balding may be due to continued spread of frontal balding, or a joining up of frontal recession and vertex balding[1,2]. Whilst human hair growth is affected by a number of factors, androgens are the most obvious regulators of normal hair growth and are a prerequisite for male pattern baldness[1]. Serum levels of total and free testosterone, sex hormone binding globulin, and dihydrotestosterone may be important especially given the strong association between free testosterone level and baldness[3].
Several studies have been conducted investigating the association between male pattern baldness (as a proxy for testosterone levels) and various health issues such as prostate cancer [3-7] and cardiovascular disease[6,8,9]. Different methods have been used to assess male pattern baldness in these studies, including the Hamilton-Norwood classification system[2]. This study was undertaken in order to determine the accuracy and reliability of assessment of balding patterns by both trained observers and men themselves, using the Hamilton-Norwood classification system[2].
Methods
Subjects were male volunteers between the ages of 30–70 years recruited from sporting clubs, shopping centres, universities, the central business district and other public areas throughout the metropolitan area of Perth, the capital city of the state of Western Australia, during March 2003 to August 2003. Of the 105 volunteers recruited, two were less than 30 years of age, and therefore were ineligible for the study.
Participants completed a questionnaire providing minimal information regarding demographic characteristics, including age, ethnicity, and level of education. Included in the questionnaire was an unlabelled Hamilton-Norwood classification scale[2] (Figure 1), which participants were asked to examine and to select the picture they believed best depicted their own balding pattern.
Figure 1 Hamilton Norwood Classification Scale (OT Norwood, 1975)
Whilst the subjects completed the questionnaire, two data collectors with some training in recognising different balding patterns independently assessed the participant's balding pattern using the same classification system. On recruitment, participants were told the study was to help with medical research, however the role of the data collectors with regard to their independent assessment of the participant's balding pattern was not revealed until after the questionnaire was complete.
Where possible, (sporting clubs and universities) volunteers were asked to bring a photograph of themselves at age 35 on the day on which the study would be done. The photographs were used by the observers to independently assess the men's balding patterns at age 35 while the participants completed the questionnaire. These participants were also asked to select the picture from the Hamilton-Norwood scale which best depicted their balding pattern at age 35, and were not permitted to look at the photographs while completing the questionnaire.
For data analysis, we compared the assessments of the two observers with each other, and the assessment of the participants with the assessments of the observers. Initially, each of the 12 reference pictures included in the Hamilton-Norwood classification system[2] were considered separately and the percentage of cases with exact agreement on these was determined between each of the observers and the subjects, and between the two observers. The classifications were then arranged into four groups according to overall balding pattern: no balding (A and B in Figure 1); vertex balding only (C); combination of frontal and vertex balding (D to H); and frontal balding only (I to L). Percent agreement within group was then calculated for each observer versus the subjects and between the two observers. The percent of agreement was compared by: age groups (younger than 50 years, and 50 years or older); self-reported ethnicity ("Australian" versus other); and education. Where possible, kappa statistics were also calculated.
Results
There were 69 participants (67%) between the ages 30–49 years, and 34 (33%) were between the ages 50–70 years. Whilst 70% of subjects considered themselves of "Australian" ethnicity (Caucasian born in Australia), the remainder was made up of representatives from Europe (20.4%), Asia (4.9%), New Zealand (3.9%) and South America (1%). A third of subjects had trade education or equivalent (30.1%); 22.3% had completed high school but not gone onto further education; 12.6% had completed junior high school only; and, because of recruitment through universities, 19.4% of subjects had an undergraduate degree and a further 15.5% had completed post-graduate study.
Overall, trained observers were found to be highly reliable at analysing balding patterns with an exact agreement of 85.4% and an agreement within balding pattern groups of 90.3% (Table 1). Compared to the observers, men were found to be moderately accurate in their ability to describe their current balding pattern with an exact agreement percentage ranging from 48.5 to 55.3% and agreement as to balding group around 70%.
Table 1 Reliability and validity of assessment of balding patterns
Exact Balding groups
% agreement κ (p) % agreement κ (p)
Observer 1 vs. Observer 2 85.4 0.828
(p < 0.001) 90.3 0.858
(p < 0.001)
Observer 1 vs. subjects 48.5 0.386
(p < 0.001) 68.0 0.520
(p < 0.001)
Observer 2 vs. subjects 55.8 0.463
(p < 0.001) 73.8 0.599
(p < 0.001)
In regards to how different demographic characteristics affect men's ability to predict their balding pattern (Table 2), the characteristic with the most influence appeared to be age, with men aged 50 or above being more accurate (exact agreement 56–62%) than men aged less than 50 years (exact agreement 45–52%). Men who finished high school were the most accurate at assessing their balding status followed by either those that had studied at technical college or university. The least accurate were men who had completed year 10 at high school or less, with men who had completed post-graduate studies also performing fairly poorly. The effect of ethnicity on ability to assess balding patterns (between Australian and non-Australian men) appeared to be of little significance.
Table 2 Validity of assessment of balding group by age, ethnicity and education.
Observer 1 vs subjects Observer 2 versus subjects
% agreement κ (p) % agreement κ (p)
Age group
<50 years 65.2 0.46(<0.001) 72.5 0.56(p < 0.001)
>50 years 73.5 0.62(<0.001) 76.5 0.66(p < 0.001)
Education
Junior high school 46.2 0.29 (p = 0.058) 61.5 0.48 (p = 0.003)
Senior high school 91.3 0.86 (p < 0.001) 87.0 0.79 (p < 0.001)
Trade school 64.5 0.46 (p < 0.001) 74.2 0.58 (p < 0.001)
Undergraduate 70.0 - 70.0 -
Postgraduate 56.2 0.37 (p = 0.013) 68.7 0.57 (p < 0.001)
Ethnicity
"Australian" 66.7 0.50 (p < 0.001) 75.0 0.61 (p < 0.001)
Other 71.0 0.57 (p < 0.001) 71.0 0.57 (P < 0.001)
There were 15 subjects who provided photos of themselves aged approximately 35 years and both observers examined 13 of these. The inter-observer reliability for exact match was 81.8% (κ = 0.766, p < 0.001) and agreement within balding pattern groups increased to 100% (κ = 1, p < 0.001). Observer 2 only examined 13 subject, so for the 15 subjects examined by Observer 1, agreement between the men and the observer was 66.7% for exact match and 86.7% for agreement within the balding pattern groups.
Discussion
In this study we have shown that trained observers are very reliable in assessing men's balding patterns. Our data also show that, when compared to the trained observers, men themselves can assess their balding patterns quite well. In particular, men are accurate in assessing which balding pattern group they have. This result is important due to previous research suggesting that it is the overall pattern of hair loss rather than extent of balding that determines whether men are at an increased risk of developing negative health outcomes including prostate cancer[3,4,7].
There have been several studies investigating the link between male pattern baldness and prostate cancer[3-5,7,10], as well as other health issues such as cardiovascular disease[6,8,9]. In these studies, balding patterns have been assessed using different techniques, some more complex than others. There has been controversy over the use of some of the more simplistic methods of assessment[6], as little research has been performed regarding their accuracy and their ability to discriminate between the types of balding (vertex, frontal, and combination of vertex and frontal). In a study performed by Hererra et al[11] assessment of subject's balding pattern involved counting the total number of bald spots found on the head. In a repeat assessment performed six years later on the same subjects, there was actually a decreased level of baldness in 12% of study participants. This apparent reversal of baldness was unable to be attributed to regrowth from treatment or other means, and so it must be concluded that the methods used to assess baldness in these participants were unreliable.
Other methods for assessing baldness have been used in clinical situations including reference grids used with standardized photographs of the scalp or in vivo[12]; and videomicroscopic[13] and macrophotographic[14] techniques in which the individual hairs are counted. While these techniques may be used in well-funded clinical trials with the aim of assessing change in balding, they are not appropriate for epidemiological studies in which often the only requirement is to classify men as to their type of balding.
The majority of studies[4,7,10,15] have used variations on the Hamilton scale as modified by Norwood in 1975[2]. This method allows for the grading of baldness in terms of severity and pattern. The scale can be used either by independent observers, or by men themselves, but no official instructions for use or training manuals are available. No previous studies have been performed to assess the accuracy and reliability of either trained independent observers or the participants themselves in the assessment of balding patterns.
The strengths of our study included the recruitment of volunteers from a broad cross-section of the population thus allowing for extrapolation of the results back to the wider community. As our research was performed as an observational study, care was taken to avoid the data collectors influencing the results. This included not informing participants that the observers would be assessing their balding pattern until after the questionnaire was complete. The data collectors also refrained from giving advice to participants when asked to help assess their balding pattern.
Older subjects appeared to be better at assessing their balding group than men less than 50 years. This may be due to greater hair loss resulting in a more straightforward distinction between balding patterns, or possibly just a greater self-awareness of degree of balding amongst those in the older age group. The results with regard to education were confusing with men who had only senior high school education seeming to be best at assessing their balding patterns. This may have just been due to small numbers in the groups. Other demographic characteristics not included in our questionnaire may have been of interest in determining factors that influence accuracy of men's self assessment of their balding pattern. These include factors such as marital status, occupation, and personality sub-types, and we would encourage any further research into the area to investigate the possible relationship between these aspects and the accuracy of men's self-assessment.
Any extrapolation of our data needs to take into account the differing methods of data collection between our study and other studies in which the man may obtain advice from partners or friends in the home environment, as well as have access to mirrors and photographs to assist in their assessment of current and past balding patterns. These factors do not negatively affect our results however, as the use of such help would ultimately increase accuracy of balding assessment from the already acceptable level shown in our results, rather than detract from it.
The study of the accuracy of previous balding was limited by small sample size. Participants had to be approached prior to the study to provide a photograph of them at age 35 years. This limited numbers of eligible participants, and also meant that the men may have viewed the photograph of themselves before completing the questionnaire, which may have increased their accuracy of retrospective assessment. Difficulties in assessing vertex balding from the photographs by the observer also became apparent, as often photographs did not provide an adequate view of the top and back section of a participant's head. It is unlikely that these limitations can be overcome, as it would be difficult to devise another method for the observer to retrospectively assess the participant's balding pattern at age 35.
Previous research has demonstrated an intra-observer rate of consistency in assessment of 98–99% using the Hamilton-Norwood scale[2,10]. Hamilton himself classified 200 balding patterns and then repeated the process three months later without reference to the original classifications. All but one of the classifications were identical. The scale was modified in 1975 by Norwood. The Hamilton scale has been found to correlate with local hair density[14]. A more simple classification of balding patterns was recently been described for use in hair restoration surgery[16] but this scale has not been used in the epidemiological context.
Conclusions
From this study, we suggest that any further work requiring assessment of male balding patterns consider the use of trained observers as the gold standard of assessment. Where this is unattainable, we have shown that men's self evaluation is accurate enough to ensure reliability and validity of results. In addition, we believe this study demonstrates that if links are found between male balding patterns and health effects, that men can reliably determine their own balding pattern and assess their own risk.
Competing interests
The author(s) declare that they have no competing interests.
Authors contributions
RT and JM planned the study, collected the data, analysed the data under supervision, and wrote the study report for a student project.
JL assisted in planning the study and drafted the manuscript
LF assisted in planning the study, supervised data collection and analysis and edited 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
Hibberts NA Howell AE Randall VA Balding hair follicle dermal papilla cells contain higher levels of androgen receptors than those from non balding scalp J Endocrinol 1998 156 59 65 9496234 10.1677/joe.0.1560059
Norwood OT Male Pattern Baldness: Classification and Incidence South M J 1975 68 1359 1365 1188424
Demark-Wahnefried W Lesko S Conaway MR Robertson CN Clark RV Lobaugh B Mathias BJ Smith Strigo B Paulson DF Serum Androgens: Associations with Prostate Cancer Risk and Hair Patterning J Androl 1997 18 495 500 9349747
Hsieh CC Thanos A Mitropoulos D Deliveliotis C Mantzoros CS Trichopoulos D Risk Factors for Prostate Cancer: A case-control study in Greece Int J Cancer 1999 80 699 703 10048970
Hoffmann R Happle R Current understanding of androgenic alopecia. Part I: etiopathogenesis Eur J Dermatol 2000 10 319 327 10846263
Rebora A Baldness and coronary artery disease: the dermatologic point of view of a controversial issue Arch Dermatol 2001 137 943 947 11453815
Giles GG Severi G Sinclair R English DR McCredie MRE Johnson W Boyle P Hopper JL Androgenetic Alopecia and Prostate Cancer: Findings from an Australian Case Control Study Cancer Epidemiol Biomarkers Prev 2002 11 549 553 12050096
Lotufo PA Lee IM Ajani UA Hennekens CH Manson JE Cigarette Smoking and Risk of Prostate Cancer in the Physician's Health Study (United States) Int J Cancer 2000 87 141 144 10861465 10.1002/1097-0215(20000701)87:1<141::AID-IJC21>3.0.CO;2-A
Ellis JA Stebbing M Harrap SB Male pattern baldness is not associated with established cardiovascular risk factors in the general population Clin Sci 2001 100 401 404 11256978 10.1042/CS20000203
Demark-Wahnefried W Schildkraut J Thompson D Lesko S McIntyre L Schwingl P Paulson D Robertson C Anderson E Walther P Early Onsert Baldness and Prostate Cancer Risk Cancer Epidemiol Biomarkers Prev 2000 9 325 328 10750672
Herrera CR D'Agostino RB Gerstman BB Bosco LA Belanger AJ Baldness and coronary heart disease rates in men from the Framingham Study Am J Epidemiol 1995 142 828 833 7572959
Van Neste D Leroy T Sandraps E Validation and clinical relevance of a novel scalp coverage scoring method Skin Res Technol 2003 9 64 72 12535287 10.1034/j.1600-0846.2003.00368.x
Vecchio F Guarrera M Rebora A Perception of baldness and hair density Dermatol 2002 204 33 36 10.1159/000051807
de Lacharriere O Deloche C Misciali C Piraccini BM Vincenzi C Bastien P Tardy I Bernard BA Tosti A Hair diameter diversity: a clinical sign reflecting the follicle miniaturization.[see comment] Arch Dermatol 2001 137 641 646 11346342
Oh BR Kim SJ Moon JD Kim HN Kwon DD Won YH Ryu SB Park IY Association of Benign Prostatic Hyperplasia with Male Pattern Baldness Urology 1998 51 744 748 9610587 10.1016/S0090-4295(98)00108-3
Koo SH Chung HS Yoon ES Park SH A new classification of male pattern baldness and a clinical study of the anterior hairline Aesthetic Plast Surg 2000 24 46 51 10742469 10.1007/s002669910009
| 15596015 | PMC539304 | CC BY | 2021-01-04 16:28:48 | no | BMC Public Health. 2004 Dec 13; 4:60 | utf-8 | BMC Public Health | 2,004 | 10.1186/1471-2458-4-60 | oa_comm |
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-4-301557919810.1186/1472-6750-4-30Methodology ArticleIdentification of proteins in laser-microdissected small cell numbers by SELDI-TOF and Tandem MS Kwapiszewska Grazyna [email protected] Markus [email protected] Ralf [email protected] Rainer M [email protected] Werner [email protected] Norbert [email protected] Ludger [email protected] Department of Pathology, Justus-Liebig University Giessen, Langhansstr.10, 35392 Giessen, Germany2 Ciphergen Biosystems GmbH, Göttingen, Germany3 Department of Internal Medicine II, Justus-Liebig University Giessen, 35392 Giessen, Germany2004 3 12 2004 4 30 30 13 7 2004 3 12 2004 Copyright © 2004 Kwapiszewska 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
Laser microdissection allows precise isolation of specific cell types and compartments from complex tissues. To analyse proteins from small cell numbers, we combine laser-microdissection and manipulation (LMM) with mass spectrometry techniques.
Results
Hemalaun stained mouse lung sections were used to isolate 500–2,000 cells, enough material for complex protein profiles by SELDI-TOF MS (surface enhanced laser desorption and ionization/time of flight mass spectrometry), employing different chromatographic ProteinChip® Arrays. Initially, to establish the principle, we identified specific protein peaks from 20,000 laser-microdissected cells, combining column chromatography, SDS-PAGE, tryptic digestion, SELDI technology and Tandem MS/MS using a ProteinChip® Tandem MS Interface. Secondly, our aim was to reduce the labour requirements of microdissecting several thousand cells. Therefore, we first defined target proteins in a few microdissected cells, then recovered in whole tissue section homogenates from the same lung and applied to these analytical techniques. Both approaches resulted in a successful identification of the selected peaks.
Conclusion
Laser-microdissection may thus be combined with SELDI-TOF MS for generation of protein marker profiles in a cell-type- or compartment-specific manner in complex tissues, linked with mass fingerprinting and peptide sequencing by Tandem MS/MS for definite characterization.
==== Body
Background
Investigation of cell-type specific gene expression and regulation in complex tissues is hampered by the lack of accuracy of cell isolation and sensitivity of post-isolation analysis. Laser-microdissection techniques have proven to be a reliable tool for selectively harvesting cell clusters or single cell profiles from stained tissue sections for mRNA and protein investigation. When combined with qualitative and quantitative PCR, mRNA can be successfully analysed from a few cells [1-3]. The combination of laser-microdissection and cDNA arrays allows investigation of differential gene expression in a cell type specific manner for a multitude of genes in parallel [4,5]. For proteome analysis, two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) has been previously performed from 50,000 to 250,000 microdissected cells, followed by peptide mass finger printing of single spots [6-8]. Isolation of such high cell numbers by laser-microdissection is extremely time consuming or even impracticable in complex tissues.
Recently, several groups have successfully combined laser-microdissection with surface-enhanced laser desorption/ionization mass spectrometry (SELDI MS) to generate reproducible MS profiles from 200–5,000 cells [9-14]. Changes in these protein profiles resulting from different biological conditions can be employed as biomarkers. Similarly, using matrix-assisted laser desorption/ionization mass spectrometry (MALDI MS), spectra could be generated from 500 to 2,500 microdissected cells [15,16]. However, the definite identification of peptides/proteins underlying single biomarkers from laser-microdissected material demands substantially higher cell amounts and laborious, time-intensive procedures. To date, only one biomarker has been identified after profiling of laser-microdissected tissue. In this case, Melle et al. combined 2D-PAGE with peptide mapping and tandem mass spectrometry to identify a protein significantly higher expressed in tumour tissue and confirmed the identity by immunodepletion assay and immunhistochemistry [13].
In addition to generating compartment-specific biomarker profiles, we aimed to develop alternative strategies circumventing the laborious 2D-PAGE for definite protein identification using limited cell numbers derived from microdissected material. The strategies were evaluated at septal and vascular compartments of the complex lung tissue.
Results
Generation of protein profiles by SELDI mass spectrometry
Laser-microdissection and manipulation was used to isolate 500–2,000 alveolar septum cells (Figure 1), which were then transferred into 15 μl of HEPES/Triton X-100 lysis buffer. Approximately 30 intrapulmonary vessels (corresponding to 500–2,000 cell profiles) were microdissected from tissue sections and lysed identically. Following the first isolation, the remaining cell pellet was subjected to Urea/Thiourea/CHAPS (UTC) buffer.
a) To assess the effect of different lysis buffers and different surface properties of the ProteinChip® Arrays, HEPES/Triton X-100 protein lysate as well as UTC lysate was applied independently to SAX (strong anionic exchanger) and WCX (weak cationic exchanger) ProteinChip® Arrays. Compared to the weaker HEPES lysis, UTC buffer resulted in a remarkably higher yield of peaks (signal to noise ratio (S/N) ≥ 3) on WCX arrays. On the other hand, HEPES buffer gave more individual spectra on SAX arrays. 500 cells were sufficient to detect more than 35 peaks on both SAX and WCX arrays. However, immobilization of 2,000 cells resulted in over 50 peaks on WCX arrays (Figure 2A). In regard of limiting cell numbers a two step extraction procedure (HEPES buffer followed by UTC) was proven to be useful to display a higher amount of peaks for differential expression analysis. Therefore this procedure was used for all further profiling experiments.
b) Comparing alveolar septum cells to intrapulmonary vessels, the profiles differed considerably, with only few overlapping peaks. Representative profiles are given in Figure 2B.
c) Four representative SELDI TOF MS spectra of alveolar septum cells from four different animals are shown in Figure 3A. These data show good reproducibility of protein detection by SELDI-MS in agreement with previous studies by Zhukov et al. [14] who also assessed reproducibility of SELDI-MS after using laser-microdissected lung material. To determine the limit in protein abundance for further identification, peaks with different intensities were chosen: one high-abundant protein with molecular weight of 15.7 kD and two low-abundant proteins of 13.8 and 14.0 kD, respectively. The three protein peaks with different intensities are presented in the zoomed area of Figure 3B. For identification of these proteins following strategies were evaluated.
Enrichment of proteins from microdissected cells by column chromatography and SDS-PAGE
Protein lysate of 10,000 to 20,000 microdissected septum cells was extracted by UTC buffer, the remaining material of the same needles was further extracted by SDS sample buffer. Protein samples from UTC and SDS extracts were separated by SDS-PAGE. Although both extracts showed several colloidal Coomassie Brilliant Blue (CBB) stained bands the SDS extract revealed several proteins in the MW region between 12–16 kD (Figure 4). Some of the clearly separated gel bands in the molecular weight region of the target proteins were excised and subjected to in-gel Trypsin digestion. Band 1 represented the 14.0 kD peak while band 2 corresponded to the 13.8 kD peak as identified later by peptide mapping and MS/MS experiments.
The 15.7 kD protein was enriched by micro-spin column chromatography using a Q HyperD® spin column (Ciphergen Biosystems, CA). The protein extract equivalent to approximately 50,000 cells was applied to the column and 6 fractions were eluted according to a stepwise pH gradient and concentrated by trichloroacetic acid precipitation. Aliquots of 3 μl (100 μl total volume) of each fraction were applied to a NP20 ProteinChip Array (hydrophilic chemistry) in order to detect the presence and enrichment of the selected proteins. In the organic fraction (last elution step) all three proteins (13.8, 14.0 and 15.7 kD) were detected on an NP20 ProteinChip. After separation of the complete organic fraction by SDS-PAGE, proteins were stained by colloidal CBB. A protein band with estimated molecular weight between 15–16 kD was excised and subjected to tryptic digestion (not shown).
Identification of the isolated target proteins by tryptic peptide mass fingerprinting on ProteinChip® Arrays and direct peptide fragmentation by Tandem MS/MS
Protein bands isolated from the gel were subjected to Trypsin digestion. Gel pieces were extracted twice and the resulting peptide fragments applied to NP20 and H4 ProteinChip® Arrays with hydrophilic and hydrophobic chromatographic properties, respectively. Peptide mass fingerprinting of the gel bands was done using the PBSIIc instrument. The results are given in Table 1. Two histone proteins (band 1 and 2) and haemoglobin beta were the first candidates in Profound database search. For unambiguous identification, selected peptides were sequenced directly from the arrays by collision-induced dissociation (CID), using a ProteinChip® Interface coupled to a Tandem mass spectrometer [17-19]. Representative MS and MS/MS spectra from band 2 are given in Figure 5. The peptide with an m/z ratio of 1692.88 (Figure 5A) was selected for sequencing by CID-MS/MS (Figure 5B). The obtained sequence was assigned to histone proteins (H2A1 or H2A4, Table 1). It is notable that the molecular masses of the identified proteins correlated well with the results obtained from the profiling experiments (Table 1 and Figure 3B). Analysis of tryptic peptide fragments from band 3 confirmed the molecular mass from profiling experiments (15.7 kD) and showed strong evidence for haemoglobin beta (Sequence coverage 45.7%).
Enrichment of the target protein markers from tissue sections by SDS-PAGE
Due to the labour-intensive requirements of isolating high cell numbers by laser-microdissection, we sought to recover the target proteins in whole lung tissue sections, intending to use this material for subsequent protein identification. Between 5–7 cryosections of lung tissue (10 μm) previously known to contain the target proteins from laser-microdissection were collected and proteins were isolated following the procedure as for the laser-microdissected cells. Applying an aliquot of approximately 4,000 cells to a spot of WCX ProteinChip® Array, we were able to recover the target protein markers within the homogenate spectrum (Figure 6). Using another aliquot of the section material for SDS-PAGE, we isolated single bands of expected weight, as already described for the microdissected cells. This material was subjected to tryptic digestion and peptide mass fingerprint. Again, the two histone proteins and haemoglobin beta were identified.
Discussion
The combination of laser-microdissection and mass spectrometry has been shown to be a reliable tool for compartment and cell type-specific biomarker profiling in complex tissues. Several groups have described the successful generation of mass spectra from as few as 500 to 2,000 cells after microdissection [11,15,16] which was reproduced in the present investigation. Such spectra may be employed to ascertain the cellular origin of microdissected samples. Moreover, when showing differential expression under various biological conditions, mass spectral peaks may serve as biomarkers, independent of their identification. This fast and convenient technique thus represents a valuable tool to provide disease- or status-specific protein marker patterns to be used for diagnostic and predictive purposes [20-22]. In the present investigation we could confirm feasibility and excellent reproducibility of this approach when analyzing lung tissue compartments.
To investigate proteins on 2D-gels, high amounts of cells are required. Therefore, using this approach as a starting point for protein profiling with subsequent MS identification, 50,000 to 250,000 microdissected cells have to be introduced per gel. Our aim was to minimize this laborious procedure, being hardly compatible with microdissection of minor cellular compartments. Thus, we first generated compartment specific profiles from laser-microdissected material by mass spectrometry and subsequently collected high cell numbers to identify the previously selected proteins. Moreover, using an interface to a Tandem MS/MS instrument, analysis of tryptic mass fingerprints can be performed from the same array without need for additional material. To reduce the demand of material on the gel, different staining techniques can be used (e.g. silver or SYPRO RUBY® fluorescence staining). While advantageous in gel staining due to higher sensitivity, the problems are shifted towards MS technique: the identification may fail due to minute amounts of protein per gel spot. Therefore, in our study we performed robust and easy CBB staining to be able to detect the required amount of material in MS techniques. Additionally, for low abundant proteins, direct Trypsin digestion of an eluted, gel-resolved protein can be performed directly on the array. Marker protein isolation and enrichment of the protein peak may also be enhanced by application of suitable pH and salt conditions directly on the array.
A promising alternative for labour-intensive 2D-gel applications can be the detection of biomarker proteins from microdissected cells and their subsequent identification in tissue slices. Since the exact weight of the target protein is known from preceding experiments with microdissected cells, the section homogenate can be screened for the respective peaks. As tissue sections are easy and fast to obtain, they were used to generate profiles on WCX arrays from the same lungs as from microdissection. While these spectra differed partially from those derived from microdissected alveolar septum cells, the peaks corresponding to the pre-defined target proteins were easily detected in the homogenate spectrum. In addition, corresponding bands could be detected by SDS-PAGE and subsequent tryptic mass fingerprinting confirmed the identity of histones H2B F and H2A1 and haemoglobin beta.
Typically, high-resolution 2D techniques require several days from sample application to the final staining of protein spots. Isolation of 500 to 2,000 cells by laser-microdissection requires minutes to hours, depending on the targeted cell type, tissue area or organ compartment. Array pre-treatment, immobilization of the lysate and washing lasts around 90 minutes and MS measurement is performed within few minutes. This calculation may reveal the time saved by our approach.
Due to the accuracy of measurement we found that the molecular weight derived from SELDI mass spectrometry corresponded very well with the exact protein mass. For the three investigated proteins, the mass accuracy was approximately 0.1% for external calibration and is thus remarkably higher than using 2D-PAGE. Nevertheless, the exact molecular weight alone is insufficient to identify the concerning protein directly via database search.
Conclusions
Combination of laser-microdissection with SELDI-TOF MS generates reproducible and credible biomarker profiles in a cell-type- or compartment-specific manner from complex tissues. For identification of underlying peptides/proteins, this approach may be combined with enrichment and isolation strategies, linked with mass fingerprinting by SELDI-TOF MS and peptide sequencing by Tandem MS/MS for definite chemical characterization. These techniques allow analysis of differential protein expression of low cell numbers microdissected from complex tissues.
Methods
Mouse lung preparation
All animal experiments were approved by local authorities [Regierungspräsidium Giessen, no. II25.3-19c20-15(1) GI20/10-Nr.22/2000]. Lung preparation was performed as described previously [5]. In brief, male BALB/c mice (Charles River, Sulzfeld, Germany, 20–22 g) were exposed to normobaric normoxia in a chamber at FiO2 = 0.21. After 1 day, animals were sacrificed; lungs were flushed via the pulmonary artery, and 800 μl prewarmed TissueTek® (Sakura Finetek, Zoeterwoude, The Netherlands) were instilled into the airways via a tracheal cannula. Afterwards, lungs were excised and frozen in liquid nitrogen immediately.
Laser-assisted microdissection
Laser-microdissection and manipulation was performed as described previously [1-3,5]. In brief, cryosections (10 μm) from lung tissue were mounted on glass slides. After hemalaun staining for 30 sec, sections were subsequently immersed in 70% and 96% ethanol and stored in 100% ethanol until use. Not more than 10 sections were prepared at the same time to restrict storage time. Alveolar septum cells and intrapulmonary vessels, respectively, were microdissected under visual control using the Laser Microbeam System (P.A.L.M., Bernried, Germany) and isolated by a sterile 30 G needle (Figure 1). Needles with adherent material were transferred into a reaction tube containing HEPES/Triton X-100 lysis buffer (50 mM HEPES, pH 7.2, 1% Triton X-100).
Comparative protein expression profiling by SELDI
Needles with adherent cells were transferred to the reaction tubes containing 15 μl of the HEPES/Triton X-100 lysis buffer. After vigorous shaking for 30 min at room temperature, samples were centrifuged at 14,000 g for 10 min. 3 μl of this supernatant were directly applied to the spots of SAX and WCX ProteinChip® Arrays and incubated in a humid chamber for 1 h. For profiling, SAX ProteinChip® Arrays were preincubated in SAX Binding Buffer (100 mM Tris, pH 8.5, 0.02% Triton X-100).
After first extraction of the microdissected material using HEPES buffer, 15 μl of UTC buffer (6 M Urea, 2 M Thiourea, 2% CHAPS, 75 mM DTT) were applied to needles/cells. The pellet in UTC buffer was centrifuged for 10 min at 14,000 g and the complete supernatant was used for profiling on WCX ProteinChip® Arrays. These were pre-treated with 10 mM HCl for 5 min and equilibrated two times with WCX binding buffer (100 mM NaOAc, pH 4.5, 0.02% Triton X-100) for 5 min. Fifteen μl of UTC supernatant were diluted 1:10 v/v in WCX binding buffer and incubated under vigorous shaking for 45 min. In order to deal with high amounts of volume (150 μl) the Bioprocessor (Ciphergen Biosystems. Inc.) was applied. After removal of the samples every spot was washed twice using binding buffer followed by a final 10 sec water rinse. After air-drying, saturated sinapinic acid (0.6 μl) dissolved in 50% acetonitrile and 0.5% trifluoracetic acid (TFA) was added twice. Subsequently, mass analysis of bound peptides/proteins was performed using the Ciphergen PBS IIc system. ProteinChip® Arrays were analyzed by averaging 100–150 laser shots collected in the positive mode. Optimization range of the time lag focusing was set between 10–30 kD. Deflector settings were used to filter out peaks with <2000 m/z. Calibration was performed externally using purified peptide and protein standards. Obtained spectra were analyzed by the ProteinChip Software version 3.01.
Small scale column chromatography and SDS-PAGE
SDS-PAGE was performed as described previously [23] with minor modifications. We used 15% acrylamid separation and 5% stacking gels in a mini-gel chamber (Roth, Karlsruhe). Supernatants of UTC buffer extraction containing approximately 2000 cells/μl were mixed 1:1 with twofold concentrated sample buffer (100 mM Tris-Cl pH 6.8, 4% SDS, 20% Glycerin, 3% DTT, 0.05% Bromphenolblue). Samples were boiled for 3 min and a total volume of 20 μl (equivalent to 20,000 cells) was loaded onto single lanes of the gel. Afterwards, 1 μl SDS sample buffer per 1,000 cells was added to the needles in order to solubilize remaining proteins after UTC buffer extraction. SDS extracts were also applied to SDS gels.
For selective enrichment of marker proteins we used small sized anionic exchange spin columns resulting in 6 fractions after elution using a stepwise pH gradient from pH 9–3 and an organic fraction. The protein amount reflecting 50,000 cells was loaded onto a Q HyperD spin column and aliquots of the eluted fractions were analyzed on NP20 ProteinChip® Arrays to reveal the enrichment of selected protein peaks in a certain fraction. The fractions containing the enriched proteins were concentrated by TCA precipitation in order to apply the complete fraction to a single lane of an SDS gel. Staining was performed by colloidal Coomassie brilliant blue (CBB, Roth).
Protein tryptic digestion (peptide mass fingerprint)
The CBB stained bands matching the expected molecular weight regions of selected proteins were excised and subjected to Trypsin digestion. Gel pieces were washed three times with 400 μl of 100 mM ammonium bicarbonate and 50% acetonitrile for 15 min followed by 15 min incubation in 100% acetonitrile. After removal, gel pieces were dried shortly in a speed-vac centrifuge. Depending on the gel volume, 10–15 μl of a Trypsin solution (20 ng Trypsin/μl in 25 mM ammonium bicarbonate) was applied and digestion was performed overnight (16 h) at 37°C. Afterwards, reaction tubes were centrifuged and 0.5–1.5 μl aliquots of each supernatant were applied to the spots of H4 (hydrophobic surface coating) and NP20 (hydrophilic surface coating) ProteinChip® Arrays. A 20% matrix solution of alpha-4 hydroxy-cinnamic acid (CHCA) was applied to the spots. The remaining gel pieces were extracted with a 60% acetonitrile and 0.2% TFA solution for 1 h with a 5 min sonication to extract remaining organic peptides. Supernatants from this second extraction step were also applied to H4 and NP20 ProteinChip® Arrays. All Arrays were measured in the PBS IIc system by averaging 150 laser shots. After subtraction of all peaks also present in the blank gel piece (e.g. Trypsin autolysis peaks), m/z values were submitted to Profound and Mascot for database searching.
Tandem MS/MS analysis
Peptides from tryptic digestion were applied to NP20 ProteinChip® Arrays and 2 × 0.6 μl of a saturated CHCA solution was added. For quality control of the peak intensities, the NP20 Arrays were analyzed in a PBS IIc instrument. Afterward, the arrays were transferred to a Tandem MS instrument. Data were acquired on a Micromass QTOF II (Manchester, UK) tandem quadrupole-time of flight (Q-TOF) mass spectrometer equipped with a PCI 1000 ProteinChip® Tandem MS Interface (Ciphergen Biosystems). Ions were created using a pulsed nitrogen laser operating with 30 pulses/sec. Nitrogen gas was used for collisional cooling of formed ions and argon gas was used for all low-energy collision-induced dissociation experiments. The system was externally calibrated in MS/MS mode using the parent ion and selected fragments of adrenocorticotropic hormone (ACTH) human fragment 18–39 (m/z = 2465.1983; Sigma Aldrich).
Authors' contributions
GK: laser-microdissection, preparation of samples, PAGE, SELDI-MS
MM: SELDI-MS and optimalisation for microdissected material
RB: tandem MS measurements
RMB: instruction to laser-microdissection
WS: design of project, preparation of the manuscript
NW: animal model, preparation of the lungs
LF: design of project, preparation of the manuscript
Acknowledgement
We thank M.M. Stein and K. Quanz for skilful technical assistance, J. Wolff and L. Marsh for critical reading of the manuscript and G. Jurat for photographic arrangements. This study was funded by the Deutsche Forschungsgemeinschaft, SFB 547, project B7 and Z1. We gratefully acknowledge the funding of the PBS IIc SELDI-MS spectrometer by the Kerckhoff-Foundation, Bad Nauheim.
Figures and Tables
Figure 1 Isolation of an alveolar septum by laser-assisted microdissection and manipulation from a hemalaun stained frozen lung section (magnification 200×). A) Alveolar septum is selected for isolation. B) Laser photolysis is used to disconnect the cells from adjacent ones. C) Septum cells adhere tightly to the approximated sterile needle and can be transferred into a reaction tube.
Figure 2 Protein profiling of laser-microdissected alveolar septum cells and intrapulmonary vessels for detection of differentially expressed proteins (m/z = mass/charge). A) SELDI-TOF spectra comparison of a SAX ProteinChip® Array profile from HEPES/Triton X-100 lysed septum cells and WCX ProteinChip Array® profile from septum cells treated with UTC buffer. B) Spectra comparison of intrapulmonary vessel profiles and alveolar septum (2,000 cells) after UTC lysis and binding onto WCX ProteinChip® Array.
Figure 3 SELDI-TOF spectra of alveolar septum cells (2,000 cells) from four different mice A) Laser-microdissected alveolar septa from 4 different mice were lysed in UTC lysis buffer and profiled on a WCX ProteinChip® Array. Spectra of the molecular range between 3–18 kD are depicted. B) Zoom of the region between 12.5–16.5 kD: A representative spectrum showing three peaks at 13.8, 14.0 and 15.7 kD which were selected for further biochemical identification.
Figure 4 SDS extract of 20,000 microdissected alveolar septum cells separated by SDS-PAGE. After first extraction with UTC buffer a second extraction step with SDS sample buffer was performed. The amount equivalent to 20,000 cells of this extract was separated on a 15% SDS gel. Band 1 and band 2 correspond to the selected protein peaks of 13.8 and 14.0 kD from WCX Array experiments (Figure 2).
Figure 5 MS/MS analysis of the tryptic fragments of band 2. A) Tryptic fragments of band 2 were applied to NP20 ProteinChip® Arrays and measured on a ProteinChip® Interface coupled to a Tandem MS. The peptide of 1692.88 D was selected for MS/MS analysis. B) CID-MS/MS spectrum from tryptic fragment with m/z 1692.88.
Figure 6 SELDI-TOF spectra comparison of profiles from microdissected cells versus lung tissue sections. Microdissected alveolar septum cells as well as five lung tissue sections were lysed in UTC buffer and bound onto WCX ProteinChip® Arrays. All three investigated target proteins could be recovered from the tissue section derived profile as well.
Table 1 Protein database query of the three investigated target protein bands. Details and specifications of the analysis that resulted in identification of the biomarker proteins are given. MW: molecular weight
Gelband No. MW calc. Sequence coverage % Score profound MS/MS sequence Acces. No. Swissprot Protein name
band 1 13.797 63 2.24 AMGIMNSFVNDIFER P10853 Histone H2B F
band 2 14.042 31 HLQLAIRNDEELNK P22752 Histone H2A.1
14.522 P10812 H2A.4
band 3 15.708 45.9 1.17 P02088 Haemoglobin β
==== Refs
Fink L Stahl U Ermert L Kummer W Seeger W Bohle RM Rat porphobilinogen deaminase gene: a pseudogene-free internal standard for laser-assisted cell picking Biotechniques 1999 26 510 516 10090993
Fink L Seeger W Ermert L Hanze J Stahl U Grimminger F Kummer W Bohle RM Real-time quantitative RT-PCR after laser-assisted cell picking Nat Med 1998 4 1329 1333 9809560 10.1038/3327
Fink L Bohle RM Internal standards for laser microdissection Methods Enzymol 2002 356 99 113 12418191
Luo L Salunga RC Guo H Bittner A Joy KC Galindo JE Xiao H Rogers KE Wan JS Jackson MR Erlander MG Gene expression profiles of laser-captured adjacent neuronal subtypes Nat Med 1999 5 117 122 9883850 10.1038/4806
Fink L Kohlhoff S Stein MM Hanze J Weissmann N Rose F Akkayagil E Manz D Grimminger F Seeger W Bohle RM cDNA array hybridization after laser-assisted microdissection from nonneoplastic tissue Am J Pathol 2002 160 81 90 11786402
Craven RA Banks RE Use of laser capture microdissection to selectively obtain distinct populations of cells for proteomic analysis Methods Enzymol 2002 356 33 49 12418186
Xia SH Hu LP Hu H Ying WT Xu X Cai Y Han YL Chen BS Wei F Qian XH Cai YY Shen Y Wu M Wang MR Three isoforms of annexin I are preferentially expressed in normal esophageal epithelia but down-regulated in esophageal squamous cell carcinomas Oncogene 2002 21 6641 6648 12242662 10.1038/sj.onc.1205818
Zhou G Li H DeCamp D Chen S Shu H Gong Y Flaig M Gillespie JW Hu N Taylor PR Emmert-Buck MR Liotta LA Petricoin EF Zhao Y 2D differential in-gel electrophoresis for the identification of esophageal scans cell cancer-specific protein markers Mol Cell Proteomics 2002 1 117 124 12096129 10.1074/mcp.M100015-MCP200
Jr GW Cazares LH Leung SM Nasim S Adam BL Yip TT Schellhammer PF Gong L Vlahou A Proteinchip(R) surface enhanced laser desorption/ionization (SELDI) mass spectrometry: a novel protein biochip technology for detection of prostate cancer biomarkers in complex protein mixtures Prostate Cancer Prostatic Dis 1999 2 264 276 12497173 10.1038/sj.pcan.4500384
Cazares LH Adam BL Ward MD Nasim S Schellhammer PF Semmes OJ Wright GLJ Normal, benign, preneoplastic, and malignant prostate cells have distinct protein expression profiles resolved by surface enhanced laser desorption/ionization mass spectrometry Clin Cancer Res 2002 8 2541 2552 12171882
Wellmann A Wollscheid V Lu H Ma ZL Albers P Schutze K Rohde V Behrens P Dreschers S Ko Y Wernert N Analysis of microdissected prostate tissue with ProteinChip arrays--a way to new insights into carcinogenesis and to diagnostic tools Int J Mol Med 2002 9 341 347 11891524
Zheng Y Xu Y Ye B Lei J Weinstein MH O'Leary MP Richie JP Mok SC Liu BC Prostate carcinoma tissue proteomics for biomarker discovery Cancer 2003 98 2576 2582 14669276 10.1002/cncr.11849
Melle C Ernst G Schimmel B Bleul A Koscielny S Wiesner A Bogumil R Moller U Osterloh D Halbhuber KJ Von Eggeling F Biomarker Discovery and Identification in Laser Microdissected Head and Neck Squamous Cell Carcinoma with ProteinChip(R) Technology, Two-dimensional Gel Electrophoresis, Tandem Mass Spectrometry, and Immunohistochemistry Mol Cell Proteomics 2003 2 443 452 12824440
Zhukov TA Johanson RA Cantor AB Clark RA Tockman MS Discovery of distinct protein profiles specific for lung tumors and pre-malignant lung lesions by SELDI mass spectrometry Lung Cancer 2003 40 267 279 12781425
Palmer-Toy DE Sarracino DA Sgroi D LeVangie R Leopold PE Direct acquisition of matrix-assisted laser Desorption/Ionization time-of-flight mass spectra from laser capture microdissected tissues Clin Chem 2000 46 1513 1516 10973907
Xu BJ Caprioli RM Sanders ME Jensen RA Direct analysis of laser capture microdissected cells by MALDI mass spectrometry J Am Soc Mass Spectrom 2002 13 1292 1297 12443019 10.1016/S1044-0305(02)00644-X
Zhang L Yu W He T Yu J Caffrey RE Dalmasso EA Fu S Pham T Mei J Ho JJ Zhang W Lopez P Ho DD Contribution of human alpha-defensin 1, 2, and 3 to the anti-HIV-1 activity of CD8 antiviral factor Science 2002 298 995 1000 12351674 10.1126/science.1076185
Weinberger SR Viner RI Ho P Tagless extraction-retentate chromatography: a new global protein digestion strategy for monitoring differential protein expression Electrophoresis 2002 23 3182 3192 12298090 10.1002/1522-2683(200209)23:18<3182::AID-ELPS3182>3.0.CO;2-5
Reid G Gan BS She YM Ens W Weinberger S Howard JC Rapid identification of probiotic lactobacillus biosurfactant proteins by ProteinChip tandem mass spectrometry tryptic peptide sequencing Appl Environ Microbiol 2002 68 977 980 11823248 10.1128/AEM.68.2.977-980.2002
Petricoin EF Liotta LA Mass spectrometry-based diagnostics: the upcoming revolution in disease detection Clin Chem 2003 49 533 534 12651801 10.1373/49.4.533
Zhang Z Bast RCJ Yu Y Li J Sokoll LJ Rai AJ Rosenzweig JM Cameron B Wang YY Meng XY Berchuck A Van Haaften-Day C Hacker NF de Bruijn HW van der Zee AG Jacobs IJ Fung ET Chan DW Three biomarkers identified from serum proteomic analysis for the detection of early stage ovarian cancer Cancer Res 2004 64 5882 5890 15313933
Wiesner A Detection of tumor markers with ProteinChip technology Curr Pharm Biotechnol 2004 5 45 67 14965209
Ossipow V Laemmli UK Schibler U A simple method to renature DNA-binding proteins separated by SDS-polyacrylamide gel electrophoresis Nucleic Acids Res 1993 21 6040 6041 8290367
| 15579198 | PMC539305 | CC BY | 2021-01-04 16:02:57 | no | BMC Biotechnol. 2004 Dec 3; 4:30 | utf-8 | BMC Biotechnol | 2,004 | 10.1186/1472-6750-4-30 | oa_comm |
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-4-231559600510.1186/1472-6947-4-23Research ArticleGPs' decisions on drug treatment for patients with high cholesterol values: A think-aloud study Backlund Lars [email protected]ånér Ylva [email protected] Henry [email protected] Johan [email protected] Lars-Erik [email protected] Center for Family Medicine, Karolinska Institutet, Alfred Nobels allé 12, SE-141 83 Huddinge, Sweden2 Department of Psychology, University of Stockholm, Sweden3 Statisticon AB, Östra Ågatan 31, SE-753 22 Uppsala, Sweden2004 13 12 2004 4 23 23 2 6 2004 13 12 2004 Copyright © 2004 Backlund 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 was to examine how General Practitioners (GPs) use clinical information and rules from guidelines in their decisions on drug treatment for high cholesterol values.
Methods
Twenty GPs were presented with six case vignettes and were instructed to think aloud while successively more information about a case was presented, and finally to decide if a drug should be prescribed or not. The statements were coded for the clinical information to which they referred and for favouring or not favouring prescription.
Results
The evaluation of clinical information was compatible with decision-making as a search for reasons or arguments. Lifestyle-related information like smoking and overweight seemed to be evaluated from different perspectives. A patient's smoking favoured treatment for some GPs and disfavoured treatment for others.
Conclusions
The method promised to be useful for understanding why doctors differ in their decisions on the same patient descriptions and why rules from the guidelines are not followed strictly.
==== Body
Background
The medical decision examined in our study is whether or not to initiate drug treatment for high cholesterol values. The topic has been the focus of much debate on the grounds that the proportion of individuals with elevated cholesterol values is high in most Western populations, and that the costs for treating all these people with drugs life-long would be enormous, with a marginal benefit in risk reduction for the majority of them [1-3]. The current Swedish guidelines [4] from 2003 are national applications of the recommendations on coronary prevention of the Third Joint European Task Force [5]. In sum, the national guidelines define a total cholesterol value above 5 mmol/l and/or an LDL value above 3 mmol/l as hypercholesterolaemia and the same values as the goals for treatment. As a comparison, the American guidelines are more aggressive in terms of treatment goals for patients with established coronary heart disease and they are more focused on the LDL levels [6]. The Swedish guidelines identify two group of patients that in general should be offered pharmacological treatment after a sufficiently long trial of life style intervention: the individuals with already established cardiovascular disease (so called secondary prevention cases) or diabetes. A third group that in general should receive medication are patients with familial hyperlipidaemia (FH). For the remaining individuals with a total cholesterol above 5 mmol/l and/or LDL above 3 mmol/l (primary prevention), the same guidelines suggest that the decision to recommend a drug or not should be based on an estimate of the combined risk stemming from the individual's different risk factors. More specifically, the recommendation is to make a numerical risk estimate of the risk for coronary heart disease (CHD) within the next ten years (or the risk projected to 60 years) with a cut-off value at 20%. Based on the results from epidemiological studies, algorithms for arriving at such risk estimates (e.g. Anderson et al. [7]) have been developed and are available, for instance, as charts in pocket format for doctors.
Thus, the decision-making task can be carried out as follows. The first step is to decide whether the patient case is an instance of secondary prevention, diabetes or FH, and if not, to estimate the numerical risk for coronary disease within ten years. A risk above 20% would justify pharmacological treatment, given that life style intervention has been tried for a sufficiently long period.
In this study we address the question of how General Practitioners (GPs), who manage most of the cholesterol testing and treatment in Sweden, make such decisions when guidelines are not physically available to them. We will try to highlight the decision-making by examining how it is affected by clinical variables describing the patient and by medical knowledge and decision rules on behalf of the doctors. The reason for studying decisions without access to written guidelines is that as experienced GPs (in the case of three of us), we have found that this is how decisions on cholesterol treatment are usually made. Furthermore, in a previous study concerning the ability of GPs to make numerical estimates of future cardiovascular risks, we asked the GPs if they had access to any tool for making numerical risk estimates at their clinic [8]. Only nine out of 84 respondents said they had such a tool. GPs' judgments regarding cholesterol treatment have been studied previously using Clinical Judgment Analysis, CJA (for a description of this research paradigm, see Cooksey [9], for an overview of the medical applications, see Wigton [10]). The variation among doctors with respect to which information about the patient influenced their decisions most (strategies) seemed to be high [11], and the strategies were not in accordance with the guidelines for a substantial proportion of the doctors [12]. The number of patient variables (cues) that influenced the decisions was two or three for most doctors [11,12]. About one-fourth of the doctors did not include coronary heart disease in the irjudgments [12], in spite of the central role of this risk factor according to the guidelines. The statistical modelling with CJA describes individual doctors' judgment strategies added together for a set of cases, but does not give direct information about the kind of rules or medical knowledge that the participants use in their decisions. In the present study we used the think-aloud technique [13] in order to learn more about the use of medical knowledge and rules. In addition, with this technique we are better able to study the decisions for individual patient cases. In a previous paper, based on the present data [14], we coded the think-aloud protocols for preferences concerning two decision alternatives, prescribing a cholesterol lowering drug or not doing so. The codings proved to be reliable. They also appeared to be valid in the sense that there was good agreement on how think-aloud data and rating data, both concerning preferences for prescribing and not prescribing, described the decision process over time for different simulated patient cases. In the present study we link such preference data from think-aloud protocols to different kinds of information describing the patient cases. We also investigate how the use of rules and guidelines can be inferred from the think-aloud protocols.
Our first set of research questions concerned how different kinds of information about the patient (e.g. age, sex, previous diseases and laboratory tests) relate to the decision to prescribe a drug or not to do so. First, we estimated the importance of the individual information categories by counting the total number of times they had, according to the coding of the verbal protocols, been valued in a positive or negative direction in relation to the decision at hand. Second, in order to get an idea about why different doctors reach different decisions when presented with identical case information, we made the following analyses. For each of the patient cases separately, the subgroup who decided to prescribe was compared with those with an opposite decision regarding how often they valued different information categories. Third, to further understand how the participants differed in their judgments, we examined which kinds of specified information about a patient (e.g. male sex) are the most likely to lead to disagreement, i.e. to be judged in a positive direction by some participants and in a negative direction by others.
Disagreement about the evaluation of data on a given variable may result from different cut-off values, e.g. for the cholesterol variable. A certain value can be considered high for one participant and thus speaking for drug treatment. The same cholesterol value may be considered almost normal by another participant and thereby speaking against drug treatment for the same patient. The age variable may also be associated with disagreement due to different cut-off values. A higher age is generally associated with a higher risk, but there is a lack of evidence for the potential benefit of treating the oldest age groups, and this may introduce different cut-off values for different doctors.
Disagreement may also be caused by what might be called different perspectives. If we take smoking as an example all doctors should recognize smoking as a factor associated with an increase in future cardiovascular risk, and should accordingly make statements with a positive directionality for drug treatment. On the other hand, some doctors in some situations may regard actions aimed at smoking cessation as more beneficial than cholesterol reduction, which may give smoking a negative directionality in relation to drug treatment. Overweight can be regarded in the same way, i.e. as an indicator of drug treatment or as indicating change of life style as preferable to drug treatment. Thus, there are two alternative treatment philosophies – drug treatment or life style change – which in turn may be associated with opposite evaluations of the same data in relation to drug treatment. To the extent that these philosophies in fact are associated with different evaluations, one may regard them as different perspectives where certain data (e.g. smoking) are seen from different angles, as risk indicators or as entities that could be changed through patient's own efforts (i.e. by changing life style) as a means to treat the his or her health problem. The latter perspective may also be associated with somewhat moralistic evaluations, e.g. that overweight or smoking is the patient's own choice or own "fault", which in turn would decrease the inclination to initiate drug treatment. Some evidence for this conjecture comes from a CJA study by Evans et al [11]. They interviewed the doctors after the case presentations regarding which factors they thought had influenced their judgment most. The GPs stated that they were generally less likely to treat people who were overweight. Most were also less likely to treat smokers, but some had the opposite policy. Those less likely to treat smokers were also less likely to treat obese patients. The traditional medical risk factors like diabetes and hypertension may also be associated with either the risk increase perspective or an alternative perspective, where other variables than the cholesterol level are in focus for treatment. Such an alternative perspective should be more likely with a poor control of the blood pressure or diabetes parameters. As this was hardly the case with our case vignettes, and as the moralistic perspective is more likely with life style factors, we expected disagreement to be more frequent with life style variables than with traditional medical factors.
Our second set of research questions concerned the use of rules, and the concept of risk as shown in the verbal protocols. Six patient cases were chosen that included two high-risk patients (secondary prevention or diabetes) for whom the guidelines can be transferred to a simple decision rule (e.g. "patients with elevated cholesterol values and previous coronary heart disease should have drug treatment recommended"). Our question was how frequently such decision rules were in the verbal protocols and their content in relation to practice guidelines for elevated blood lipids. For the remaining four cases (primary prevention) no such simple guideline-based rule can be applied and instead, a numerical risk calculation is suggested. We examined the extent to which references to risk estimates were made in the think-aloud protocols. For both secondary and primary prevention cases we were interested in determining how the decisions corresponded to what is indicated by guidelines and risk algorithms.
In sum, our research questions concerned: Importance of information (which categories of information about the patients seem to be most important for the decisions?). Patterns of importance for "Yes" and "No" decisions (when each case is analyzed separately, can we get an idea of the reasons behind different decisions by comparing the information evaluation for doctors who chose prescription with those with the opposite decision?). Disagreement (which categories of information give rise to disagreement?). Use of rules (their frequency and contents). Risk estimation (for cases that should be decided by use of a numerical risk calculation, according to the guidelines, how frequent is a referral to the concept of risk estimation in the verbal protocols?).
Our approach to analyse think-aloud protocols in a medical decision task for the relative importance of different information categories and the amount of disagreement in the evaluation of these categories has not been tried before as far as we know. We believe that the results can be useful for understanding why doctors reach different decisions in response to the same patient cases and why they often do not act in accordance to guidelines. This knowledge should be useful as an aid to design guidelines and teaching.
Methods
Design
The 20 participants received the same six patient cases and the order of the cases was the same for all participants. Cases with "Yes"- and "No" decisions as the recommended treatment according to the guidelines were mixed as evenly as possible. Ten of the participants were randomly assigned to a condition where, in addition to thinking aloud, they also rated their willingness to prescribe a drug at regular intervals during each case. As was described in a previous paper [14], this group did not differ from the group without the rating task as regards the prescription decision or the directionality in the think-aloud data. In the present paper the think-aloud data from these two groups have been compiled, whereas the analyses of the rating data are confined to this previous paper.
Participants
Twenty GPs working in the southern Stockholm area participated. There were 10 males and 10 females. Their ages varied between 34 and 60 years (M = 48.3) and they had practiced between one and 22 years (M = 11.4) as specialists in family medicine. A total of 36 doctors were contacted by telephone. They were selected so as to have a relatively even distribution across different districts in the area and according to gender, but the selection was not random. Twenty-four agreed to participate, but before the session four of these later declined to participate.
Cases
Six clinical cases were selected from an original set of 40 authentic cases with cholesterol values above normal (at least 5.5 mmol/L). The original set was used in a Clinical Judgment Analysis design with a different sample of doctors and is described in Backlund et al [12]. Two of the cases, GM (with diabetes mellitus) and AR (with angina pectoris), were obvious high-risk patients and it would be reasonable to use the guidelines in a straightforward manner and recommend treatment for these cases (under the assumption that lifestyle modification had already been tried). Case SH was distinguished by the absence of risk factors other than a moderate increase in cholesterol level, and it would therefore be reasonable to refrain from drug treatment. In the remaining three cases (IS, TW and PU) additional risk factors existed such as smoking and hypertension, and the recommended line of management would be to let the decision be guided by a numerical estimate of the future risk for coronary heart disease. The presently available risk-charts that are referred to in the Swedish [4] and European guidelines [5] indicate a 10–20% ten-year risk for case IS (treatment not justified), and a 20–40% risk for case TW (treatment justified). For case PU the calculated risk is 10–20%, which would suggest refraining from drug treatment. However, case PU had a strong family history of coronary heart disease, which is an important risk factor although it is not directly included in the chart. A decision to prescribe a drug is probably justified for this case. Case SH had a calculated risk of 5–10%. For all six cases lifestyle intervention as well as advice concerning diet and exercise had been tried for at least six months before the visit in question.
The different kinds of clinical information presented on the six successive screens were divided time-wise in the same way for all six cases. The order in which this information was presented was arranged so as to be as realistic as possible in relation to clinical practice (including how patient cases are described in written referrals to other clinics and in clinical conferences and tutoring). Table 1 shows case IS as it was presented to participants in the study. All previously shown information about a case was repeated on the later screens to reduce and control for memory effects. This part of the text was placed at the top of the screen and was a different colour.
Procedure
The study was conducted at the doctor's office or in a room nearby. All visits and recordings were made by one of the authors (LB). The cases were presented on a computer screen (Software Question Asker™ (QA) [15]. In the course of six screens, more clinical information was gradually added to the case. The participants were instructed that authentic cases of hypercholesterolaemia would be presented and that their task was to voice aloud all their thoughts about the case, and that each case would end with the question as to whether or not they would prescribe a drug for this patient. They controlled the shift to a new screen by using a mouse click. When the participant had finished a screen, he or she clicked on a "continue" button. If a participant was silent for more than 10–15 sec, he or she was reminded to voice aloud all thoughts about the information presented.
The study was approved by the local ethics committee.
Response measures and coding of data
Decision
Each case ended with a screen with the following question: "Would you prescribe a cholesterol-lowering drug for this patient?" The participant responded by clicking on one of two response alternatives, "Yes" or "No".
Think-aloud protocols
The sessions were tape-recorded. A secretary then transcribed the recorded sessions into a written, word-by-word format. The protocols were segmented into statements. The next step was to categorize the statements into one of ten categories concerning the general characteristic of the statement (cp. Cognition Categories): Attention, Evaluation, Rule, Explanation, Action pharmacological treatment, Action non-pharmacological treatment, Action other, Want of information, Rating (valid for the ten participants with an additional rating task) and Other. The set of categories is described in more detail in Backlund et al [14]. Each statement was also assigned one of the values +, -, 0 or x, denoting directionality in relation to the decision task (to prescribe or not to prescribe). The majority of statements that could be assigned a positive or negative directionality had been categorized into either "evaluation" or "application of a rule". However, the most frequent outcome of the categorization was "attention" (the participant read the information aloud, or retrieved it from memory, with neutral or no reformulation), and no directionality could be assigned (i.e. coded as "x"). Zero directionality indicated an explicit statement that the information was neutral in relation to the decision. Finally, each statement was coded with respect to the information referred to. A certain information category could be coded more than once for a given doctor and a given case, unless we regarded the statement as a mere repetition (close in time and identically or almost identically phrased as an earlier statement). The original set contained 21 information categories within the areas of background data, medical conditions, previous diseases, lifestyle factors, physical examination and laboratory tests.
Results
Reliability of the coding system
Two of the authors (LB and YS) independently coded the protocols from the first six participants. Reliability was computed separately for directionality (+, -, 0 or x) and for information category (one of 21) as the percentage of statements that were coded into the same directionality/information category. For these first six participants, the inter-judge reliability was 92% for directionality and 94% for information category. The reliability measures were considered to be satisfactory and therefore only one of the authors (LB) performed the remaining coding.
Information categories
The original set of 21 different information categories was reduced to eleven. When we ranked the information categories with regard to the frequency of positive or negative evaluations there was a great leap between weight (frequency 14 and rank eleven) and triglycerides (frequency four and rank twelve). We therefore excluded triglycerides and information categories with fewer evaluations. Examples of such excluded categories were information about physical examination of the heart and lungs (normal outcome for each of the cases) and test results concerning liver or thyroid function. The remaining eleven information categories were Cholesterol, LDL (low density lipo-protein), HDL (high density lipo-protein), weight, smoking, CHD (coronary heart disease), diabetes, hypertension, heredity, sex and age.
Treatment decisions
Table 2 summarizes the information for each case as regards these eleven categories and shows the number of doctors who decided to prescribe a drug as well as the recommended decisions according to the Swedish guidelines. The frequency varied from zero (case SH) to 17 (case AR). "Yes"-decisions and "No"-decisions were equal in frequency when summarized over participants and cases. Figure 1 shows that the majority of doctors chose to prescribe for two, three or four of the six cases. No participant decided to prescribe for all six cases and one participant chose not to prescribe for any of the cases.
Importance of information
Figure 2 shows that information about cholesterol was evaluated most frequently, both in the positive and the negative direction. It can also be seen that positive evaluations were more frequent than negative ones, except for sex, age and weight, which in part can be due how the final question was worded. For each of the 20 participants we calculated the number of evaluative statements (i.e. with a positive or negative directionality) for each of the eleven information categories as an index of the importance of the information category. A 2 (Direction: Positive/Negative) × 11 (Information category 1–11) × 6 (Case 1–6) ANOVA with three within-group variables and the number of evaluated statements as dependent variable was performed. The main effect of Information category was highly significant, F (4.4; 84.1) = 8.80, p < .01, as was Information category × Case interaction, F (9.1; 172.2) = 11.80, p < .01. (In ANOVA with repeated measures in this study, the degrees of freedom were adjusted according to the Greenhouse-Geisser Epsilon). Thus, as could be expected, the different information categories were evaluated unequally often, and the pattern of relative importance differed in the six individual patient cases. All other main effects and interaction effects were also significant with p < .01.
Patterns of importance for "Yes" and "No" decisions
In the following, the six patient cases will be analysed separately. This may allow us to detect possible differences in the pattern of importance between different information categories for participants who decided to prescribe a drug and those who made the opposite decision. For each case the number of evaluative statements was the dependent variable in a 2 (Decision: Yes/No) × 11 (Information category 1–11) × 2 (Direction: positive/negative) ANOVA, with the first as a between-group variable and the latter two as within-group variables. The statistical effects are summarized in Table 3.
The main effect of Decision was not significant in any of the cases, indicating that there was no evidence of an association between the number of evaluative statements and decision outcome. For four of the cases the effect of Direction was significant, indicating that positive and negative statements were of unequal frequency. Except for case SH (the case for which all 20 participants decided not to prescribe), positive statements were more frequent than negative ones. For all six cases the different information categories were evaluated unequally often (i.e. main effect of Information). A significant Decision × Direction Interaction, indicating that the decision to prescribe or not to prescribe was associated with different distributions between positive and negative statements, was found in only two of the cases. Direction × Information Interaction was significant or nearly significant for all five cases, suggesting that the distribution of positive and negative directionality was unequal across the different information categories.
The most interesting part of these analyses, however, is whether different decisions were associated with different evaluative patterns across the information categories. Statistically, this corresponds to two- or three-way interaction effects including Decision and Information.
A significant Decision × Information interaction for a case would indicate that participants with a "Yes"-decision had their number of evaluative statements differently distributed across Information categories compared to participants with a "No"-decision, regardless of whether the direction was positive or negative. The three-way interaction includes the directionality of the statements as well. As can be seen from Table 3, most of these interaction effects were non-significant. However, the number of evaluations per patient case is probably too small to give enough power for such interaction effects. Three of the cases will be selected to illustrate how this approach may provide hypotheses about information strategies.
Case IS represents a 67-year-old female with hypertension as a central risk factor in addition to her cholesterol elevation. She also had a modest heredity. As Figure 3 shows, the 12 participants who decided to prescribe had more positive evaluations of the central risk factor hypertension. In addition, the group who decided not to prescribe seemed to make negative evaluations of information about heredity, suggesting that this information may have been an "argument" against pharmacological treatment for some of the participants in the "No"-subgroup.
Case TW (Figure 3) represents a case with several risk factors in addition to cholesterol elevation (e.g. smoking and hypertension). An important difference between the groups may be that the "No"-group evaluated the patient's relatively low cholesterol level more often and in the negative direction in relation to pharmacological treatment.
Case AR (Figure 3) represents a case with CHD (in this case angina pectoris). A comparison between the response groups suggests that greater emphasis was put on CHD by the "Yes"-group, and at the same time there was a negative evaluation, as regards pharmacological treatment, of the patient's (over-) weight by some participants in the "No"-group.
Thus, an analysis of the response patterns for these three cases suggests that the "Yes" and "No"-groups differed not only in how much they evaluated the central risk factor(s) (in addition to cholesterol elevation) as favouring drug treatment but also in that the "No"-group seemed to have identified at least one information category as evidence against treatment. As regards the remaining cases, the evaluative pattern for case PU (young woman with severe heredity for CHD) could be interpreted in a similar way, with the patient's (young) age as an argument against drug treatment, whereas case GM (diabetic case) did not invite any such interpretation. For Case SH, no comparison between response groups could be made as all participants decided not to prescribe.
Disagreement
We defined agreement as the degree to which the same information about a patient case was evaluated with the same directionality. We hypothesized that disagreement would be more common for information about lifestyle-related factors like smoking and weight than would be the case for medical conditions like hypertension and diabetes. There was only one case with a clear overweight and one case where the patient smoked, and the number of evaluative statements concerning these two information categories was therefore rather low. Regarding case TW's smoking, there were 21 statements with a positive direction and ten with a negative direction. For case AR's overweight there were three positive and six negative statements. For hypertension, each of the cases had either only positive or only negative directions (or no statements with directionality at all). The same pattern was found for diabetes and CHD, except that for diabetes two statements concerning case GM were negative compared to 24 statements with positive directionality, and for CHD (case AR) one statements was negative whereas 31 were positive. Thus, with minor exceptions the participants agreed on the evaluations of these three information categories. The data were consequently in line with our hypothesis.
There were few evaluative statements concerning the sex of the cases. With one single exception all statements were negative and concerned female cases, which is in line with the known lower risk for female patients to suffer from cardiovascular diseases. A corresponding tendency towards positive evaluations of the male cases was not clearly demonstrated in this material, which could suggest a possible bias in how sex is evaluated as a risk factor. As far as the age variable is concerned, there were both positive and negative statements (i. e. disagreement) for four of the six cases, which was in accord with our expectations. Among the information categories concerning laboratory values, cholesterol was evaluated most often by far, with a fairly even distribution of positive (total 46) and negative (total 39) statements. For at least four of the cases, there were approximately the same numbers of positive and negative evaluations of the same cholesterol value. In other words, according to our definition there is evidence of disagreement among participants in the evaluation of the different cholesterol values.
At the level of individual participants, there were eleven instances where doctors made both positive and a negative evaluation(s) of one case. Four of these eleven concerned smoking, two cholesterol, two LDL and one each of hypertension, coronary heart disease and diabetes.
Use of rules
A total of 32 statements (i.e. not more than 1.6 per participant) were coded as rules. According to our judgment, 18 of these 32 statements were derived from or were compatible with the guidelines (including those rules that were not entirely correct in detail, e.g. regarding the cut-off limits for LDL and HDL) and twelve of these 18 were a more or less directly referring to secondary prevention or diabetes (e.g "He has angina pectoris and should be below 5 in cholesterol"). Examples of other contents for the statements coded as rules were age limit for cholesterol treatment, importance of looking for secondary hypercholesterolaemia, the role of LDL/HDL ratio, priority of smoking vs. pharmacological treatment, the desired blood pressure value for diabetics blood pressure and the cut-off value for ten-year risk for primary prevention. For two of the patient cases (case GM with diabetes mellitus, and case AR with a history of angina pectoris), the guidelines allow a simple decision rule to be applied. Of the 32 instances of reference to a rule, 24 were in connection with these two patient cases.
Risk estimation
For the four primary prevention cases, IS, TW, SH and PU, a number of statements referring to numerical risk estimate (guidelines say 20% within the next ten years) could have been expected. Only two participants referred to numerical risk estimates.
Discussion
We discuss first how the doctors evaluated the available information in relation to the decision to be made (i.e., in terms of directionality). When each case was analyzed separately, there was some evidence of different patterns of information use shown by prescribers and non-prescribers. The non-prescribers seemed to evaluate central risk factors with a positive directionality less often than prescribers, and they also appeared to identify at least one information category that was given a negative directionality. This is compatible with theories that describe decision-making as search for arguments or reasons for one or the other decision alternative [16,17].
Paradoxically, the information categories that some doctors used as arguments against treatment were used as arguments for treatment by other doctors. We interpret this finding as showing that prescribing and not prescribing doctors evaluate given information from different perspectives, i.e., from different viewing angles that will put different aspects of the given information in the foreground and background, respectively [18]. If we take smoking or overweight as examples, they could be seen as risk indicators for CHD (which is what is naturally seen from a drug treatment perspective) or as possibilities for life-style change, which in turn will reduce the patient's future coronary risk (which is what is naturally seen from a life-style change perspective). It may be noted that the doctors did in general not consider both ways of evaluating the information to assess their relative weight for and against a decision to prescribe. Instead, only the aspects supporting this decision or an alternative decision were focused, which is in line with the assumption that the doctors viewed the information from a certain perspective that favoured the decision to be made. For example, we can compare the protocol by participant 6 (decision Yes) regarding Case AR: "He has angina and he has overweight so I will treat him" with the protocol from participant 3 (decision No) regarding the same case: "I would like him to reduce his weight first".
The use of life style factors as arguments for prescription decisions was further illustrated in a separate analysis based on the same verbal protocols that also include a task where the doctors were asked to describe freely how they usually reason when they meet patients with high cholesterol values [19]. The protocols were coded for knowledge of guidelines content and for arguments for the decision to prescribe or not. In several instances the doctor seemed to be fully aware of the contents of the guidelines but still decided to refrain from a strict application of it. The arguments for the decisions in these cases often concerned life style factors like smoking or overweight – either as risk increasing factors or as alternative strategies for intervention.
Disagreement was also shown for the age variable. Age is generally considered as positively and monotonically related to risk for future cardiovascular events. At the same time, the guidelines make the reservation that the benefit of giving drugs to very old people is unclear. As far as young patients are concerned, the perspective of ten-year risk appears to be too narrow. The recommended procedure is to increase or project the age to 60 years in order to estimate the risk [4,5]. For doctors with limited experience in using the risk charts, this might be confusing.
Cholesterol was another variable that was ambiguous which could be explained in part by the selection of patient cases. Four of the six cases had cholesterol values in the range of 5.0–6.5 mmol/L, which is often labeled as a mild elevation. This might have formed the basis for negative evaluations, i.e. when a value was close to normal a decision to refrain from drug prescription could have been favored. A few participants also commented that the cholesterol values were lower than they had expected, or lower than those of their own patients.
Ambiguity in the decision situation due to seeing the situation in terms of different treatment perspectives (relevant for life style factors) or different ideas about the optimal cut-off points (relevant for age and laboratory values) could possibly be reduced by clearer guidelines, which in turn accentuate the need for more research on a number of issues. These issues include the role of life-style factors for coronary heart disease, as well as how patients should be motivated to change their life style, and cost-benefit outcomes of using drug treatment of patients in different age groups and with different cholesterol values.
We will now consider the second set of research questions addressed in the present study, viz., the extent to which the participants used certain rules as a basis for their decisions. Based on the verbal protocols, the frequency of statements classified as a rule was rather low, on average 1.6 per participant, and most of the rules concerned secondary prevention. Part of the explanation for the low number of rules might be that the participants were uncertain about the contents of the guidelines and were therefore less willing to talk about them. However, from our separate coding for knowledge of guidelines content referred to above [19], the conclusion was that the doctors were in general well aware of the distinction between primary and secondary prevention.
The low frequency of statements containing a reference to the risk concept could be explained in the same way, since the participants had no immediate access to an aid for calculating risk and were possibly unsure about the general content of such an aid (e.g. the numerical value for ten-year risk that would put the patient into a high-risk category and justify drug treatment). Another reason for the low number of rules might be that the instructions did not encourage the participants to explain their decisions, but simply to state aloud their thoughts about the presented information, which is generally considered as the best method to ensure that the verbal protocols reflect the cognitive processes of interest [13]. A third possible influence on the use of rules may be that cases that could be handled in a straight-forward way by applying rules from the guidelines were at he same time characterized by having cholesterol values that were only marginally increased above normal, which might have introduced conflict in the decision situation.
From the view of evidence-based decisions and quality of care, we can say that many of the cases were difficult and that a considerable spread in the decisions was to be expected. In fact, most of the participants found it difficult to decide about several of the cases, which was evident from interviews after the sessions. On the other hand, the only case (SH) with a mild risk (5–10%) was correctly judged by every participant as not being a candidate for drug treatment. Case AR with angina pectoris represents a decision situation where the guidelines could justify pharmacological treatment in a straightforward manner, and 17 out of 20 chose to prescribe. The presence of diabetes in Case GM could similarly justify drug treatment, but this was the choice for only half the participants. The reason could be that the recommendations concerning diabetes as a risk factor in parity with coronary heart disease is rather new. The Swedish guidelines were published in 1999 and the study was conducted in 2000.
One limitation of the present analyses is that most of the conclusions are based on pooled data from groups of participants, while the principal interest is in strategies at the individual level. For example, the opposing evaluations of the same patient data could only be demonstrated between doctors due to the low number of patient cases. After completion of the six cases, the participants were asked to relate in their own words how they usually reason regarding pharmacological treatment when confronted with patients with high cholesterol values. In a forthcoming paper these narratives will be analyzed at the individual level as "scripts" for dealing with cholesterol treatment. We will then have a better understanding of how knowledge and guidelines in this area of medicine are represented in memory, and how these cognitive structures are related to actual decisions and to the individual doctor's think-aloud protocols from processing the cases.
Conclusions
In this study we have used a new method to analyse a medical treatment decision. Verbal protocols were coded with respect to how different patient variables seemed to favor or not to favor the decision to prescribe a drug or not. The method promised to be fruitful for understanding why doctors reach different decisions in response to the same patient descriptions and why guidelines are not followed.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All authors participated in the design of the study. LB carried out the data collection. LB and YS performed the coding of the protocols. LB performed the rest of the data analyses and drafted the manuscript. All authors participated in the discussion of the draft. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank the participating GPs. The study was supported by the Stockholm County Council and the Swedish Heart Lung Foundation.
Figures and Tables
Figure 1 The number of participants with 0–6 "Yes"-decisions The number of participants with 0–6 "Yes"-responses, out of 6 possible, to the question whether or not to recommend drug treatment for the different patient cases.
Figure 2 Relative importance of the different information categories. For each of the information categories, its importance for the decision to recommend drug treatment or not, was defined as the mean number of statements per participant that were coded with a positive or negative directionality in relation to the treatment decision.
Figure 3 Patterns of importance for "Yes" and "No" decisions. As in Figure 2, the mean numbers of statements with a positive and negative directionality are shown for the 11 different information categories. In Figure 3, the patterns of positive and negative evaluations across the information categories are shown separately for the subgroups with "Yes" decisions and "No" decisions, respectively. Three of the six patient cases are shown.
Table 1 Example of a case (IS)
Screen Information
1 The patient is a 67-year-old woman whose recent cholesterol value was 7.3 mmol/L. She has had the diagnosis hypercholesterolaemia for two years. She has been given advice concerning diet but she has not been prescribed a cholesterol-lowering drug. Her cholesterol value has decreased from an initial value of 7.8 mmol/L
2 The patient has been on medication for hypertension for 10 years (Seloken ZOC* 50 mg and Plendil** 5 mg). She is now on a visit to check her blood pressure and hypercholesterolaemia.
3 The patient has no other known diseases apart from osteoarthritis of her knees. Her mother suffered from hypertension and reached the age of 84 years.
4 The patient is a non-smoker. She very seldom drinks alcohol. She does not exercise on a regular basis but she is fond of taking walks.
5 Physical examination: Good general condition. A few kilograms overweight. Blood pressure 145/75. Heart and lung auscultation normal.
6 Laboratory values: Total cholesterol 7.3 mmol/L. LDL 5.4 mmol/L. HDL 1.2 mmol/L. Triglycerides 1.6 mmol/L. TSH, creatinine and liver-function tests were normal.
7 Would you prescribe a cholesterol-lowering drug for this patient?
Yes No
Table 2 Characteristics of the cases and the number of doctors who decided to prescribe a drug
Information category Case
IS GM TW SH AR PU
Age 67 53 67 51 56 41
Sex Female Female Male Female Male Female
Heredity Slight Slight No Slight No Strong
Hypertension Yes No Yes No No No
Diabetes No Yes No No No No
CHD No No No No Yes No
Smoking No No Yes No No No
Overweight Slight No No No Yes No
HDL 1.2 1.1 1.0 1.0 0.9 1.2
LDL 5.4 4.1 3.6 4.3 4.3 5.3
Cholesterol 7.3 5.9 6.0 6.5 5.9 7.2
Decision according to guidelines No Yes Yes No Yes No?*
Percentage of doctors who decided to prescribe 60 50 35 0 85 70
Table 3 The frequency of statements with positive or negative directionality. Summary of test statistics The table summarizes the results of ANOVA with the frequency of evaluative statements as dependent variable and Decision, Direction and Information as independent variables. F-values are given with the results of significance tests within parentheses.
Case
Statistical effects IS GM TW SH AR PU
Decision .73 (ns) 1.86 (ns) .87 (ns) * .01 (ns) .01 (ns)
Direction 10.54 (<.01) 3.98 (.06) 6.09 (<.05) 7.39 (<.01) .84 (ns) 8.31 (<.01)
Information 7.03 (<.01) 16.80 (<.01) 11.17 (<.01) 1.99 (.09) 5.87 (<.01) 11.15 (<.01)
Decision × Direction 2.55 (ns) .69 (ns) 7.89 (<.05) * 12.65 (<.01) 2.42 (ns)
Direction × Information 6.40 (<.01) 5.78 (<.01) 4.57 (.01) 2.18 (.07) 6.28 (<.01) 12.90 (<.01)
Decision × Information 2.06 (<.09) .61 (ns) .74 (ns) * 4.38 (<.01) 1.15 (ns)
Decision × Direction × Information 1.54 (ns) .85 (ns) 2.18 (ns) * 3.05 (<.05) 1.09 (ns)
==== Refs
Pharoah PDP Hollingswoth W Cost effectiveness of lowering cholesterol concentration with statins in patients with and without pre-existing coronary heart disease: life table method applied to health authority population BMJ 1996 312 1443 1448 8664620
Muir J Applying the Sheffield tables to data from general practice Br J Gen Pract 1999 162 156 159
Garber A Using cost-effectiveness analysis to target cholesterol reduction Ann Intern Med 2000 132 833 835 10819708
Läkemedelsverket Behandling av hyperlipidemi – nationella och internationella guidelines (Treatment of hyperlipidaemia – national and international guidelines, in Swedish) Information från Läkemedelsverket 2003 7 21 68
Third Joint Task Force of European and Other Societies on Cardiovascular Disease Prevention in Clinical Practice European guidelines on cardiovascular disease prevention in clinical practice Eur H J 2003 24 1601 1610 10.1016/S0195-668X(03)00347-6
Grundy SM Balady GJ Criqui MH Fletcher G Greenland P Guide to primary prevention of cardiovascular diseases: a statement for healthcare professionals from the task force on risk reduction Circulation 1997 95 2329 2331 9142014
Anderson KM Wilson PWF Odell PM Kannel WB An updated coronary risk profile. A statement for health professionals Circulation 1991 83 356 362 1984895
Backlund L Bring J Strender L-E How accurately do general practitioners and students estimate coronary risk in hypercholesterolaemic patients? Primary Health Care Research and Development 2004 5 145 152 10.1191/1463423604pc179oa
Cooksey RW Judgment analysis: theory methods, and applications 1966 Boston:Academic Press
Wigton RS Use of linear models to analyze physicians' decisions Med Decis Making 1988 8 241 252 3054396
Evans JS Harries C Dennis I Dean J General practitioners' tacit and stated policies in the prescription of lipid lowering agents Br J Gen Practice 1995 45 15 18
Backlund L Danielsson B Bring J Strender L-E Factors influencing GPs' decisions on the prescription of lipid-lowering drugs Scand J Prim Health Care 2000 18 87 93 10944062 10.1080/028134300750018963
Ericsson KA Simon HA Protocol Analysis: Verbal reports as data 1984 Cambridge: The MIT Press
Backlund L Skånér Y Montgomery H Bring J Strender L-E Doctors' decision processes in a drug prescription task: The validity of rating scales and think-aloud reports Organ Behav Hum Dec Process 2003 91 108 113 10.1016/S0749-5978(02)00529-0
Question Asker 2000. Computer software and manual
Lipshitz R Klein GA, Orasanu J, Calderwood R Decision making as argument-driven action In Decision making in action: Models and methods 1993 Norwood: Ablex 172 181
Shafir E Simonson I Tversky A Reason-based choice Cognition 1993 49 11 36 8287671 10.1016/0010-0277(93)90034-S
Montgomery H Towards a perspective theory of decision making and judgment Acta Psychol (Amst) 1994 87 155 178 7810350 10.1016/0001-6918(94)90049-3
Backlund L Skånér Y Montgomery H Bring J Strender L-E The role of guidelines and the patient's life-style in GPs' management of hypercholesterolaemia BMC Family Practice 2004 5 3 15113452 10.1186/1471-2296-5-3
| 15596005 | PMC539306 | CC BY | 2021-01-04 16:03:41 | no | BMC Med Inform Decis Mak. 2004 Dec 13; 4:23 | utf-8 | BMC Med Inform Decis Mak | 2,004 | 10.1186/1472-6947-4-23 | oa_comm |
==== Front
BMC Med GenetBMC Medical Genetics1471-2350BioMed Central London 1471-2350-5-271558830010.1186/1471-2350-5-27Research ArticleNeural network analysis in pharmacogenetics of mood disorders Serretti Alessandro [email protected] Enrico [email protected] Istituto Scientifico Universitario Ospedale San Raffaele, Department of Neuropsychiatric Sciences, Milano, Italy2 Università Vita-Salute San Raffaele, School of Medicine, Milano, Italy2004 9 12 2004 5 27 27 23 7 2004 9 12 2004 Copyright © 2004 Serretti and Smeraldi; licensee BioMed Central Ltd.2004Serretti and Smeraldi; 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 number of available genotypes for genetic studies in humans requires more advanced techniques of analysis. We previously reported significant univariate associations between gene polymorphisms and antidepressant response in mood disorders. However the combined analysis of multiple gene polymorphisms and clinical variables requires the use of non linear methods.
Methods
In the present study we tested a neural network strategy for a combined analysis of two gene polymorphisms. A Multi Layer Perceptron model showed the best performance and was therefore selected over the other networks. One hundred and twenty one depressed inpatients treated with fluvoxamine in the context of previously reported pharmacogenetic studies were included. The polymorphism in the transcriptional control region upstream of the 5HTT coding sequence (SERTPR) and in the Tryptophan Hydroxylase (TPH) gene were analysed simultaneously.
Results
A multi layer perceptron network composed by 1 hidden layer with 7 nodes was chosen. 77.5 % of responders and 51.2% of non responders were correctly classified (ROC area = 0.731 – empirical p value = 0.0082). Finally, we performed a comparison with traditional techniques. A discriminant function analysis correctly classified 34.1 % of responders and 68.1 % of non responders (F = 8.16 p = 0.0005).
Conclusions
Overall, our findings suggest that neural networks may be a valid technique for the analysis of gene polymorphisms in pharmacogenetic studies. The complex interactions modelled through NN may be eventually applied at the clinical level for the individualized therapy.
==== Body
Background
The increasing number of available genotypes for genetic studies in humans requires more advanced techniques of analysis [1]. Moreover, genes interact in a complex way, with some gene variants acting additively with others, in a multiplicative way or with a compensatory effect [2,3]. Traditional statistical techniques are not appropriate for detecting such effects [4], because they rely on the basic assumption of linear combinations only [5]. Investigation in multifactorial disorders in fact evidenced that non linear interactions are not detected by traditional regression analyses [6].
In particular, psychiatric disorders are characterized by a non mendelian, multifactorial genetic contribution with a number of susceptibility genes interacting with each other [7,8]. In the process of disentangling the contribution of environment versus genes, it has been recently suggested to focus on endophenotypes instead of psychiatric syndromes as a whole [9,10]. One interesting endophenotype is drug response, a field that gained much attention due to the possible clinical applications, ranging from individualized therapy to new drug development [11-14]. However, notwithstanding the promising results observed in the pharmacogenetic field, no single major effect gene was identified, but a variable number of polymorphisms in various genes are supposedly involved in modulating the response and/or side effects to drugs [15-20].
Since our initial study [21] we investigated the short term response to Selective Serotonin Reuptake Inhibitors (SSRIs) and a number of candidate genes, observing both positive and negative associations [22].
However, both the increasing number of genes associated with response and the limitations of traditional methods of analysis are factors requiring the use of new techniques of analysis that more closely resemble to the underlying biological process, i.e. that allows for non-linear interactions.
Neural networks (NN) have been proposed for such studies [1,23,24]. The main advantage of neural networks is that complex non-linear relationships can be modelled, potentially incorporating high-order interactions between predictive variables. This is of particular importance in a complex phenotype such as antidepressant response [22,25].
NN have been used in other fields of medicine, for example to predict cyclosporine dosage in patients after kidney transplantation [26], perspective outcome in AIDS research [27] but also in a genetic analysis in heart disease analysing 10 candidate genes simultaneously [28]. More complex models including gene-environment interactions have been developed [29].
In fact, neural networks proved to outperform single marker association tests, particularly in the case of a complex mode of inheritance or where multiple mutations result in more than one haplotype associated with the disease [25,30,31].
In the present paper we have re-analysed our sample where polymorphism in the transcriptional control region upstream of the 5HTT coding sequence (SERTPR) and in the Tryptophan Hydroxylase (TPH) gene were analized [32], in that paper we observed an association of both polymorphisms with drug response but we could not evaluate their possible non linear interactions. In the present paper we had the aim of evaluating the validity of NN models and of comparing them with traditional statistical techniques (multiple regression and discriminant function analysis).
Methods
Sample
The sample was already described in the original paper [32]. Briefly, two hundred and seventeen depressed inpatients were included in this study (age = 52.11 ± 12.04; onset = 37.97 ± 12.16; female/male: 144/73; bipolars: delusional/non delusional = 40/33, major depressives: delusional/non delusional = 71/73). All patients were evaluated at baseline and weekly thereafter until the sixth week using the 21-item Hamilton Rating Scale for Depression (HAM-D-21) [33] administered by trained senior psychiatrists blind to genetic data and to treatment (fluvoxamine 300 mg daily from day 8 plus pindolol 7.5 mg to one third of the sample). A decrease in HAM-D scores to 8 or less was considered the response criterion. After the procedure had been fully explained to all subjects, informed consent was obtained.
Plasma fluvoxamine levels were determined by high-performance liquid chromatography after 2 weeks of stable 300 mg daily dose [34]. Nine patients with extreme plasma levels (more than 2 standard deviations) were removed from the study in order to avoid biases due to side effects that are present at high doses, also subjects with plasma levels below 20 ng/ml were excluded as this may indicate non compliance, but no cases with such low doses were observed. The influence of both SERTPR and TPH polymorphisms was limited to subjects not taking pindolol [32] therefore we included in the present study the 121 subjects including fluvoxamine alone (81 responders/40 non responders). DNA analysis was performed as described in the original paper [32].
Review of the models used
Multilayer Perceptrons
This is one of the most popular network architecture in use today, though relatively recent [35]. In MLP the units each perform a biased weighted sum of their inputs and pass this activation level through a transfer function to produce their output, and the units are arranged in a layered feedforward topology. The first step of the analysis is the choice of the number of layers and nodes. This is performed searching for a minimum in the error/performance hyperplane. Once the number of layers, and number of units in each layer, have been selected, the weight and threshold of the network must be set so as to minimize the prediction error made by the network. This is the role of the training algorithms. The best-known example of a neural network training algorithm is back propagation. In back propagation, the gradient vector of the error surface is calculated and used to decrease the error. A sequence of such moves (slowing as we near the bottom – epochs) will eventually find a minimum. A large number of epochs with no further improvement in the performance suggests that the optimum set of weights has been reached.
Linear Networks
Originally developed about 60 years ago by Fisher [36], in classification, the hyperplane is positioned to divide the two classes (a linear discriminant function) while in regression, it is positioned to pass through the data. A linear model is typically represented using an NxN matrix and an Nx1 bias vector. The linear network provides a good benchmark against which to compare the performance of your neural networks.
Radial Basis Function Networks
In a radial basis function network the response surface of a single radial unit is a Gaussian (bell-shaped) function, peaked at the center, and descending outwards. RBF networks have advantages and disadvantages over MLPs. First, they can model any non-linear function using a single hidden layer, which removes some design-decisions about numbers of layers. Second, the simple linear transformation in the output layer can be optimized fully using traditional linear modelling techniques, which are fast and do not suffer from problems such as local minima which plague MLP training techniques. However the clumpy approach also implies that RBFs are not inclined to extrapolate beyond known data: the response drops off rapidly towards zero if data points far from the training data are used, therefore they are less reliable for clinical samples such our one. Detailed review of the models are reported elsewhere [24,37].
Model development and selection
An "intent-to-treat" analysis was carried out for all patients who had a baseline assessment and at least 1 assessment after randomization, with the last observation carried forward on the HAM-D. For the current application the inputs to the first layer of the neural network consist of SERTPR and TPH genotypes while the target outputs consist of response status. The network is then trained to attempt to predict response from genotypes. Each node of the input layer of the network is set to a value representing the genotype of each polymorphism. For each polymorphism and for each subject this value is set to genotypes aa, ab or bb. If a marker genotype is missing then the input is assigned a value equal to the average of the values for all subjects in the dataset, however no missing data were present in our sample. The target output for the network is set to 1 or 2 depending on whether the subject is responding or not.
The best network was selected on the basis of its discriminating error and performance, positive and negative predictive values were also reported for each model. This last was expressed as area under the Receiving Operator Characteristic (ROC) Curve. The area under a ROC curve ranges from zero to one, with values close to unity indicating better predictive power; an area of 0.5 indicates that the model is not predicting better than a random choice.
However, one major problem of NN analyses is to establish if the prediction from genotypes is greater than would be expected by chance. If the whole sample is used for training, the network will to some extent "learn to recognise" particular features of each member of the dataset and can use these to predict response in a way which may not reflect any general association between marker genotypes and disease. Generally, this problem is faced by a set of strategies: dividing the dataset (50:50, 80:20...), Jackknife, bootstrapping, cross-validation and so on. However those methods present some disadvantages, in particular if only a part of the data is used to train the network this leads to a loss of power given that subjects in the validating part have different patterns of association between genotypes and drug response.
In order to remedy these problems, in the case of MLP, it has been suggested to perform both training and testing on the entire dataset. The statistical significance of any observed association between outputs and affection status can be estimated using a permutation test [25].
Once the network was defined, a statistic, denoted T, is calculated to compare the outputs for responders and non responders in the same way as an unpaired t statistic, although the statistic is not expected to follow a t distribution under the null hypothesis. Instead, in order to estimate statistical significance a permutation procedure is performed. A large number of replicate data sets are generated from the original data and the obtained network model by randomly permuting genotypes with respect to affection status. For each of these replicate data sets we can then train and test the data set as before, each time calculating T. Since each permuted data set will have only random association between genotype and affection status we obtain N values of T which provide a distribution of T under the null hypothesis. We count the number of times any of these values exceeds the value of T we obtained for the real dataset and denote this number R. Then (R + 1)/(N + 1) provides an unbiased estimate of the statistical significance of the association between genotype and affection status in the real dataset.
In order to estimate a p-value of alpha, one should carry out approximately 10/alpha replicates. Typically, in order to detect association at a significance of 0.01 one would perform 1000 replicates (including the real dataset and 999 permuted datasets). In the case of the present paper we performed 10000 replicates.
Multiple regression and discriminant function analyses were performed to compare the results obtained with the NN strategy with traditional techniques. Responder status was the dependent variable with SERTPR and TPH as independent variables. Genotypes were scored in the following way according to the hypothesis of codominance (SERPR*l/l = 1, SERPR*l/s = 2, SERPR*s/s = 2, TPH*C/C = 1, TPH*C/A = 2, TPH*A/A = 2).
Calculations for the NN selection were performed using STATSOFT (Kernel release 5.5 A). Evaluation was performed using the NNPERM package [31].
Results
MLP showed the best performance and was therefore selected over the other networks (see table 1). The MLP selected over the other models on the basis of error and performance was composed by 1 hidden layer with 7 nodes (Figure 1) after testing about 150 different MLP models. The network showed a very good basic performance (Error 0.430, Performance 0.685).
Table 1 Comparison of NN models. PPV = Predictive Positive Value, NPV = Predictive Negative Value, ROC = area under the ROC curve.
Network type Error Performance sensitivity specificity PPV NPV ROC Youden's J
Linear 0.447 0.636 67.12 56.34 75.97 45.46 0.687 0.23
RBF 0.449 0.691 85.61 35.21 73.10 54.35 0.664 0.21
MLP (1 – 7) 0.439 0.682 77.50 51.20 76.35 52.17 0.731 0.28
Figure 1 MLP composed by 1 hidden layer with 7 nodes used for the analysis.
After, we trained the network with the back propagation algorithm. Initially we used a learning rate of 0.1 (momentum 0.3, noise set to 0), after 5000 epochs we reduced it to 0.01 but after 5000 further epochs we observed no improvement and therefore we finished the selection process and retained the network. Both polymorphisms contributed substantially to the model (SERTPR error= 0.532, ratio = 1.21; TPH error= 0.450, ratio = 1.02). This was expected since both markers were individually associated with response. In detail single marker significance, calculated as simple allelic chi-square, was p = 0.00058 for SERTPR and p = 0.025 for TPH. The classification of subjects in responders and non responders was 77.5 % for responders and 51.2% for non responders. Classification may vary depending from the selected threshold, therefore the area under the ROC curve is a better indicator of performance, in this case the area was 0.731. We also evaluated the predictive power of the network with the SERTPR polymorphism only, in this case the area under the ROC curve was 0.698. We may therefore observe that the add of TPH polymorphism increases the predictive power of the system.
In order to evaluate the significance of the network we applied a permutation test with 10000 replicates. The t statistic for the network was 4.35, it was achieved in 81 out of 10000 simulations yielding a network p-value = (81+1)/(10000+1) = 0.0082.
Finally, we performed a comparison with traditional techniques. A multiple regression analysis showed a significant correlation (p = 0.0004) with a variance explained of 12.5%. The discriminant function analysis correctly classified 34.1 % of responders and 68.1 % of non responders (F = 8.16 p = 0.0005).
Following, we tested the possible impact of clinical variables on response. We included in the model the following variables: Age, age at onset, sex, education, diagnosis, presence of delusional features, recurrence index (defined as number of episodes per year), pindolol augmentation and baseline HAM-D. With those variables no satisfactory network was identified. They were therefore not considered as possible confounding factors in the genetic analysis.
Discussion
This paper reports the first attempt to use NN in pharmacogenetic analyses. We applied this technique to short term antidepressant response in mood disorders. Our analyses suggest that MLP network is the most appropriate for this kind of data, in according with previous observations [25]. The growing number of polymorphisms (about 3.000.000) and the growth of simultaneous techniques such as gene arrays ask for appropriate techniques of analysis. Traditional ones have strong limitations not allowing for non linear interactions and the risk of overfitting in the case of multiple polymorphisms analysed in necessary limited sample sizes. We observed that a relatively simple MLP NN is able to predict response in a way comparable to traditional techniques. The lack of non linear interactions in the simple model we analysed [32] explains why did not observe a marked superiority of NN over traditional analyses. However the most promising result of the strategy we tested in the present paper is the possibility to add a large number of polymorphism to the network and to evaluate the improvement in the prediction, showed by the area under the ROC curve. Moreover the significance of the network can be evaluated with the permutation test [25,31]. Moreover the MLP model we used is quite parsimonious in terms of parameters used (2 input variables, 1 output variable and 1 hidden layer with 7 nodes).
Further developments of this strategy are the inclusion of more detailed information on the phenotypic side. The classification results we obtained are not sufficient in clinical terms were in particular much higher specificities are needed in order to recognize in advance non responders. To reach this target we should consider that we previously observed that some polymorphism influence only part of the whole depressive symptomatology [38]. Further clinical variables should also be considered as reported to influence the short term antidepressant outcome [39], even if previous NN studies failed to identify clinical predictors of antidepressant response [40]. Our analyses are in line with this view, in fact the clinical variables we analysed were not significantly associated with outcome.
The relatively small sample we used does not guarantee against a possible overfitting phenomenon, therefore enlargement of the sample is a priority. Moreover we used the same sample for testing and validating our result, this is not a standard technique [41], this problem is usually faced by a set of strategies such as dividing the dataset, Jackknife, bootstrapping, cross-validation and so on. However those methods present some disadvantages, in particular if only a part of the data is used to train the network this leads to a loss of power in the case that subjects in the validating part have different patterns of association between genotypes and drug response. Therefore in the present paper we performed both training and testing on the entire dataset with the use of a permutation test to validate the results [25]. Another limitation of the present paper is that we compared NN with multiple regression only, other techniques could be tested as well such as set association [30], multifactor dimensionality reduction [42], and logic regression [43].
Differences in allele frequency for different populations have been reported [44]. However our sample was composed of subjects mainly collected in the North of Italy with Italian antecedents for at least two generations, though genetic heterogeneity have been evidenced for some isolate populations (such as Sardinia, not included in our sample) Italy is characterized by a substantial genetic homogeneity [45]. Another caveat is linked to the characteristics of our sample. In fact the Center for Mood Disorders of San Raffaele Hospital is a tertiary structure, therefore we cannot exclude a potential selection bias associated with illness severity and possible extension to outpatients or drug abusers are not warranted [46].
Conclusions
Overall, our findings suggest that neural networks may be a valid technique for the analysis of gene polymorphisms in pharmacogenetic studies. The complex interactions modelled through NN may be eventually applied at the clinical level for the individualized therapy [47].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AS conceived the study, drafted the manuscript and participated in the design of the study and performed the statistical analysis. ES participated in its design and coordination. 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 acknowledge Cristina Lorenzi for the genetic analyses.
==== Refs
Ott J Neural networks and disease association studies Am J Med Genet 2001 105 60 61 11425001
Moore JH The ubiquitous nature of epistasis in determining susceptibility to common human diseases Hum Hered 2003 56 73 82 14614241 10.1159/000073735
Frankel WN Schork NJ Who's afraid of epistasis? Nature genetics 1996 14 371 373 8944011 10.1038/ng1296-371
Schaid DJ Olson JM Gauderman WJ Elston RC Regression models for linkage: issues of traits, covariates, heterogeneity, and interaction Hum Hered 2003 55 86 96 12931047 10.1159/000072313
Ritchie MD White BC Parker JS Hahn LW Moore JH Optimization of neural network architecture using genetic programming improves detection and modeling of gene-gene interactions in studies of human diseases BMC Bioinformatics 2003 4 28 12846935 10.1186/1471-2105-4-28
Moore JH Lamb JM Brown NJ Vaughan DE A comparison of combinatorial partitioning and linear regression for the detection of epistatic effects of the ACE I/D and PAI-1 4G/5G polymorphisms on plasma PAI-1 levels Clin Genet 2002 62 74 79 12123491 10.1034/j.1399-0004.2002.620110.x
Merikangas KR Risch N Will the Genomics Revolution Revolutionize Psychiatry? Am J Psychiatry 2003 160 625 635 12668348 10.1176/appi.ajp.160.4.625
Botstein D Risch N Discovering genotypes underlying human phenotypes: past successes for mendelian disease, future approaches for complex disease Nat Genet 2003 33 Suppl 228 237 12610532 10.1038/ng1090
Lenox RH Gould TD Manji HK Endophenotypes in bipolar disorder Am J Med Genet 2002 114 391 406 11992561 10.1002/ajmg.10360
Waterwort DM Bassett AS Brzustowicz LM Recent advances in the genetics of schizophrenia Cell Mol Life Sci 2002 59 331 348 11915947 10.1007/s00018-002-8426-6
Weber WW Pharmacogenetics 1997 New York, Oxford University Press
Lerer B Macciardi F Pharmacogenetics of antidepressant and mood-stabilizing drugs: a review of candidate-gene studies and future research directions Int J Neuropsychopharmacol 2002 5 255 275 12366879 10.1017/S1461145702002936
Arranz MJ Munro J Birkett J Bolonna A Mancama D Sodhi M Lesch KP Meyer JF Sham P Collier DA Murray RM Kerwin RW Pharmacogenetic prediction of clozapine response Lancet 2000 355 1615 1616 10821369 10.1016/S0140-6736(00)02221-2
Veenstra-VanderWeele J Anderson GM Cook EH Pharmacogenetics and the serotonin system: initial studies and future directions Eur J Pharmacol 2000 410 165 181 11134668 10.1016/S0014-2999(00)00814-1
Alda M Pharmacogenetics of lithium response in bipolar disorder J Psychiatry Neurosci 1999 24 154 158 10212559
Arranz MJ Munro J Osborne S Collier D Kerwin RW Applications of pharmacogenetics in psychiatry: personalisation of treatment Expert Opin Pharmacother 2001 2 537 542 11336604
Basile VS Masellis M McIntyre RS Meltzer HY Lieberman JA Kennedy JL Genetic dissection of atypical antipsychotic-induced weight gain: novel preliminary data on the pharmacogenetic puzzle J Clin Psychiatry 2001 62 Suppl 23 45 66 11603885
Segman R Neeman T Heresco-Levy U Finkel B Karagichev L Schlafman M Dorevitch A Yakir A Lerner A Shelevoy A Lerer B Genotypic association between the dopamine D3 receptor and tardive dyskinesia in chronic schizophrenia Mol Psychiatry 1999 4 247 253 10395214 10.1038/sj.mp.4000511
Segman RH Heresco-Levy U Finkel B Goltser T Shalem R Schlafman M Dorevitch A Yakir A Greenberg D Lerner A Lerer B Association between the serotonin 2A receptor gene and tardive dyskinesia in chronic schizophrenia Mol Psychiatry 2001 6 225 229 11317227 10.1038/sj.mp.4000842
Serretti A Artioli P Predicting response to lithium in mood disorders: role of genetic polymorphisms Am J Pharmacogenomics 2003 3 17 30 12562213 10.1038/sj.tpj.6500139
Smeraldi E Zanardi R Benedetti F Dibella D Perez J Catalano M Polymorphism Within the Promoter of the Serotonin Transporter Gene and Antidepressant Efficacy of Fluvoxamine Molecular Psychiatry 1998 3 508 511 9857976 10.1038/sj.mp.4000425
Serretti A Lilli R Smeraldi E Pharmacogenetics in affective disorders European Journal of Pharmacology 2002 438 117 128 11909602 10.1016/S0014-2999(02)01309-2
Lucek P Hanke J Reich J Solla SA Ott J Multi-locus nonparametric linkage analysis of complex trait loci with neural networks Human Heredity 1998 48 275 284 9748698 10.1159/000022816
Haykin S Neural Networks: A Comprehensive Foundation 1998 , Prentice Hall
Curtis D North BV Sham PC Use of an artificial neural network to detect association between a disease and multiple marker genotypes Ann Hum Genet 2001 65 95 107 11415525 10.1046/j.1469-1809.2001.6510095.x
Camps-Valls G Porta-Oltra B Soria-Olivas E Martin-Guerrero JD Serrano-Lopez AJ Perez-Ruixo JJ Jimenez-Torres NV Prediction of cyclosporine dosage in patients after kidney transplantation using neural networks IEEE Trans Biomed Eng 2003 50 442 448 12723055 10.1109/TBME.2003.809498
Sardari S Sardari D Applications of artificial neural network in AIDS research and therapy Curr Pharm Des 2002 8 659 670 11945163
Yoon Y Song J Hong SH Kim JQ Analysis of multiple single nucleotide polymorphisms of candidate genes related to coronary heart disease susceptibility by using support vector machines Clin Chem Lab Med 2003 41 529 534 12747598 10.1515/CCLM.2003.080
Sing CF Stengard JH Kardia SL Genes, environment, and cardiovascular disease Arterioscler Thromb Vasc Biol 2003 23 1190 1196 12730090 10.1161/01.ATV.0000075081.51227.86
Hoh J Ott J Mathematical multi-locus approaches to localizing complex human trait genes Nat Rev Genet 2003 4 701 709 12951571 10.1038/nrg1155
North BV Curtis D Cassell PG Hitman GA Sham PC Assessing Optimal Neural Network Architecture for Identifying Disease-associated Multi-marker Genotypes using a Permutation Test, and Application to Calpain 10 Polymorphisms Associated with Diabetes Ann Human Genet 2003 67 348 356 12914569 10.1046/j.1469-1809.2003.00030.x
Serretti A Zanardi R Rossini D Cusin C Lilli R Smeraldi E Influence of tryptophan hydroxylase and serotonin transporter genes on fluvoxamine antidepressant activity Molecular Psychiatry 2001 6 586 592 11526473 10.1038/sj.mp.4000876
Hamilton M Development of a rating scale for primary depressive illness British Journal of Social & Clinical Psychology 1967 6 278 296 6080235
Lucca A Lucini V Catalano M Alfano M Smeraldi E Plasma tryptophan to large neutral amino acids ratio and therapeutic response to a selective serotonin uptake inhibitor Neuropsychobiology 1994 29 108 111 8022529
Rumelhart DE McClelland J Parallel Distributed Processing 1986 1 Cambridge, MA, MIT Press
Fisher RA The use of multiple measurements in taxonomic problems Annals of Eugenics 1936 7 179 188
Fausett LV Fundamentals of Neural Networks: Architectures, Algorithms and Applications 1993 , Pearson Higher Education
Serretti A Benedetti F Mandelli L Lorenzi C Pirovano A Colombo C Smeraldi E Genetic dissection of psychopathological symptoms: Insomnia in mood disorders and CLOCK gene polymorphism Am J Med Genet 2003 121B 39 43 12898573 10.1002/ajmg.b.20053
Goodwin FK Jamison KR Manic-depressive illness 1990 New York, Oxford University Press
Winterer G Ziller M Linden M Classification of observational data with artificial neural networks versus discriminant analysis in pharmacoepidemiological studies--can outcome of fluoxetine treatment be predicted? Pharmacopsychiatry 1998 31 225 231 9930637
Hastie T Tibshirani R Friedman J The Elements of Statistical Learning. Data Mining, Inference, and Prediction Springer Series in Statistics 2001 New York, Springer-Verlag
Hahn LW Ritchie MD Moore JH Multifactor dimensionality reduction software for detecting gene-gene and gene-environment interactions Bioinformatics 2003 19 376 382 12584123 10.1093/bioinformatics/btf869
Kooperberg C Ruczinski I LeBlanc ML Hsu L Sequence analysis using logic regression Genet Epidemiol 2001 21 Suppl 1 S626 31 11793751
Kunugi H Hattori M Kato T Tatsumi M Sakai T Sasaki T Hirose T Nanko S Serotonin transporter gene polymorphisms: ethnic difference and possible association with bipolar affective disorder Molecular Psychiatry 1997 2 457 462 9399688 10.1038/sj.mp.4000334
Gasparini P Estivill X Volpini V Totaro A Castellvi-Bel S Govea N Mila M Della Monica M Ventruto V De Benedetto M Stanziale P Zelante L Mansfield ES Sandkuijl L Surrey S Fortina P Linkage of DFNB1 to non-syndromic neurosensory autosomal-recessive deafness in Mediterranean families European Journal of Human Genetics 1997 5 83 88 9195157
Solomon DA Keller MB Leon AC Mueller TI Lavori PW Shea T Coryell W Warshaw M Turvey C Maser JD Endicott J Multiple recurrences of major depressive disorder American Journal of Psychiatry 2000 157 229 233 10671391 10.1176/appi.ajp.157.2.229
Pickar D Rubinow K Pharmacogenomics of psychiatric disorders Trends Pharmacol Sci 2001 22 75 83 11166851 10.1016/S0165-6147(00)01603-5
| 15588300 | PMC539307 | CC BY | 2021-01-04 16:31:07 | no | BMC Med Genet. 2004 Dec 9; 5:27 | utf-8 | BMC Med Genet | 2,004 | 10.1186/1471-2350-5-27 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1566016110.1371/journal.pbio.0030008Research ArticleBioinformatics/Computational BiologyBiophysicsNeuroscienceInsectsDisentangling Sub-Millisecond Processes within an Auditory Transduction Chain Disentangling an Auditory Transduction ChainGollisch Tim [email protected]
1
Herz Andreas M. V
1
2
1Institute for Theoretical Biology, Humboldt University, BerlinGermany2Bernstein Center for Computational Neuroscience, BerlinGermanyMeister Markus Academic EditorHarvard UniversityUnited States of America1 2005 4 1 2005 4 1 2005 3 1 e89 3 2004 21 10 2004 Copyright: © 2005 Gollisch and Herz.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.
Computation Provides a Virtual Recording of Auditory Signaling
Every sensation begins with the conversion of a sensory stimulus into the response of a receptor neuron. Typically, this involves a sequence of multiple biophysical processes that cannot all be monitored directly. In this work, we present an approach that is based on analyzing different stimuli that cause the same final output, here defined as the probability of the receptor neuron to fire a single action potential. Comparing such iso-response stimuli within the framework of nonlinear cascade models allows us to extract the characteristics of individual signal-processing steps with a temporal resolution much finer than the trial-to-trial variability of the measured output spike times. Applied to insect auditory receptor cells, the technique reveals the sub-millisecond dynamics of the eardrum vibration and of the electrical potential and yields a quantitative four-step cascade model. The model accounts for the tuning properties of this class of neurons and explains their high temporal resolution under natural stimulation. Owing to its simplicity and generality, the presented method is readily applicable to other nonlinear cascades and a large variety of signal-processing systems.
Comparing auditory stimuli that give the same neural response within the framework of a computational model, the authors extract intermediary signal-processing steps with sub- millisecond temporal resolution
==== Body
Introduction
Animals and human beings rely on accurate information about their external environment and internal state for proper behavioral reactions. This vital requirement has led to a large variety of highly sophisticated sensory systems [1]. A common feature, though, is the step-by-step conversion of the incoming signal through multiple sequential transformations. In auditory systems, for example, air-pressure fluctuations induce oscillations of mechanical resonators such as the eardrums, basilar membranes, and hair sensilla [2,3,4,5]. These oscillations cause the opening of mechanosensory ion channels in auditory receptor cells [6,7,8]. The resulting electrical currents change the cells' membrane potentials. This, in turn, activates voltage-dependent ion channels that eventually trigger action potentials, which are passed to higher brain areas for further information processing (Figure 1). Each processing step induces a transformation of the stimulus representation that may include rectification, saturation, and temporal filtering. In the mammalian ear, this processing sequence is extended by nonlinear mechanical amplification and feedback [9], which influence the individual processing steps. Similar multi-step sequences of biophysical or biochemical transduction processes underlie the proper function of all sensory and many other signaling systems.
Figure 1 Sequential Processing in the Auditory Transduction Chain
A sequence of several steps transforms an incident sound wave into a neural spike response.
(1) Mechanical coupling. The acoustic stimulus induces vibrations of a mechanical membrane (basilar or tympanic membrane).
(2) Mechanosensory transduction. The deflections cause the opening of mechanosensory ion channels in the membrane of a receptor neuron. Many details of this transduction process are still unknown. The depicted schematic coupling follows the gating-spring model proposed for mechanosensory transduction in hair cells [43].
(3) Electrical integration. The electrical charge due to the transmembrane current accumulates at the cell membrane.
(4) Spike generation. Action potentials are triggered by voltage-dependent currents.
Each of these four steps transforms the signal in a specific way, which may be nearly linear (as for the eardrum response) or strongly nonlinear (as for spike generation, which is subject to thresholding and saturation). In general, the illustrated steps may contain further sub-processes such as cochlear amplification or synaptic transmission between hair cells and auditory nerve fibers. For the auditory periphery of locusts investigated in the present study, this schematic picture resembles anatomical findings [18], which reveal that the receptor neurons are directly attached to the eardrum and that they send their action potentials down the auditory nerve without any further relay stations.
We here show that it is possible to extract fine temporal details of individual processes within such signal-processing chains from observing the output activity alone. This progress results from a new method that extends an experimental strategy well known from measuring threshold curves in neurobiology [10] or applying equivalence criteria in psychophysics [11]: varying stimulus parameters such that the investigated pathway, cell, or system stays at a constant level of output activity. The key to the new method is to compare different stimuli within these measured iso-response sets in such a way that single processing steps can be dissociated. A cascade model is used as a mathematical framework to infer the salient features of the individual processes. This allows us to quantitatively characterize the signal-processing dynamics even under in vivo conditions.
Unlike many classical approaches of systems identification, the method is not based on temporal correlations between the input and output; hence, the time resolution of the method is not limited by the output precision of the system under study. In a spike-based analysis of neural response properties, this allows us to assess the dynamical features of the involved processes with considerably higher resolution than suggested by the spike jitter.
A particularly fine temporal resolution is needed to analyze signal processing in auditory systems that solve complex tasks such as sound localization, echolocation, and acoustic communication [12,13,14,15]. Here, even single receptor cells display extraordinary sub-millisecond precision [14,16,17], with the underlying signal-processing steps involving yet shorter time scales. How these individual steps operate over short times and eventually allow such remarkable precision is largely unknown because of the high vulnerability of the auditory periphery. This calls for methods based on neurophysiological measurements from a remote downstream location such as the auditory nerve, so that the mechanical structures of the ear remain intact.
As a suitable model system to study signal processing in the ear, we chose the auditory periphery of the locust (Locusta migratoria). Its anatomy is well characterized [18], and the auditory nerve is easily accessible for electrophysiological recordings. The nerve contains the axons of the receptor cells. These can be roughly divided into two groups according to their frequency of maximum sensitivity, which lies near 5 kHz for low-frequency receptor cells and around 15 kHz for high-frequency receptor cells. The mechanical structure of the locust system is simpler than that of mammals, as the receptor cells are directly attached to the tympanic membrane, the animal's eardrum. Also, in contrast to the signal amplification in the vertebrate cochlea, there are no known feedback loops, a circumstance which facilitates the modeling. General features of mechanoreceptors, on the other hand, are surprisingly similar across species and are also shared by hair cells in the mammalian inner ear [8].
Results
To analyze signal processing in the locust ear, we performed intracellular recordings in vivo from single receptor-cell axons in the auditory nerve. The stimuli consisted of two short clicks. The clicks were sound-pressure pulses with peak amplitudes A
1 and A
2, respectively, and were separated by a short time interval, Δt (Figure 2A; see also Figure S1 for microphone recordings). For such stimuli, the receptor cell fired at most one action potential per double click; stimulus intensity hardly influenced spike timing, but strongly affected spike probability, as shown in Figure 2B. The response strength may thus be described by the probability that a spike occurs within a certain time window after the two clicks.
Figure 2 Receptor Neuron Responses for Two-Click Stimuli
(A) Stimulus parameters. Acoustic stimuli consisted of two short clicks with amplitudes A
1 and A
2, respectively, separated by a peak-to-peak interval Δt. The clicks were triangular and had a total width of 20 μs. The peak-to-peak interval was generally less than 1.5 ms.
(B) Raster plots of spike responses. Spike times obtained from a single receptor neuron with four different peak intensities (83–86 dB SPL) are shown for 30 runs each. For the different intensities, both click amplitudes were varied while their ratio was kept fixed, with intensity values referring to the larger click amplitude. The inter-click interval in this example was 40 μs. The values of p denote the measured spike probabilities. The inset displays spike times from the strongest sound stimulus at higher magnification. All spikes fall in a temporal window between 4.5 and 5.5 ms after stimulation. Spike times were recorded with a temporal resolution of 0.1 ms. These data illustrate that the response of the receptor cell is well described by the occurrence probability of a single spike in a rather broad time window, for example, between 3 and 10 ms after stimulus presentation. As is often observed for these receptor cells, there is virtually no spontaneous activity.
For fixed time interval Δt, an iso-response set consists of those combinations of A
1 and A
2 that lead to the same predefined spike probability p. Since the spike probability increases with the click amplitudes, A
1 and A
2 can easily be tuned during an experiment to yield the desired value of p (see Materials and Methods). The tuning scheme was applied for stimulus patterns with different relative sizes of the two clicks, so that a multitude of different combinations of A
1 and A
2 corresponding to the same p was obtained. Rapid online analysis of the neural responses and automatic feedback to the stimulus generator made it possible to apply this scheme despite the time limitations of the in vivo experiments.
Figure 3 shows typical examples of such iso-response sets, measured for different time intervals Δt. For each of the three cells displayed, two distinct values of Δt were used. The sets can be used to identify stimulus parameters that govern signal processing at a particular time scale. Most importantly, the iso-response sets exhibit specific shapes that vary systematically with Δt. For short intervals (below approximately 60 μs), the sets generally lie on straight lines, at least for low-frequency receptor cells. High-frequency receptor cells do not display straight lines even at the smallest Δt used in the experiment (40 μs) for reasons that will become apparent later. For long intervals (between approximately 400 and 800 μs, depending on the cell), the iso-response sets fall onto nearly circular curves. Note that in Figure 3C, the iso-response set for Δt = 500 μs deviates from the symmetry between A
1 and A
2. In Figure 3D, the inter-click interval of Δt = 120 μs fell in neither of the two regimes discussed above, and the corresponding iso-response set shows a particularly bulged shape. Recordings from a total of eight cells agree with the observations from the three examples displayed in Figure 3.
Figure 3 Measurements of Iso-Response Sets and Identification of Relevant Stimulus Parameters
(A) Acoustic stimuli. The stimuli consisted of two short clicks with amplitudes A
1 and A
2 that were separated by a peak-to-peak interval Δt, here shown for Δt = 40 μs (upper trace) and Δt = 750 μs (lower trace).
(B–D) Examples of iso-response sets from three receptor cells. Here, as throughout the paper, iso-response sets correspond to a spike probability of 70%. Each panel shows iso-response sets from a single receptor cell for two different values of Δt, one smaller than 100 μs (filled circles) and one larger (open squares). The solid lines denote fits to the data of either straight lines or circles. The values for Δt used in the experiments are indicated in the respective panels. All error measures display 95% confidence intervals. For the short intervals, the data are well fitted by straight lines (A
1 + A
2 = constant). For the long intervals in (B) and (C), circles (A
1
2 + A
2
2 = constant) yield good fits; a slight asymmetry is clearly visible in (C). The data for the intermediate inter-click interval Δt = 120 μs in (D) are not well fitted by either of these shapes. Here, the measured points are connected by a dashed line for visual guidance. Note that in (B) the overall sensitivity of the neuron seems to have changed; the intersections of the straight line and the circle with the x- and y-axis do not match exactly although the stimulus in these cases is the same, a single click. The reason may be either a slow adaptation process or a slight rundown of the recording over the experimental time of around 30 min. However, this does not account for the more prominent differences in shape of the two iso-response sets. These examples demonstrate that on different time scales, different stimulus parameters are relevant for the transduction process, the amplitude A of a sound stimulus for short times and its energy A
2 for long times.
The two prominent shapes of the iso-response sets—straight lines and circles—reflect two different processing steps in the auditory transduction chain. A straight line implies that the linear sum, A
1
+
A
2, of both click amplitudes determines the spike probability and demonstrates that the sound pressure is most likely the relevant stimulus parameter. Such linear summation of the pressure on short time scales is not surprising, considering the mechanical properties of the eardrum; owing to its mechanical inertia, rapidly following stimuli can be expected to superimpose. This interpretation is in agreement with laser-interferometric and stroboscopic observations of the eardrum, which have demonstrated that it reacts approximately linearly to increases in sound pressure [3,19].
For the longer intervals, on the other hand, the iso-response sets are circles to good approximation, indicating that the quadratic sum, or A
1
2
+
A
2
2, now determines the spike probability. It follows that the sound energy, which is proportional to the squared pressure, is the relevant stimulus parameter on this time scale. This quadratic summation represents a fundamentally different way of stimulus integration from that of the linear summation on short time scales and indicates the involvement of a different biophysical process. A process that can mediate stimulus integration over longer intervals is the accumulation of electrical charge at the neural membrane. According to this explanation, the electrical potential induced by a click is proportional to the click's energy; contributions from consecutive clicks are then summed approximately linearly because of the passive membrane properties. This is in accordance with earlier investigations for stationary sound signals that revealed an energy dependence of the neurons' firing rate [20]. We conclude that in between the mechanical vibration of the eardrum and the accumulation of electrical charge at the neural membrane, there is a squaring of the transmitted signal. This squaring may be attributed to the core process of mechanosensory transduction, i.e., the opening of ion channels by the mechanical stimulus.
The above findings motivate the following mathematical model, which describes how a stimulus consisting of two sound clicks is transformed into a spike probability. Within the model, a single click of amplitude A generates a vibration of the tympanum with strength X = c
1
·A, i.e., linear in the amplitude with a proportionality constant c
1. This mechanical vibration leads to a membrane potential, whose effect on the generation of the spike some time T after the click is given by J = c
2·X
2
= c
2
·(c
1
·A)2, i.e., quadratic in the amplitude with an additional proportionality constant c
2. The square follows from the circular shape of the iso-response sets for longer time scales, which indicated that a quadratic operation must take place before the accumulation of charge at the neural membrane. Finally, the spike probability p is given by a yet unknown function p = g(J). As J is the relevant quantity determining spike probability, we also refer to it as “effective stimulus intensity.” The model contains a freedom of scaling; any proportionality constants in J can be absorbed into the function g(J). To simplify the notation, we thus set c
1 = c
2 = 1 and obtain X = A for the strength of the mechanical vibration and J = X
2
= A
2 for the effective stimulus intensity in response to a single click.
Note that in this picture, the mechanical vibration and the membrane potential are each captured by a single quantity that does not describe the time course of the corresponding processes, but rather their integrated strength in response to a click. In general, the conversion of the mechanical vibration into a membrane potential as well as the spike generation are dynamical processes that do not happen at a single moment in time. For simplicity, however, one may think of X as describing the velocity of the mechanical vibration immediately after the click and J as capturing the membrane potential at the time of spike generation.
For the two-click stimulus with amplitudes A
1 and A
2, respectively, we choose the first click to be small enough so that it does not lead to a spike by itself. The measured action potential is thus elicited at some time T after the second click. To derive the model equation for this experimental situation, we divide the time from the first click to spike generation into the period between the two clicks and the period following the second click.
Let us start by focusing on the inter-click interval. After the first click, the mechanical vibration has the strength X
1 = A
1. However, how much electrical charge accumulates during the inter-click interval to influence spike generation at time T after the second click depends on the length Δt of the inter-click interval. This effect is incorporated by a Δt-dependent scaling factor Q(Δt) into the model and results in a first contribution from the first click to spike generation given by J
1
= A
1
2
·Q(Δt). Since Q(Δt) denotes the effect of the first click within the inter-click interval only, it should vanish in the limit of very small Δt.
Let us now consider the remaining time before spike generation. After the second click, the mechanical vibration is due to a superposition of both clicks. For short inter-click intervals, the straight iso-response lines suggest a simple addition of the two click amplitudes; in general, however, the contribution of the first click to the membrane vibration after the second click will again depend on the inter-click interval Δt. This is modeled by a scaling factor L(Δt), i.e., the vibration after the second click has a strength X
2 = A
1· L(Δt) + A
2. Accordingly, the effect of the two-click vibration on the membrane potential at time T after the second click is J
2
= (X
2)2
= (A
1
· L(Δt) + A
2)2. For very small Δt, L(Δt) should approach unity to account for the equal contribution of both clicks for vanishing inter-click intervals. The total effective stimulus intensity is then given by
This quantity determines the spike probability p via the relation p = g(J).
How does this model explain the particular shapes of the iso-response sets in Figure 3? The linear and the circular iso-response sets apparently correspond to the two special cases: (1) L(Δt) = 1 and Q(Δt) = 0 (straight line) and (2) L(Δt) = 0 and Q(Δt) = 1 (circle).
We can therefore regard equation 1 as a minimal model incorporating linear as well as quadratic summation, as suggested by the measured iso-response sets. Based on the experimental data, we expect that the first case is approximately fulfilled for small Δt and the second case in some range of larger Δt. In our biophysical interpretation, the first case means that the two clicks are added at the tympanic membrane (L(Δt) ≈ 1), but the short interval between the two clicks prevents a substantial accumulation of charge from the first click alone (Q(Δt) ≈ 0), as already discussed above. The second case may be found for Δt long enough that the mechanical vibration has already decayed (L(Δt) ≈ 0). The two clicks are then individually squared, i.e., they independently lead to two transduction currents. The currents add up if the time constant of the neural membrane is significantly longer than the inter-click interval (Q(Δt) ≈ 1).
In the two limiting cases, equation 1 is symmetric with respect to A
1 and A
2, reflecting the symmetry of, e.g., the data in Figure 3B. However, for values of Δt where neither of the two cases is strictly fulfilled, this symmetry of the iso-response sets will be distorted, as is noticable for the longer Δt in Figure 3C. Other sets of values for L(Δt) and Q(Δt) may lead to very different iso-response shapes, as in Figure 3D.
Equation 1 presents a self-contained model for click stimuli and is sufficient to analyze the temporal characteristics of the individual steps. It can be interpreted as a signal-processing cascade that contains two summation processes, one linear in the click amplitudes and one quadratic. For click stimuli, the functions L(Δt) and Q(Δt) are thus filter functions associated with the linear and quadratic summation, respectively.
Despite the simple structure of the model, the filters L(Δt) and Q(Δt) can be expected to retain the salient features of the underlying biophysical processes such as frequency content and integration time. In Protocol S1, we show that equation 1 can be obtained in an a posteriori calculation from a generalized cascade model and that this derivation leads to an interpretation of L(Δt) as the velocity of the mechanical vibration and of Q(Δt), at least for large enough Δt, as the time course of the membrane potential following a click. In this generalized model, the input signal is an arbitrary sound pressure wave A(t), and the effective stimulus intensity is a continuous function of time, J(t), which is given by
Here, the input A(t) is first convolved with a temporal filter, l(τ), the result is squared and subsequently convolved with a second filter, q(τ), as depicted in Figure 4. The filters l(τ) and q(τ) have characteristics similar to the click-version filters L(Δt) and Q(Δt), but are not identical to them. Their relations follow from the calculation in Protocol S1. As we here focus on click stimuli, we will use the simpler equation 1 to evaluate the temporal structures of L(Δt) and Q(Δt).
Figure 4 Generalized Cascade Model of the Auditory Transduction Chain
The model is composed of a sequence containing two linear temporal filters, l(τ) and q(t), and two static nonlinear transformations, namely a quadratic nonlinearity and an output nonlinearity g˜(·), which may differ from the nonlinearity g(·) of the click-stimulus model (see Protocol S1). First, the stimulus A(t) is convolved with the filter l(τ) (linear integration). Second, the result is squared (nonlinear transformation). Third, the result of the previous step is convolved with the filter q(τ), yielding the effective stimulus intensity J(t) (linear integration). Fourth, a final transformation g˜ of J(t) (nonlinear transformation) determines the response, which in this generalized model is the time-dependent firing rate r(t). The model thus corresponds to an LNLN cascade. This abstract structure directly follows the sequential configuration of the biophysical processing steps shown in Figure 1.
Note that we interpret equation 1 to yield the spike probability after the second click. If the first click is large and the second small, however, the first click alone may account for some of the observed spikes; clearly this is the case when the second click vanishes. This is not captured by equation 1, and one might expect that, for large values of A
1, these additional spikes lead to measured values of A
2 that are slightly smaller than expected for a circular iso-response set. The data in Figure 3, however, suggest that this effect is small and not picked up by our experiment. Nevertheless, for the following quantitative study, we will keep the first click always on a level where the click by itself does not contribute substantially to the spike probability.
The previous experiment showed that the separate effects of the two summation processes can be discerned for short and long time intervals. For intermediate Δt, however, their dynamics may largely overlap. Is it nevertheless possible to design an experiment that directly reveals the whole time course of the mechanical vibration L(Δt) and the electrical integration Q(Δt)? This would provide a parameter-free description of both processes and advance the quantitative understanding of the auditory transduction dynamics. To reach this goal, we again measure iso-response sets. As before, we exploit that for fixed Δt, any pair of click amplitudes (B
1, B
2) should result in the same spike probability p as the pair (A
1, A
2) as soon as J(A
1, A
2) = J(B
1, B
2). It is this straightforward relation that allows us to determine both L(Δt) and Q(Δt) independently of each other. In fact, some appropriate set of measurements that fulfill the iso-response relation is all that is needed to calculate L(Δt) and Q(Δt). Illustrating this concept, we now proceed with a particularly suited choice of stimulus patterns, which keeps the mathematical requirements for the calculation at a minimum. For each Δt, we measure two different iso-response stimuli, and as a key feature, one of these has a “negative” second click, i.e., a sound-pressure pulse pointing in the opposite direction as the first click, as depicted in Figure 5A. Mathematically, this choice of stimulus patterns leads to two simple equations for the two unknowns L(Δt) and Q(Δt), which can be solved explicitly, as explained in Materials and Methods. By repeating such double measurements for different values of Δt, the whole time course of L(Δt) and Q(Δt) is obtained.
Figure 5 Temporal Structure of the Mechanical Oscillation and Electrical Integration
(A) Stimulus patterns. Two clicks were presented, separated by a time interval Δt. The first click (amplitude A
1) was held constant throughout this experiment. The second click was presented in the same direction as the first click (solid line, amplitude A
2) or in the opposite (“negative”) direction (dashed line, amplitude Ã
2). The click amplitudes A
2 and Ã
2 were adjusted to fall in the desired iso-response set.
(B–G) Mechanical oscillation and electrical integration of a high-frequency (B and E) and two low-frequency (C and F, and D and G, respectively) receptor neurons.
(B–D) Time course of the eardrum vibration. The individual values (circles) were calculated from the measured values of A
2 and Ã
2 for each Δt. The results are compared with a theoretical curve from a damped harmonic oscillator (solid line) with fundamental frequency f and decay time constant τ
dec fitted to the data.
(E–G) Time course of the electrical integration process. The measured data are compared to an exponential fit (solid line) with a time constant τ
int.
Figure 5 shows examples of L(Δt) and Q(Δt) for three different cells. L(Δt) displays strong oscillatory components, as was observed for all cells. This property presumably reflects the eardrum's oscillation at the attachment site of the receptor cell. The detailed temporal structure of L(Δt) now allows us to investigate the salient features of this oscillation. To quantify our findings, we fit a damped harmonic oscillation to the measured data for L(Δt) and extract the fundamental frequency as well as the decay time constant. We can use these values to predict the neuron's characteristic frequency (the frequency of highest sensitivity) and the width of its frequency-tuning curve. Figure 6 shows the comparison of these predictions with traditional measurements of the tuning curves for all 12 cells measured under this experimental paradigm with sufficient sampling to extract L(Δt). The remarkable agreement confirms that the new analysis faithfully extracts the relevant, cell-specific properties of the transduction sequence. The correspondence between the tuning characteristics and the filter L(Δt) also explains why high-frequency receptor cells do not feature straight lines for their iso-response sets even at the shortest inter-click interval (40 μs) used in the experiment. For those cells, L(Δt) decays rapidly, thus not allowing access to the region where L(Δt) ≈ 1.
Figure 6 Predictions of Tuning Characteristics
(A) Tuning curves for the same two cells as in Figure 5B and 5E, and 5C and 5F, respectively. The data show the intensity required to drive a receptor cell at a firing rate of 150 Hz for different sound frequencies in the range of 1 to 40 kHz. The characteristic frequency f
CF is determined as the minimum of the tuning curve, and the tuning width Δf
3dB as the width of the curve 3 dB above the minimum value.
(B) Comparison of the predicted and measured characteristic frequency and the tuning width. The predictions were obtained from the fundamental frequency and decay time constant of the measured filter L(Δt); the measured values are taken from the tuning curves as in (A) (n = 12). The encircled data points correspond to the three examples shown in Figure 5. The width of the tuning curves is notoriously difficult to assess quantitatively, as it depends sensitively on an accurate determination of the intensity minimum of the tuning curve. This contributes strongly to the differences of the tuning-width values.
The short initial rise phase of the measured Q(Δt) in Figure 5E and 5F illustrates the rapid buildup of the membrane potential after a click. The exponential decay following this phase suggests that the accumulated electrical charge decays over time owing to a leak conductance. Previously, the time constant could not be measured because of difficulties in obtaining recordings from the somata or dendrites of the auditory receptor cells. Using our new method, we find time constants in the range of 200 to 800 μs. These values are small compared to time constants in more central parts of the nervous system, reflect the high demand for temporal resolution in the auditory periphery, and explain the high coding efficiency of the investigated receptor neurons under natural stimulation [21].
In most of our recordings, the temporal extent of the filter L(Δt) was considerably smaller than that of Q(Δt). This usually leads to a region around a Δt of 400–800 μs, depending on the specific cell, where L(Δt) ≈ 0 and Q(Δt) is still near unity. These findings correspond to the circular iso-response sets of the initial experiment.
Towards very small Δt, on the other hand, the data show that Q(Δt) usually decreases strongly. As explained earlier, this is expected from the linear iso-response sets, and it is observed exemplarily in the data shown in Figure 5E and 5F. In addition, the first few 100 μs of the data may show considerable fluctuations of Q(Δt) for some recordings, as in Figure 5G. Different effects may influence this early phase of Q(Δt). (1) The electrical potential might be shaped by further dynamics in addition to the low-pass properties of the neural membrane, such as inactivation of the transduction channels or electrical resonances as found in some hair cells [6]. (2) The fluctuations could reflect the oscillatory influx of current following from the oscillation of the eardrum. In other words, the low-pass filtering of the neural membrane may not be strong enough to quench all oscillatory components of the transduction currents. The resulting effect on the filter Q(Δt)—though too small to be picked up reliably by the present experiments—can be observed in simulations of the processing cascade, see Figure S2. At present, we cannot distinguish between these two interpretations. More detailed future experiments, however, may allow a quantitative test of these hypotheses.
Measuring the mechanical and electrical response dynamics, L(Δt) and Q(Δt), completes the model. In order to test its validity and suitability to make quantitative predictions, we investigated the model's performance on a different class of stimuli, namely combinations of three short clicks. Having measured the required values for L(Δt) and Q(Δt) with two-click stimuli as in the previous experiment (see Figure 5), we now ask the following question: if we keep the first two clicks small enough that they do not lead to a spike response, can we predict the size of the third click required to reach a given spike probability? We can use the measured values of L(Δt) and Q(Δt) to calculate these predictions and experimentally test them by performing a series of three-click iso-response measurements. This experiment was performed on three different cells; one cell featured an unusually high response variability, and results from the other two cells are shown in Figure 7. The agreement between the predicted and the true click amplitudes shows that the model yields quantitatively accurate results.
Figure 7 Model Predictions for Three-Click Stimuli
(A) Stimulus patterns. The stimuli consisted of three clicks with amplitudes A
1, A
2, and A
3 that were separated by time intervals Δt
1 and Δt
2, respectively. The second and third clicks were either given in the same or opposite (“negative”) direction as the first click. A
1 and A
2 were set equal and held constant, and A
3 was adjusted to yield a spike probability of 70%. The following pairs of time intervals (Δt
1, Δt
2) were applied: (100 μs, 100 μs), (100 μs, 200 μs), and (200 μs, 100 μs).
(B and C) Predicted and measured amplitudes of the third click for two different cells. Predictions were made after L(Δt) and Q(Δt) had been measured with two-click experiments such as in Figure 5. The comparison between predicted and measured values for A
3 therefore contains no free parameters. The model equation for three-click stimuli is presented in Materials and Methods. As demonstrated by these data, the model allows quantitatively accurate predictions.
Discussion
We have presented a novel technique to disambiguate single processing steps within a larger sensory transduction sequence and to analyze their detailed temporal structures. Our approach is based on measuring particular iso-response sets, i.e., sets of stimuli that yield the same final output, and on specific quantitative comparisons of such stimuli to dissociate the individual processes. For the investigated auditory transduction chain in the locust ear, this strategy led to a precise characterization of two consecutive temporal integration processes, which we interpret as the mechanical resonance of the eardrum and the electrical integration of the attached receptor neuron. The method revealed new details of these processes with a resolution far below 1 ms. The results for the time course of the mechanical resonance agree with traditional measurements of tuning curves and show the decay of the oscillation with a temporal precision much higher than expected from the jitter of the measured output signal, the spikes. The time constants of the electrical integration that were extracted from the data had not been accessible by other means.
The analysis resulted in a four-step model of auditory transduction in locusts. The model comprises a series of two linear filters and two nonlinear transformations. The quadratic nonlinearity that separates the two linear filters suggests that the mechanosensory transduction can be described by an energy-integration mechanism, as the squared amplitude corresponds to the oscillation energy of the tympanum. This quadratic form was derived from the circular shape of the iso-response sets for longer time scales Δt and is in accordance with the energy-integration model that was found to capture the sound-intensity encoding of stationary sound signals in these cells [20]. Furthermore, the direct current component of the membrane potential in hair cells is also proportional to sound energy [22], and in psychoacoustic experiments, energy integration accounts for hearing thresholds [23,24,25,26]. However, a recent analysis of response latencies in auditory nerve fibers and auditory cortex neurons in cats suggests an integration of the pressure envelope for determining thresholds [27]. This effect may be attributable to the synapse between the hair cell and the auditory nerve fiber in the mammalian ear. In the locust ear, this synapse does not exist, as the fibers are formed by the axons of the receptor neurons themselves.
Although the quadratic nonlinearity is fully consistent with our data, there is a second possibility within the general cascade model, equation 2, namely, squaring after rectification. From a biophysical point of view, this would be expected if the mechanosensory ion channels can only open in one direction. Based on the current data, we cannot distinguish between these two possibilities. As the two scenarios should lead to slightly different response characteristics, future high-resolution experiments should be able to resolve this question.
The linear filters L(Δt) and Q(Δt) were interpreted as the mechanical oscillation of the tympanum and the electrical integration at the neural membrane. Their oscillatory and exponential decay characteristics, respectively, support this view. In principle, however, other processes may well contribute to these characteristics, e.g., electrical resonances as seen in hair cells of the turtle and bullfrog [6,28]. These electrical amplification processes would be expected to influence the filter Q(Δt), but our data generally provide little evidence for such effects. Deviations from the exponential decay characteristics in Q(Δt) may in part be attributable to the oscillatory influx of charge resulting from the tympanic vibration. This may lead to a small oscillatory component in the early phase of the filter (cf. Protocol S1; Figure S2).
The mechanical coupling in the first step of our model is linear. This is in accordance with mechanical investigations of the tympanum using laser interferometry [3] and stroboscopic measurements [19]. As the short clicks used in our study produce reliable spiking responses only at high sound pressure, however, we cannot exclude the influence of nonlinear coupling at low sound pressure, which has been hypothesized on the basis of distortion-product otoacoustic emissions [29]. In addition, the mechanical properties of the tympanum seem to change slightly under prolonged stimulation and give rise to mechanical adaptation effects with time scales in the 100-ms range [30]. Spike-frequency adaptation also adds a nontrivial feedback term to the minimal feedforward model of Figure 4. Similarly, specific potassium currents and sodium-current inactivation induced by sub-threshold membrane potential fluctuations may complicate the transduction dynamics for more general inputs, but do not leave a signature in the present click-stimulus data.
The model was quantitatively investigated by using combinations of short clicks. The particular structure of these stimuli allowed a fairly simple mathematical treatment. The derivation of equation 1 relied on capturing the mechanical vibration and the membrane potential, respectively, by single quantities in each time period following a click. This was possible because of the expected stereotypic evolution of the dynamic variables during the “silent phases” between and after the clicks. A generalization to arbitrary acoustic stimuli would require a more elaborate model in the form of equation 2 as well as extensions that account for neural refractoriness and adaptation.
Besides its applicability under in vivo conditions, the presented framework has several advantageous properties. First, the method effectively decouples temporal resolution on the input side from temporal precision on the output side by focusing on spike probabilities. In all our measurements, for example, spike latencies varied by about 1 ms within a single recording set owing to cell-intrinsic noise (see Figure 2). Still, we were able to probe the system with a resolution down to a few microseconds. This would not have been possible using classical techniques such as poststimulus time histograms, reverse correlation, and Wiener-series analysis. All these methods are intrinsically limited by the width of spike-time jitter and thus cannot capture the fine temporal details of rapid transduction processes. With our method, the resolution is limited only by the precision with which the sensory input can be applied. For the investigated system, the achievable temporal resolution thus increases by at least two orders of magnitude.
Second, the method is robust against moderate levels of spontaneous output activity, as this affects all stimuli within one iso-response set in the same way. Methods that require measurements at different response levels, on the other hand, are likely to be systematically affected because the same internal noise level may have a different influence at different levels of output activity.
Finally, in many input–output systems, the last stage of processing can be described by a monotonic nonlinearity. Here, this is the relation between the effective stimulus intensity J and the spike probability p = g(J), which includes thresholding and saturation. By always comparing stimuli that yield the same output activity, our analysis is independent of the actual shape of g(J). Preceding integration steps may thus be analyzed without any need to model g(J). This feature is independent of the specific output measure and applies to spike probabilities, firing rates, or any other continuous output variable.
Let us also note that the method does not require that the time scales of the individual processes be well separated. For the studied receptor cells, mechanical damping was on average about two times faster than electrical integration, and even for cells with almost identical time constants, iso-response measurements led to high-quality data and reliable parameter fits. Nor is the method limited to particularly simple nonlinearities. All that is needed are solid assessments of the iso-response sets. Mathematically, it is straightforward to substitute some or all of the analytical treatments of this work by numerical approaches, if required by the complexity of the identified signal-processing steps. This extension allows one to use a general parametrization of the full processing chain when the nonlinear transformation cannot be estimated from iso-response sets at large and small Δt. Instead, performing more than the two measurements at each intermediate Δt in the second experiment (see Figure 5) will provide additional information that can be exploited to improve the numerical estimates of the nonlinearity.
As in many other approaches of nonlinear systems identification, the development of a quantitative model relies on the prior determination of the appropriate cascade structure. Unfortunately, there is no universal technique for doing so. In many cases, intuition is required to find suitable models, which should eventually be tested by their predictive power. In the present case, the findings of characteristic shapes of the iso-response sets gave a clear signature of two distinct linear filters with a sandwiched quadratic nonlinearity. In addition, this structure was supported by its amenability to straightforward biophysical interpretation. Generalizing our results, specific iso-response sets may aid structure identification in conjunction with a priori anatomical and physiological knowledge. Once the cascade structure is established, the individual constituents can be quantitatively evaluated by specific comparisons of iso-response stimuli. Comparing responses to clicks in positive and negative directions as in this study is in essence similar to the approach used by Gold and Pumphrey [31], who evaluated the perceptual difference between short sine tones with coherent phase relations and sine tones that contained phase-inverted parts in order to estimate the temporal extent of the cochlear filters.
A yet open problem is the inclusion of feedback components. The present approach relies on the feedforward nature of the system to disentangle the individual processing steps. In particular cases, however, the iso-response approach may also aid in separating feedforward and feedback contributions, namely, when the feedback depends purely on the last stage of the processing cascade [30]. In this situation, iso-response measurements lead to a constant feedback contribution, and the analysis of the feedforward components may be carried out as in the present case. The experiment may then be repeated for different output levels to map out the feedback characteristics.
The feedforward model that we have proposed here for the auditory transduction chain has the form of an LNLN (where “L” stands for linear and “N” stands for nonlinear) cascade, composed of two linear temporal integrations and two nonlinear static transformations [32]. Similar signal-processing sequences combining linear filters and nonlinear transformations are ubiquitous at all levels of biological organization, from molecular pathways for gene regulation to large-scale relay structures in sensory systems. In neuroscience, applications range from the sensory periphery, including frog hair cells [33], insect tactile neurons [34], and the mammalian retina [35,36,37], over complex cells in visual cortex [38,39], to psychophysics [40]. These studies are restricted to models that contain a single nonlinear transformation, corresponding to NL, LN, or LNL cascades [32,41]. An extension of these analyses was presented by French et al. [42], who derived an NLN cascade for fly photoreceptors.
Complementary to the correlation techniques underlying the parameter estimations in those models, the method presented in this work provides a new way of quantitatively evaluating and testing cascade models. The increased complexity of the LNLN cascade identified in the present case was made accessible by invoking particular iso-response measurements, and a higher temporal resolution was achieved by focusing on how spike probabilities depend on the temporal stimulus structure instead of relying on temporal correlations between stimulus and response.
Our experimental technique will be most easily applicable to systems whose signal processing resembles the cascade structure investigated here. The general concept of combining different measurements from within one iso-response set covers, however, a much larger range of systems. With increasingly available high-speed computer power for online analysis and stimulus generation, this framework therefore seems well suited to solve challenging process-identification tasks in many signal-processing systems.
Materials and Methods
Electrophysiology
We performed intracellular recordings from axons of receptor neurons in the auditory nerve of adult Locusta migratoria. Details of the preparation, stimulus presentation, and data acquisition are described elsewhere [20]. In short, the animal was waxed to a Peltier element; head, legs, wings, and intestines were removed, and the auditory nerves, which are located in the first abdominal segment, were exposed. Recordings were obtained with standard glass microelectrodes (borosilicate, GC100F-10, Harvard Apparatus, Edenbridge, United Kingdom) filled with 1 mol/l KCl, and acoustic stimuli were delivered by loudspeakers (Esotec D-260, Dynaudio, Skanderborg, Denmark, on a DCA 450 amplifier, Denon Electronic, Ratingen, Germany) ipsilateral to the recorded auditory nerve. The reliability of the sound signals used in this study was tested by playing samples of the stimuli while recording the sound at the animal's location with a high-precision microphone (40AC, G.R.A.S. Sound and Vibration, Vedbæk, Denmark, on a 2690 conditioning amplifier, Brüel and Kjær, Langen, Germany). See Figure S1 for example recordings.
Spikes were detected online from the recorded voltage trace with the custom-made Online Electrophysiology Laboratory software and used for online calculation of spike probabilities and automatic tuning of the sound intensities. The measurement resolution of the timing of spikes was 0.1 ms. During the experiments, the animals were kept at a constant temperature of 30 °C by heating the Peltier element. The experimental protocol complied with German law governing animal care.
Measurement of iso-response sets
Since the spike probability p of the studied receptor neurons increases monotonically with stimulus intensity, parameters of iso-response stimuli corresponding to the same value of p can be obtained by a simple online algorithm that tunes the absolute stimulus intensity. For fast and reliable data acquisition, we chose p = 70%. The response latency of the neurons varied by 1–2 ms, so that spike probabilities could be assessed by counting spikes over repeated stimulus presentations in a temporal window from 3 to 10 ms after the first click.
In the first set of experiments, stimulus patterns were defined by fixed ratios of A
1 and A
2, and the tuning was achieved by adjusting the two amplitudes simultaneously. The ratios were chosen so that the angles α in the A
1–A
2 plane given by tanα = A
2/A
1 were equally spaced. In the second set of experiments, A
1 was kept fixed, and only A
2 was adjusted; similarly, in the three-click experiments, only A
3 was adjusted. In the following, the intensity I always refers to the peak amplitude A
max of the stimulus pattern, measured in decibel sound pressure level (dB SPL),
For each stimulus, the absolute intensity I
70 corresponding to a spike probability of 70% was determined online in the following way. Beginning with a value of 50 dB SPL, the intensity was raised or lowered in steps of 10 dB, depending on whether the previous intensity gave a spike probability lower or higher than 70% from five stimulus repetitions. This was continued until rough upper and lower bounds for I
70 were found. From these, a first estimate of I
70 was obtained by linear interpolation. Seven intensity values in steps of 1 dB from 3 dB below to 3 dB above this first estimate were then repeated 15 times. From the measured spike probabilities, a refined estimate of I
70 was obtained by linear regression. Nine intensities from 4 dB above to 4 dB below this value were repeated 30 times (in some experiments 40 times). The final estimate of I
70 was determined offline from fitting a sigmoidal function of the form
with parameters α and β to these nine intensity-probability pairs. This relation between p and I was then inverted to find the intensity and thus the absolute values of the amplitudes that correspond to p = 0.7.
Extraction of L(Δt) and Q(Δt) from iso-response sets
The response functions L(Δt) and Q(Δt) can be obtained independently of each other by combining the results from different measurements within one iso-response set. Here, we derive explicit expressions based on a specific choice of stimuli that are particularly suited for our system. Two measurements are needed to obtain both L(Δt) and Q(Δt) for given time interval Δt. Each stimulus consists of two clicks. The first click has a fixed amplitude A
1; the amplitude A
2 of the second click at time Δt later is adjusted so that a predefined spike probability p is reached. For the second measurement, the experiment is then repeated with a “negative” second click, i.e., a click with an air-pressure peak in the opposite direction from the first click. The absolute value of this click amplitude is denoted by Ã
2. We thus find the two pairs (A
1,
A
2) and (A
1, Ã
2) as elements of an iso-response set. Since the spike probability increases with the effective stimulus intensity J, equal spike probability p implies equal J. The two pairs (A
1,
A
2) and (A
1, Ã
2) therefore correspond to the same value of J. According to the model, equation 1, the click amplitudes thus satisfy the two equations
Setting the two right sides equal to each other, we obtain
or
The first solution of this mathematical equation, Ã
2 = − A
2, does not correspond to a physical situation as both A
2 and Ã
2 denote absolute values and are therefore positive. The remaining, second solution reads
Solving for L(Δt), we obtain
Substituting L(Δt) from equation 10 in equation 5 or equation 6, we find
This yields
with c = J/A
1
2. As we keep A
1 and J constant throughout the experiment, this determines Q(Δt) up to the constant c. It can be inferred from an independent measurement with a single click: by setting A
1 = 0 in equation 5, we see that J corresponds to the square of the single-click amplitude that yields the desired spike probability. Alternatively, c can be estimated from the saturation level of Q(Δt) for large Δt, as was done in the present study.
The specific form of the effective stimulus intensity, equation 1, led to particularly simple expressions for the response functions L(Δt) and Q(Δt); see equation 10 and equation 12, respectively. Other nonlinearities may result in more elaborate expressions or implicit equations, but this technical complication does not limit the scope of the presented approach.
Data fitting
The datasets for L(Δt) were fitted with velocity response functions of a damped harmonic oscillator
where ω and δ were optimized for minimizing the total squared error. From these, the fundamental frequency f and the decay time constant τ
dec were determined as f = ω/(2π) and τ
dec = 1/δ. A simpler fit function of the form
led to essentially indistinguishable results for f and τ
dec.
The resonance frequency, which corresponds to the characteristic frequency, f
CF, of the tuning curve, and the tuning width, Δf
3dB, can be predicted from the fitted values of ω and δ according to the theory of harmonic oscillators:
The datasets for Q(Δt) were fitted with an exponential decay
where the parameters a, τ
int, and c were adjusted. Here, only data points for Δt > 150 μs were taken into account, as Q(Δt) initially shows a rising phase. The obtained value for c was used to determine the constant J/A
1
2 in equation 12.
For comparing these predicted values with measurements, the minimum and width of the tuning curves (see Figure 6A) were determined by fitting a quadratic function to the five data points closest to the data point with smallest intensity.
Model predictions for three-click stimuli
For stimuli consisting of three clicks with amplitudes A
1, A
2, and A
3 that are separated by time intervals Δt
1 and Δt
2, respectively (see Figure 7A), an approximate equation for the effective stimulus intensity J can be derived in the following way: The first click induces a tympanic vibration proportional to A
1 and a membrane potential proportional to A
1
2. Following the second click, the tympanic deflection has become A
1· L(Δt
1) and is augmented by A
2. This yields a membrane potential proportional to (A
1·L(Δt
1) + A
2)2. After the third click, the tympanic deflection has evolved to A
1·L(Δt
1 + Δt
2) + A
2·L(Δt
2) so that the membrane potential is increased by (A
1·L(Δt
1 + Δt
2) + A
2·L(Δt
2) + A
3)2. Summing up the different contributions and approximating the influence of the inter-click intervals on the membrane potential by appropriate factors of Q, we find for the effective stimulus intensity
The value of J for a predefined spike probability can be measured from a single-click experiment by setting A
1
= A
2 = 0 and tuning A
3 until the desired spike probability is reached. After having measured L(Δt) and Q(Δt) from two-click experiments, the above equation can be used to predict the amplitude A
3 needed to reach this predefined spike probability for any combination of A
1, A
2, Δt
1, and Δt
2.
Supporting Information
Protocol S1 General Cascade Model
(50 KB PDF).
Click here for additional data file.
Figure S1 Examples of Click Stimuli
The four panels show different examples of stimuli used in our study. Each panel illustrates the computer-generated pulse signal that drives the loud speaker (upper trace) and the resulting air-pressure fluctuations as measured with a high-precision microphone at the site of the animal's ear (lower trace). The computer-generated clicks are triangular with a total width of 20 μs. The stimuli shown are (A) a single click, (B) a double click with a peak-to-peak interval Δt = 50 μs, (C) a double click with Δt = 500 μs, and (D) another double click with Δt = 500 μs whose second click points in the oppositve (“negative”) direction. The measurements of air-pressure fluctuations indicate a slight broadening of the click width and some residual vibrations, but they nevertheless present a good approximation of the sharp original pulses.
(10 KB PDF).
Click here for additional data file.
Figure S2 Simulation and Analysis of the General Cascade Model in Response to Two-Click Stimuli
The general cascade model, equation 2 in the main text, was used with filters modeled as l(t) = sin(2πft)exp(−t/τ
dec) and q(t) = exp(−t/τ
int). The parameters were taken from the first two cells presented in detail in the main text: f = 14.5 kHz, τ
dec = 100 μs, and τ
int = 300 μs for Cell 1 (left column) and f = 5.1 kHz, τ
dec = 154 μs, and τ
int = 590 μs for Cell 2 (right column).
(A and B) Responses of tympanic vibration. x(t) denotes the signal after application of the linear filter l(t), arbitrary units, for positive second click (solid line) and negative second click (dashed line). Inter-click intervals in these two shown examples were Δt = 80 μs for Cell 1 and Δt = 130 μs for Cell 2.
(C and D) Corresponding responses of J(Δt). The second click was tuned so that the maximum of J(Δt) was equal for positive and negative second clicks. This required click amplitudes of size 1.92 and −2.49 relative to the first click for Cell 1 and 2.09 and −1.27 for Cell 2.
(E–H) Filters L(Δt) and Q(Δt) extracted according to equation 1 in the main text from tuning the maximum of J(Δt) for many different values of Δt (gray dots). The parameters f, τ
dec, and τ
int indicated in the plots were obtained by fitting a damped harmonic oscillator and an exponential function to L(Δt) and Q(Δt), respectively (black lines). The initial part of Q(Δt) shows small fluctuations that result from the oscillatory influx of charge following the tympanic vibrations. In (G), a magnified view of the initial section is shown in the inset.
(138 KB PDF).
Click here for additional data file.
We thank C. D. Brody, M. J. Chacron, G. M. Klump, K. P. Körding, C. K. Machens, I. Segev, M. Stemmler, and H. Wagner for stimulating discussions and J. Benda, C. K. Machens, and H. Schütze for technical assistance with the experiments. This work was supported by Boehringer Ingelheim Fonds (TG) and by the Deutsche Forschungsgemeinschaft through SFB 618 (AH).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. TG and AH conceived and designed the experiments. TG performed the experiments and analyzed the data. TG and AH wrote the paper.
Citation: Gollisch T, Herz AVM (2005) Disentangling sub-millisecond processes within an auditory transduction chain. PLoS Biol 3(1): e8.
Abbreviation
dB SPLdecibel sound pressure level
==== Refs
References
Hudspeth AJ Logothetis NK Sensory systems Curr Opin Neurobiol 2000 10 631 641 11084326
Robeles L Ruggero MA Mechanics of the mammalian cochlea Physiol Rev 2001 81 1305 1352 11427697
Schiolten P Larsen ON Michelsen A Mechanical time resolution in some insect ears J Comp Physiol 1981 143 289 295
French AS Transduction mechanisms of mechanosensilla Annu Rev Entomol 1988 33 39 58
Robert D Göpfert MC Novel schemes for hearing and orientation in insects Curr Opin Neurobiol 2002 12 715 720 12490264
Hudspeth AJ The cellular basis of hearing: The biophysics of hair cells Science 1985 230 745 752 2414845
Hill KG The physiology of locust auditory receptors. I. Discrete depolarizations of receptor cells J Comp Physiol A 1983 152 475 482
Gillespie PG Walker RG Molecular basis of mechanosensory transduction Nature 2001 413 194 202 11557988
Ricci A Active hair bundle movements and the cochlear amplifier J Am Acad Audiol 2003 14 325 338 14552426
Evans EF Keidel WD Neff WD Cochlear nerve and cochlear nucleus Auditory system. Volume 5.2, Handbook of sensory physiology 1975 Berlin Springer 1 108
Jameson D Hurvich LM editors Visual psychophysics. Volume 7.4, Handbook of sensory physiology 1972 Berlin Springer-Verlag 812
Lord Rayleigh On our perception of the direction of a sound source Nature 1876 14 32 33
Carr CE Processing of temporal information in the brain Annu Rev Neurosci 1993 16 223 243 8460892
Grothe B Klump GM Temporal processing in sensory systems Curr Opin Neurobiol 2000 10 467 473 10981615
Machens CK Schütze H Franz A Kolesnikova O Stemmler MB Single auditory neurons rapidly discriminate conspecific communication signals Nat Neurosci 2003 6 341 342 12652305
Köppl C Phase locking to high frequencies in the auditory nerve and cochlear nucleus magnocellularis of the barn owl, Tyto alba
J Neurosci 1997 17 3312 3321 9096164
Mason AC Oshinsky ML Hoy RR Hyperacute directional hearing in a microscale auditory system Nature 2001 410 686 690 11287954
Gray EG The fine structure of the insect ear Philos Trans R Soc London B Biol Sci 1960 243 75 94
Breckow J Sippel M Mechanics of the transduction of sound in the tympanal organ of adults and larvae of locusts J Comp Physiol A 1985 157 619 629 3837103
Gollisch T Schütze H Benda J Herz AVM Energy integration describes sound-intensity coding in an insect auditory system J Neurosci 2002 22 10434 10448 12451143
Machens CK Stemmler MB Prinz P Krahe R Ronacher B Representation of acoustic communication signals by insect auditory receptor neurons J Neurosci 2001 21 3215 3227 11312306
Dallos P Response characteristics of mammalian cochlear hair cells J Neurosci 1985 5 1591 1608 4009248
Garner WR The effect of frequency spectrum on temporal integration of energy in the ear J Acoust Soc Am 1947 19 808 814
Plomp R Bouman MA Relation between hearing threshold and duration for tone pulses J Acoust Soc Am 1959 31 749 758
Zwislocki J Theory of temporal auditory summation J Acoust Soc Am 1960 32 1046 1060
Florentine M Fastl H Buus S Temporal integration in normal hearing, cochlear impairment, and impairment simulated by masking J Acoust Soc Am 1988 84 195 203 3411048
Heil P Neubauer H A unifying basis of auditory thresholds based on temporal summation Proc Natl Acad Sci U S A 2003 100 6151 6156 12724527
Lewis RS Hudspeth AJ Voltage- and ion-dependent conductances in solitary vertebrate hair cells Nature 1983 304 538 541 6603579
Kössl M Boyan GS Acoustic distortion products from the ear of a grasshopper J Acoust Soc Am 1998 104 326 335
Gollisch T Herz AVM Input-driven components of spike-frequency adaptation can be unmasked in vivo J Neurosci 2004 24 7435 7444 15329390
Gold T Pumphrey RJ Hearing. I. The cochlea as a frequency analyzer Proc R Soc Lond B Biol Sci 1948 135 462 491
Korenberg MJ Hunter IW The identification of nonlinear biological systems: LNL cascade models Biol Cybern 1986 55 125 134 3801533
van Dijk P Wit HP Segenhout JM Dissecting the frog inner ear with Gaussian noise. I. Application of high-order Wiener-kernel analysis Hear Res 1997 114 229 242 9447936
Korenberg MJ French AS Voo SK White-noise analysis of nonlinear behavior in an insect sensory neuron: Kernel and cascade approaches Biol Cybern 1988 58 313 320 3382702
Spekreijse H Rectification in the goldfish retina: Analysis by sinusoidal and auxiliary stimulation Vision Res 1969 9 1461 1472 5367436
Marmarelis PZ Naka K White-noise analysis of a neuron chain: An application of the Wiener theory Science 1972 175 1276 1278 5061252
Benardete EA Kaplan E The receptive field of the primate P retinal ganglion cell, II: Nonlinear dynamics Vis Neurosci 1997 14 187 205 9057279
Emerson RC Korenberg MJ Citron MC Identification of complex-cell intensive nonlinearities in a cascade model of cat visual cortex Biol Cybern 1992 66 291 300 1550879
Touryan J Lau B Dan Y Isolation of relevant visual features from random stimuli for cortical complex cells J Neurosci 2002 15 10811 10818
Neri P Estimation of nonlinear psychophysical kernels J Vis 2004 4 82 91 15005649
Korenberg MJ Identification of biological cascades of linear and static nonlinear systems Proc Midwest Symp Circuit Theory 1973 18 2 1
French AS Korenberg MJ Järvilehto M Kouvalainen E Juusola M The dynamic nonlinear behavior of fly photoreceptors evoked by a wide range of light intensities Biophys J 1993 65 832 839 8218908
Markin VS Hudspeth AJ Gating-spring models of mechanoelectrical transduction by hair cells of the internal ear Annu Rev Biophys Struct 1995 24 59 83
| 15660161 | PMC539322 | CC BY | 2021-01-05 08:21:18 | no | PLoS Biol. 2005 Jan 4; 3(1):e8 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030008 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1566016210.1371/journal.pbio.0030009Research ArticleEcologyEvolutionZoologyBirdsAncient DNA Provides New Insights into the Evolutionary History of New Zealand's Extinct Giant Eagle New Zealand's Extinct Giant EagleBunce Michael [email protected]
1
2
Szulkin Marta
1
Lerner Heather R. L
3
Barnes Ian
4
Shapiro Beth
1
Cooper Alan
1
Holdaway Richard N
5
1Henry Wellcome Ancient Biomolecules Centre, Department of ZoologyUniversity of OxfordUnited Kingdom2Department of Anthropology, McMaster UniversityOntarioCanada3Department of Ecology and Evolutionary Biology, University of MichiganAnn Arbor, MichiganUnited States of America4Department of BiologyUniversity College LondonUnited Kingdom5Palaecol ResearchChristchurchNew ZealandPenny David Academic EditorMassey UniversityNew Zealand1 2005 4 1 2005 4 1 2005 3 1 e99 4 2004 1 11 2004 Copyright: © 2005 Bunce 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.
Ancient DNA Tells Story of Giant Eagle Evolution
Prior to human settlement 700 years ago New Zealand had no terrestrial mammals—apart from three species of bats—instead, approximately 250 avian species dominated the ecosystem. At the top of the food chain was the extinct Haast's eagle, Harpagornis moorei. H. moorei (10–15 kg; 2–3 m wingspan) was 30%–40% heavier than the largest extant eagle (the harpy eagle, Harpia harpyja), and hunted moa up to 15 times its weight. In a dramatic example of morphological plasticity and rapid size increase, we show that the H. moorei was very closely related to one of the world's smallest extant eagles, which is one-tenth its mass. This spectacular evolutionary change illustrates the potential speed of size alteration within lineages of vertebrates, especially in island ecosystems.
A phylogenetic analysis reveals the rapid increase in size of a New Zealand eagle, demonstrating the speed at which evolution can act on islands
==== Body
Introduction
Since the discovery of the first fossil in 1872 the sheer size of Harpagornis moorei (Figure 1A) has fuelled speculation about its evolutionary history, ecology, and extinction, which like many other New Zealand bird species, is linked ultimately to human arrival in the 13th century. Even though morphology indicates that H. moorei was approaching the upper limit of body mass for powered flight [1], it was still an efficient predator. Consistent with other members of the family Accipitridae, it killed by piercing and crushing its prey with its large talons (Figure 1B). Rock art, Maori oral history, and bone artefacts prove early Polynesians co-existed with the eagle; however, there is no evidence that humans were targets for this aerial predator. An exploratory skeletal analysis, using representative genera within the Accipitridae (but lacking Australasian representatives of the genera Hieraaetus), placed H. moorei as a sister species to Aquila audax, the Australian wedge-tailed eagle (circa 4.5 kg; 2 m wingspan) [2]. However, shifts in body size are common in island ecosystems [3] and may distort skeletal characters used in phylogenetic reconstructions.
Figure 1 Images and Phylogenetic Analysis of New Zealand's Extinct Giant Eagle, H. moorei
(A) An artist's impression of H. moorei attacking the extinct New Zealand moa. Evidence of eagle strikes are preserved on skeletons of moa weighing up to 200 kg. These skeletons show the eagle struck and gripped the moa's pelvic area, and then killed with a single strike by the other foot to the head or neck. (Artwork: John Megahan.)
(B) Comparison of the huge claws of H. moorei with those of its close relative the Hieraaetus morphnoides, the “little” eagle. The massive claws of H. moorei could pierce and crush bone up to 6 mm thick under 50 mm of skin and flesh.
(C) Maximum-likelihood tree based on cyt b data (circa 1 kb), depicting phylogenetic relationships within the “booted eagle” group. Extraction numbers or GenBank accession numbers are shown along with taxa name. Harpagornis moorei (red) groups exclusively with the small Hieraaetus eagles, and genetic distances suggest a recent common ancestor about 0.7–1.8 million years ago (early to mid Pleistocene). The tree uses an HKY + Γ4 + I likelihood model enforcing a molecular clock; maximum-likelihood bootstrap consensus values greater than 60% are shown.
Results/Discussion
To further investigate the evolutionary history of this raptor we performed an ancient DNA study on the fossil remains of two extinct H. moorei specimens together with 16 extant eagles (Table S1). Short, overlapping segments of the mitochondrial cytochrome b (cyt b) and ND2 genes were PCR-amplified, sequenced and analysed with data available on GenBank to build a maximum-likelihood tree (Figures 1C and S1). Appropriate ancient DNA controls were undertaken [4], including multiple, overlapping amplifications and independent replication (see Materials and Methods). Surprisingly, the resulting phylogeny firmly placed the H. moorei in a “clade” with a group of small eagles of the genus Hieraaetus, H. morphnoides, the little eagle, and H. pennatus, the booted eagle (both circa 1kg; 1.2m wingspan), and not with A. audax. Moreover, the genetic distance separating H. moorei and the most recent common ancestor of the related Hieraaetus eagles is relatively small (1.25%). The lack of fossil calibration points for accipitrids precludes direct estimates of divergence times; however, when a molecular rate of 0.7%–1.7% per million years, as previously estimated for avian cyt b [5], is applied to the tree, a divergence estimate of approximately 0.7–1.8 million years ago is obtained. Although such indirect molecular dating estimates are error-prone, we believe that this range is the best available approximation of the “true” date when the lineages diverged (early to mid Pleistocene); however, additional molecular data and control region sequences may further clarify the topology and timing of the splits. The arrival of H. moorei into New Zealand's South Island appears to have been a recent event, probably involving a small bird-eating Asian/Australian Hieraaetus eagle that thereafter increased rapidly in size.
An analysis of mean body mass using independent contrasts clearly demonstrated that the size of H. moorei is an anomaly in the context of the eagle phylogeny shown in Figure 1C (see Materials and Methods). Factors that may have influenced such rapid morphological evolution include the size of potential prey, competition with smaller harriers (Circus spp.), and a complete lack of terrestrial predatory mammals in New Zealand. The speed and magnitude of the increase in body mass seems unique within vertebrate lineages, and is more significant because it occurred in a species still capable of flight. The avian faunas of Islands such as Hawaii (Hawaiian goose), the Galápagos (Galápagos finches), Mauritius (dodo), and New Zealand (moa) are often cited as examples of rapid evolution of flightlessness, shifts in body size, and other specialisations in birds [1,6,7]. Isolated island faunas derived from vagile colonists often have vacant niches for large herbivores and predators. In New Zealand the large herbivore niches were occupied for millions of years by the moa. Clearly, in the absence of mammalian predators, selection for large body size in an avian predator with a relatively generalised body form was not limited by competition. Other large predatory birds have evolved on islands in the absence of mammalian competitors, notably giant eagles and owls on Cuba [8], but the magnitude of the size increase in H. moorei over its sister taxa is unrivalled. H. moorei therefore represents an extreme example of how freedom from competition on island ecosystems can rapidly influence morphological adaptation and speciation.
The phylogeny in Figure 1C also reveals considerable problems with the current classification of the “booted eagle” group (eagles with feathered tarsi), especially the genera Aquila, Hieraaetus, and Spizaetus, which are clearly paraphyletic. Assignment of species within these genera has traditionally been problematic (see [9]), so this observation was not unexpected. However, it is apparent the name for the extinct New Zealand eagle should be amended to Hieraaetus moorei (Haast, 1872) (see Materials and Methods). The inclusion of H. moorei with the small Hieraaetus eagles implies the body size of the New Zealand species has changed by almost an order of magnitude since these lineages diverged. This spectacular evolutionary change illustrates the potential speed of size alteration within lineages of vertebrates and represents yet another example of the remarkable evolutionary processes that occur within island ecosystems.
Materials and Methods
DNA extraction and amplification
DNA was extracted and amplified from two H. moorei bones (Table S1) as previously described [10] using appropriate ancient DNA techniques. “Modern” toepad tissue (museum specimens) was extracted using Qiagen (Valencia, California, United States) DNeasy tissue extraction kits. Multiple negative extraction and amplification controls were included, to detect contamination. All PCR reactions were conducted as described in [10] using Platinum Taq HiFi (Invitrogen, Carlsbad, California, United States) together with the cyt b and ND2 primers listed in Table S2. Thermal cycling conditions were typically 40 cycles of 95 °C/55–60 °C/68 °C (30–45 s each). Sequences were determined using ABI Big Dye (v.3.1) on an ABI 3100 or 3730 (Applied Biosystems, Foster City, California, United States), according to manufacturer's instructions. Modern samples, and ancient samples subsequent to PCR amplification, were analysed in the Zoology Department, Oxford University. A single Harpagornis bone was sent to an ancient DNA facility at University College London (I. Barnes) for independent replication, where identical sequences were obtained for two cyt b amplifications. Similar cyt b and ND2 tree topologies, in addition to multiple overlapping sequences, make it unlikely that we are detecting a nuclear pseudogene.
Phylogenetic methods
Maximum-likelihood trees for cyt b and ND2 were selected using a heuristic search as implemented in PAUP*4.0b10 [11] under the HKY + Γ4 + I substitution model. The assumption of a molecular clock was tested using a likelihood ratio test in which the χ2 test statistic was two times the log likelihood difference between clock and non-clock models. For the cyt b tree the assumption of rate constancy was not rejected. Node support was evaluated for 1,000 bootstrap replicates. Bayesian Markov Chain Monte Carlo phylogenies were also generated on the cyt b dataset using BEAST [12] and MrBayes [13] under similar substitution models—the topology of these trees was consistent with Figure 1C and generated posterior support values higher than the bootstrap values.
Using the maximum-likelihood tree in Figure 1C, an independent-contrasts analysis was employed to determine whether correlations existed between phylogenetic position and body mass. Mean live weight estimates were obtained from the literature, and the average mass of H. moorei was estimated from femur length [2]. A test to measure the index of phylogenetic dependence was conducted; this measures the degree to which traits vary across taxa (in a phylogeny) in accordance with predictions of a neutral Brownian model according to [14]. The results (not shown) clearly demonstrate that the mass of H. moorei is clearly an “outlier” in the context of the phylogeny presented here.
Hieraaetus systematics
The type species for the genus Hieraaetus is H. pennatus (Gmelin, 1788); therefore, the taxa grouping strongly with H. pennatus must remain in that genus. The close genetic relationship of H. morphnoides with H. pennatus firmly embeds this species in Hieraaetus. However, the New Guinea subspecies presently recognised as H. morphnoides weiskei is genetically, geographically, and morphologically distinct and warrants species status, which necessitates the new combination Hieraaetus weiskei (Reichenow, 1900). Harpagornis moorei is included in the clade with H. pennatus and H. morphnoides, and hence its generic assignment must reflect that. The name for the extinct Harpagornis moorei of New Zealand should therefore be amended to Hieraaetus moorei (Haast, 1872).
Supporting Information
Figure S1 Maximum-likelihood tree generated using 434 bp of ND2 data from a subset of eagle taxa. The tree topology seen here is identical to that seen in Figure 1C and is an independent verification of the tree topology
(89 KB PDF).
Click here for additional data file.
Table S1 List of Eagle Taxa Used in This Study along with Museum Accession Numbers and Sample Provenance
(64 KB PDF).
Click here for additional data file.
Table S2 List of Avian cyt b and ND2 Primers Used in This Study
(54 KB PDF).
Click here for additional data file.
Accession Numbers
Sequences have been deposited in GenBank (http://www.ncbi.nlm.nih.gov/Genbank/index.html) under accession numbers AY754044 to AY754056.
We would like to acknowledge the following institutions for samples: Museum of New Zealand Te Papa Tongarewa (P. Millener, A. J. D Tennyson, and J. A. Bartle), Oxford Museum of Natural History (M. Nowak-Kemp), and the Natural History Museum, Tring (R. Prys-Jones, M. Adams, and S. Parry). We thank Elisabeth Haring for discussing unpublished data, Trevor Worthy, Matthew Phillips, Rob Freckleton, David Mindell, and the Ancient Biomolecules Centre for helpful discussions, Rod Morris and Becky Williams for contributions of photographs, and John Megahan for artwork. Financial support was provided by Natural Environment Research Council (MB, AC, BS), the Wellcome Trust (IB, BS, AC), Leverhulme Trust (AC), and Foundation for Research Science and Technology New Zealand (RNH).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MB, AC, and RNH conceived and designed the experiments. MB, MS, HRLL, IB, and BS performed the experiments. MB and BS analyzed the data. MB and RNH wrote the paper.
Citation: Bunce M, Szulkin M, Lerner HRL, Barnes I, Shapiro B, et al. (2005) Ancient DNA provides new insights into the evolutionary history of New Zealand's extinct giant eagle. PLoS Biol 3(1): e9.
Abbreviations
cyt bcytochrome b
==== Refs
References
Worthy TH Holdaway RN The lost world of the moa: Prehistoric life of New Zealand 2002 Bloomington (Indiana) Indiana University Press 718
Holdaway RN Meyburg BU Chancellor RD An exploratory phylogenetic analysis of the genera of the Accipitridae, with notes on the biogeograhy of the family Raptor conservation today 1994 Berlin WORLD Working Group on Birds of Prey and Owls and the Pica Press 601 649
Palkovacs EP Explaining adaptive shifts in body size on islands: A life history approach Oikos 2003 103 37 44
Cooper A Poinar HN Ancient DNA: Do it right or not at all Science 2000 289 1139 1139 10970224
Krajewski C King DG Molecular divergence and phylogeny: Rates and patterns of cytochrome b evolution in cranes Mol Biol Evol 1996 13 21 30 8583894
Paxinos EE James HF Olson SL Sorenson MD Jackson J mtDNA from fossils reveals a radiation of Hawaiian geese recently derived from the Canada goose (Branta canadensis)
Proc Natl Acad Sci U S A 2002 99 1399 1404 11818543
Shapiro B Sibthorpe D Rambaut A Austin J Wragg GM Flight of the dodo Science 2002 295 1683 1683 11872833
Arredondo O Olson SL The great predatory birds of Cuba Collected papers in avian paleontology honoring the 90th birthday of Alexander Wetmore 1976 Washington (D.C.) Smithsonian Institution Press 169 187
Brown LH Amadon D Eagles, hawks and falcons of the world, Volume 2 1968 London Country Life 945
Bunce M Worthy TH Hoppitt W Willerslev E Drummond A Extreme reversed sexual size dimorphism in the extinct New Zealand moa Dinornis
Nature 2003 425 172 175 12968178
Swofford DL PAUP*: Phylogenetic analysis using parsimony (* and other methods), version 4 [computer program] 2001 Sunderland (Massachusetts) Sinauer
Drummond AJ Rambaut A BEAST, version 1.1.2 [computer progam]. Available: http://evolve.zoo.ox.ac.uk/beast/
2003 Accessed 12 November 2004
Huelsenbeck JP Ronquist F MrBayes: Bayesian inference of phylogeny Bioinformatics 2001 17 754 755 11524383
Freckleton RP Harvey PH Pagel M Phylogenetic analysis and comparative data: A test and review of evidence Am Nat 2002 160 712 726 18707460
| 15660162 | PMC539324 | CC BY | 2021-01-05 08:21:18 | no | PLoS Biol. 2005 Jan 4; 3(1):e9 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030009 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1566015210.1371/journal.pbio.0030010Research ArticleBioinformatics/Computational BiologyEvolutionGenetics/Genomics/Gene TherapyNoneA Model of the Statistical Power of Comparative Genome Sequence Analysis Comparative Genome Sequence AnalysisEddy Sean R
1
1Howard Hughes Medical Institute and Department of Genetics, Washington University School of MedicineSaint Louis, MissouriUnited States of AmericaHardison Ross C. Academic EditorPennsylvania State UniversityUnited States of America1 2005 4 1 2005 4 1 2005 3 1 e109 6 2004 2 11 2004 Copyright: © 2005 Sean R. Eddy.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.
Genome Sequencing: Using Models to Predict Who's Next
Comparative genome sequence analysis is powerful, but sequencing genomes is expensive. It is desirable to be able to predict how many genomes are needed for comparative genomics, and at what evolutionary distances. Here I describe a simple mathematical model for the common problem of identifying conserved sequences. The model leads to some useful rules of thumb. For a given evolutionary distance, the number of comparative genomes needed for a constant level of statistical stringency in identifying conserved regions scales inversely with the size of the conserved feature to be detected. At short evolutionary distances, the number of comparative genomes required also scales inversely with distance. These scaling behaviors provide some intuition for future comparative genome sequencing needs, such as the proposed use of “phylogenetic shadowing” methods using closely related comparative genomes, and the feasibility of high-resolution detection of small conserved features.
The mathematical model presented in this work will help to inform comparative genomics strategies for identifying conserved DNA sequences
==== Body
Introduction
Comparative genome sequence analysis is a powerful means of identifying functional DNA sequences by their evolutionary conservation [1,2,3]. It will be instrumental for achieving the goal of the Human Genome Project to comprehensively identify functional elements in the human genome [4]. How many comparative genome sequences do we need? Where is the point of diminishing returns, after which sequencing another koala or bat does not contribute significant information to human genome analysis? Since sequencing is expensive and capacity remains limited, one would like to address this issue as rigorously as possible.
Empirical evaluations of candidate comparative genomes have become important in allocating sequencing resources. Pilot sequencing and analysis in Saccharomyces and Drosophila species were done to choose appropriate species for comparative genome sequencing [5,6]. A pilot sequencing effort is underway for a number of mammalian genomes to evaluate their utility for human genome analysis [4]. Given the complexity of genomes, empirical studies are necessary. However, one would also like to complement this with higher-level, general insights that are independent of the details of particular analysis programs, organisms, and genomic features.
Cooper et al. proposed a mathematical model of one important type of comparative genome analysis [7]. They framed a question amenable to quantitative modeling: how many comparative genomes, and at what distances, are required to detect that an individual base in a target genome is “neutral” (inferred to be evolving at the neutral rate) as opposed to “conserved” (inferred to be under purifying selection)? Their model infers a nucleotide site to be conserved if it is 100% identical to homologous sites in N comparative genomes. The key parameters are the independent branch lengths (di) contributed to a phylogeny by each new comparative genome (i), measured in neutral substitutions per site. More neutral evolutionary distance makes it more likely that neutral sites will have one or more substitutions in the alignment. Analytical strength increases as a function of the total neutral branch length in the phylogeny (Σidi
), because the probability that a neutral site has no changes in any branch of the phylogeny (and thus would be misclassified as conserved) is taken to be approximately e
−Σidi
. Based on the model, they concluded that 5.0 neutral substitutions/site of total branch length (about 10–20 well-chosen mammalian genomes) would approach “single nucleotide resolution” for human genome analysis, with a false positive probability (FP) of less than e
−5.0 per invariant site.
This model has some limitations that seem serious enough to question the proposed target of 10–20 mammalian genomes. Most importantly, it assumes that conserved sites are invariant. Few conserved features are absolutely invariant. If invariance is required to infer conservation, the fraction of truly conserved sites that are wrongly inferred to be neutral (because a substitution is seen in one of the comparative genomes) asymptotically approaches one as the number of comparative genomes or their evolutionary distance increases. We want to consider not just our FP, but our statistical power—our ability to successfully detect features that are conserved.
Additionally, single nucleotide resolution may not be the most relevant goal. It is useful to consider single nucleotide resolution as an ultimate limit on comparative analyses—one can imagine plausible analyses of single bases, and certainly individual codons—but we are mostly concerned with identifying conserved features of greater length, such as exons or transcription factor binding sites.
Nonetheless, the level of abstraction introduced by Cooper et al. is attractive. There is a need for better intuitions for planning comparative genome sequencing. How many more comparative genomes are needed as one looks for smaller and smaller conserved features—from exons to regulatory sites to single codons or even single nucleotides? How many more genomes are needed as one uses more and more closely related comparative genomes, in order to improve the chances that homologous lineage-specific features are found and correctly aligned [8,9]? Precise answers will be elusive, because genome biology is complex, but perhaps there are rough, useful scaling relationships amongst comparative genome number, evolutionary distance, and feature size. To explore this, I have extended the ideas introduced by Cooper et al. and developed an abstract model that seems to capture the essential flavor of comparative genome analysis.
Results/Discussion
Description of the Model
A “feature” is a sequence of L nucleotide sites in the target genome. We assume we have a correct, ungapped multiple sequence alignment of this sequence to N homologous features from N additional comparative genomes, and that the L sites are independent.
In the NL nucleotides in the aligned comparative sequences, we count how many changes are observed relative to the target feature sequence; call this c. If c is greater than some threshold C, we infer the feature is evolving at the neutral rate. If, on the other hand, c is less than or equal to C, we infer the feature is conserved.
We assume that each comparative genome is independently related to the target genome by a branch length of D neutral substitutions per site, that is, a uniform star topology, with the target at the root, and equal length branches to the comparative genomes at the leaves. A uniform star topology allows us to model how evolutionary distance affects comparative analysis at an abstract level, as a single variable D, independent of the details of real phylogenies. The biologically unrealistic placement of the known target at the root simplifies the mathematics, and does not significantly affect the results compared to making the more realistic assumption of an unknown ancestor at the root of a tree with N + 1 leaves, including the target.
We assume that the only difference between conserved features and neutral features is that conserved features evolve more slowly, by a relative rate coefficient ω. A conserved site accumulates an average of ωD substitutions, whereas a neutral site accumulates an average of D substitutions. ω = 0 for an absolutely conserved feature; ω = 1 for a neutrally evolving feature. At short evolutionary distances, we expect about c = DNL changes in neutral features, and c = ωDNL changes in conserved features, with binomial densities for P(c) around those values.
To model the probability that two nucleotides diverged by D or ωD substitutions will be observed to be identical (to deal with multiple substitutions at one site), we assume a Jukes-Cantor process in which all types of base substitution occur at the same rate [10]. Under a Jukes-Cantor model, the probability that two sites that have diverged by D substitution events are identical is , which approaches 25% at infinite divergence.
Given these assumptions, the FP in a comparative analysis (the probability that we erroneously infer that a neutral feature is conserved) is the probability that a neutral feature happens to have C or fewer observed changes (a cumulative binomial distribution):
and the false negative probability (FN; the probability that we erroneously infer that a conserved feature is neutral) is the probability that a conserved feature happens to have more than C observed changes:
The model therefore depends on four parameters: the size of the conserved feature, L, the relative rate of evolution of the conserved feature, ω, the number of comparative genomes, N, and the neutral distance of the comparative genomes from the target genome, D. The threshold C is usually not an input parameter (except in the special case of invariance; C = 0). Rather, we find the minimum genome number N (or feature size L) at which there exists any cutoff C that can satisfy specified FN and FP thresholds.
The Cooper et al. model is essentially a special case where L = 1 (single nucleotide resolution), ω = 0 (conserved sites are always invariant), C = 0 (only invariant sites are inferred to be conserved), and FN = 0 by definition (if all conserved sites are invariant, and all invariant sites are inferred to be conserved, then all conserved sites are detected). Also, instead of using an evolutionary model to account for multiple substitutions at one site (saturation), Cooper et al. make a Poisson assumption that the probability of observing no change at a comparative site is e−D, which is only valid for small D.
The model discriminates features based on their relative rate of evolution. The same equations could be used to detect features evolving faster than the neutral rate (positively selected features), or to detect highly conserved features on a background of less strongly conserved sequence, as, for instance, transcription factor binding sites in an upstream region often appear [11,12]. For simplicity, I will only talk about discriminating “conserved” from “neutral” features here.
Reasonable Parameter Values
The feature length L and conservation coefficient ω abstractly model the type of feature one is looking for. I use L = 50, L = 8, and L =1 as examples of detecting small coding exons, transcription factor binding sites, and single nucleotides, respectively, solely by sequence conservation. On average, conserved exons and regulatory sites appear to evolve about 2- to 7-fold slower than neutral sequences (ω = 0.5–0.15) [7,8,13,14,15]. I use 5-fold slower (ω = 0.2) in most cases discussed below. Typically, one doesn't know L or ω when looking for novel features. These two parameters behave as bounds: if one can detect a specified feature, larger and/or more conserved features are also detected.
The model's single distance parameter, D, abstractly represents the independent neutral branch length contributed by each comparative genome [7]. In a phylogenetic tree of the target with N > 1 comparative genomes that are as independent from each other as possible, we can roughly consider the independent branch length contributed by each comparative genome to be one-half its pairwise distance to the target genome, because in a real tree (with unknown common ancestors, as opposed to placing the target at the root of a uniform star topology) all comparative genomes share at least one branch leading to the target. Thus the figures highlight D = 0.03, 0.19, and 0.31 as “baboon-like,” “dog-like,” and “mouse-like” distances from human, 50% of one set of pairwise neutral distance estimates of 0.06, 0.38, and 0.62, respectively, arbitrarily chosen from the literature [7]. These labels are solely to give some intuition for what the model's D parameter means. The correspondence between D and real branch lengths is crude. Real neutral distance estimates are a subject of substantial (up to about 2-fold) uncertainty in the literature, and there are regional variations and strong context effects on neutral substitution rates in mammalian genomes [16,17]. More importantly, the model's uniform star topology, though it allows a high-level analysis in terms of just two parameters, D and N, makes direct comparison to real phylogenies difficult. Large numbers of equidistant, independently evolved mammalian genomes do not occur in reality. Real genomes are not independent, and will generally contribute an independent neutral branch length of less than one-half of their pairwise distance to the target genome.
Critically, the model assumes that homologous features are present, correctly detected, and correctly aligned. In reality, with increasing evolutionary distance, features can be gained, lost, or transposed [14,18,19,20,21], the ability to detect homology by significant sequence similarity decreases, and alignments become less reliable [22]. The frequency of effects like loss, gain, and transposition depend on the biology of particular types of features, so departures from the model's “alignment assumptions” are difficult to model abstractly. However, minimally, we can posit a maximum neutral distance, D
max, beyond which the alignment assumptions will not hold, based just on the ability of alignment programs to recognize and align homologous DNA sequences. Roughly speaking, reliability of DNA sequence alignments begins to break down at about 70% pairwise identity. For alignments of conserved features evolving 5-fold slower than neutral, this suggests D
max ∼ 0.15/0.2 = 0.75; Figures 1 and 2 show results out a little further, to D
max = 1.0.
Figure 1 Number of Genomes Required for Single Nucleotide Resolution
The red line plots genome number required for identifying invariant sites (ω = 0) with a FP of 0.006, essentially corresponding to the Cooper model [7]. Black lines show three more parameter sets: identifying 50% (FN < 0.5) of conserved sites evolving 5-fold slower than neutral (ω = 0.2) with FP < 0.006, doing likewise but with a more-stringent FP of 0.0001, and identifying 99% of conserved sites instead of just half of them. Values of N at baboon-like, dog-like, and mouse-like neutral distances are indicated with diamonds, squares, and circles, respectively. Jaggedness of the lines here and in subsequent figures is an artifact of using discrete N, L, and cutoff threshold C to satisfy continuous FP and FN thresholds.
Figure 2 Number of Genomes Required for 8-nt or 50-nt Resolution
Top: identifying 8-nt conserved features (“transcription factor binding sites”; L = 8); bottom: identifying 50-nt conserved features (“exons”; L = 50). Parameter settings are indicated at top right, in same order as the plotted lines. The parameters are the same as those used in Figure 1.
Two different FP settings are used as illustrative examples: 0.006 (the e
−5 threshold used by Cooper et al. [7]) and the more stringent 10−4. For consistency, the same two FP thresholds are used to illustrate scaling behaviors for all three feature sizes (L = 1, L = 8, and L = 50). However, for a real analysis, one wants to consider the appropriate choice of FP carefully. In a genome sequence of length M, the total number of false positive feature predictions in all overlapping possible windows of length L is M − L + 1, multiplied by FP per feature. In most analyses, we would probably merge overlapping predicted features into a single predicted conserved region, resulting in a lower number of false positive regions in a genome. This overlap correction (from the number of false features to the number of false regions) depends on the parameters, but for the parameters in Figures 1 and 2 it varies from 1.5- to 2-fold less for L = 8 sites and 4- to 8-fold less for L = 50 sites, based on simulations. Thus, for example, FP = 10−4 corresponds to one false positive feature per 10 kb, and (for the parameters here) somewhere between one false positive conserved region per 20–100 kb, depending on the feature. For “small exon” detection, this means 40,000–300,000 false region/feature predictions in the 3-Gb human genome; for “transcription factor binding sites,” this means one false positive feature or region per 10–20 kb. FP = 10−4 therefore seems a reasonable stringency for L = 8 or L = 50 feature analyses. If one carried out a single nucleotide resolution analysis on a genome-wide scale, FP = 10−4 would mean that 99.8% of the predictions for conserved bases in the 3-Gb human genome would be correct, assuming about 5% of the bases are truly conserved and detected with high sensitivity. However, it is likely that one would actually carry out single nucleotide resolution analyses on a subset of conserved features that had already been identified (exons, for example), so a less stringent FP might be required. The setting of FP = 0.006 might therefore be more appropriate for evaluating single nucleotide resolution, where FP is closer to the traditional statistical choices of a 0.01 or 0.05 significance level.
Single Nucleotide Resolution Requires Many Genomes
The Cooper model concluded that for invariant conserved sites, sequencing comparative genomes to achieve a total branch length of five neutral substitutions per site would give single nucleotide resolution, with a FP of e
−5 (0.006) [7]. Under my model, detection of invariant nucleotides takes about 17 genomes at mouse-like distances, essentially as predicted by Cooper et al. (Figure 1).
However, the picture changes when one considers comprehensive detection of features that are conserved but not invariant (Figure 1). To detect 50% of sites evolving 5-fold slower than neutral, we need 25 comparative genomes at mouse-like distances at the same (arbitrary) false positive threshold of less than 0.006. For a comprehensive screen that would detect 99% of conserved single nucleotides with a FP of less than one per 10 kb, the model predicts about 120 comparative genomes at mouse-like distances are needed.
Detectable Feature Size Scales Inversely with Genome Number
The large genome numbers in Figure 1 might appear to conflict with the known power of comparing just two genomes, such as human and mouse. This is because recognizing conserved sequences is easier than recognizing conserved single nucleotides; the size of the conserved feature matters.
Figure 2 shows how many genomes are required to detect small features like transcription factor binding sites (L ∼ 8) or larger features like short coding exons (L ∼ 50). One genome at about human/mouse distance is sufficient for reasonable strength in coding exon detection. For a range of reasonable sensitivity and specificity stringencies, three to 15 genomes at human/mouse distance are sufficient for detecting transcription factor binding sites.
There is a general, intuitive explanation for this. The strength of an analysis will depend on the difference in the expected number of substitutions in neutral features versus conserved features. This difference will be proportional to NL, the total number of aligned sites. Thus, for a constant stringency, the required number of comparative genomes is expected to scale inversely with the size of the feature to be detected (N ∝ 1/L): to detect conserved features ten times smaller, it takes ten times as many comparative genomes. (This scaling behavior is seen directly later.)
No Clear Optimum for Evolutionary Distance, but Close Distances Disfavored
Figures 1 and 2 show two other notable behaviors. First, there is no sharp optimum for the neutral distance D. The number of genomes required is relatively flat for a wide range, from about 0.4 to well beyond 1.0. Within a broad range, the exact choice of one comparative genome versus another has little impact.
This is shown more directly in Figure 3, in which a measure of overall statistical strength is plotted against neutral distance over an unrealistically long range of D, out to 4.0 substitutions/site. For conserved features evolving 5-fold slower than neutrality, assuming that alignment assumptions hold, the optimum distance according to the model is about 1.4 neutral substitutions/site, four to five times the mouse-like distance. However, for many kinds of features, at such long evolutionary distances the alignment assumptions are likely to break down. Because the mathematically optimal distance for discriminating idealized conserved and neutral features lies outside the range where the alignment assumptions are likely to hold, it may not be particularly meaningful to imagine a uniquely optimal choice of evolutionary distance for comparative genome analysis; optimal choices will be problem-dependent. (This is not surprising, of course, but perhaps useful to see clearly in a simple model.)
Figure 3 A Measure of Statistical Strength As a Function of Neutral Evolutionary Distance
One convenient threshold-independent measure of the strength of a comparative analysis is an expected Z score, the expected difference Δc in the number of substitutions in a neutral feature alignment versus a conserved feature alignment, normalized to units of standard deviations. E(Z) is readily calculated for the binomial distribution:
where pn and pc are the probabilities of observing a change at one aligned comparative nucleotide according to the Jukes-Cantor equation.
The plots here are for N = 5 and L = 8. The shape of the curve is independent of N and L, while the absolute magnitude of Z scales as √NL
. The x-axis is shown from D = 0 to D = 4, beyond the more realistic range of Figures 1 and 2, to show the mathematically optimum D if homologous conserved features were present, recognized, and accurately aligned at any D.
The second behavior worth noting in Figures 1 and 2 is that at close evolutionary distances, the necessary number of comparative genomes needed ramps up steeply. For instance, at human/baboon distances of 0.03, achieving equivalent statistical strength requires about seven times as many comparative genomes as when using human/mouse distances (see Figure 2).
There is another general intuition behind these results. For D ≪ 1, the expected number of substitutions is DNL in a neutral feature and ωDNL in a conserved feature. So, for a constant statistical stringency, the number of genomes required will scale inversely with evolutionary distance, when the distance is small. At larger distances, this scaling ceases as the number of observed changes saturates.
The strong scaling of N at small distances D has implications for the use of “phylogenetic shadowing” using closely related genomes [8,9]. It is clear that the use of closely related genomes is advantageous in several ways: alignments are more accurate, one can accurately align a surrounding neutral region to detect small embedded conserved regions, and homologous features are more likely to be present (for instance, primate-specific features in human analyses). However, the model illustrates how these advantages are accompanied by a significant cost in statistical strength (see Figure 3). When using comparative species at short evolutionary distances, species choice matters a lot. Within primates, for example, divergence times from human vary about 10-fold (∼6 to ∼65 million years); if one aims to use “primate sequences” for human genome analysis, there is a large difference between using distant primates (lemurs or New World monkeys) versus close primates (great apes).
Resolution and Stringency as a Function of Genome Number
How much additional information does each new comparative genome sequence give us? The top panel in Figure 4 plots sensitivity and specificity as the number of comparative genomes increases, for an analysis of transcription factor binding site–like features. The scaling behavior is expected to be (roughly speaking) log(FP or FN) ∝ − N, based on the cumulative binomial expressions for FP and FN. That is, each additional genome reduces FP or FN by a roughly constant multiplier; for the parameters used here, every three or four more comparative genomes reduces FP by 10-fold. The bottom panel in Figure 4 plots resolution L as a function of N, showing the expected L ∝ 1/N scaling. Each doubling of the number of comparative genomes increases resolution about 2-fold.
Figure 4 Increase in Stringency and Resolution with Increasing Genome Number
Top: black line shows improvement in specificity (FP) for transcription factor (TF) binding site–like features (L = 8, ω = 0.2) as comparative genome number increases, for FN = 0.01 (99% of sites detected), and genomes of D = 0.31 (mouse/human-like distance). Red line shows improvement in sensitivity (FN) for the same parameters and a FP threshold of 0.0001. Shown as a log-linear plot to show the expected rough log(FP or FN) proportional to −N scaling.
Bottom: resolution (size of detectable feature, L) as a function of comparative genome number, plotted on log-log axes to show the fit to the expected L ∝ 1/N scaling. All four lines assume goals of FN < 0.01 and FP < 0.0001. Black lines are for identifying conserved features evolving 5-fold slower than neutral (ω = 0.2), using baboon-like (D = 0.03), dog-like (D = 0.19), or mouse-like (D = 0.31) genomes. Red line is for identifying invariant features with mouse-like genomes.
Good Agreement with More Realistic Simulations
The model's simplicity is useful. By just counting the number of substitutions in conserved versus neutral features, the reasons for the scaling behaviors are more intuitively obvious. However, the assumptions required for this level of simplicity are questionable. In real DNA sequences, the Jukes-Cantor model's simple assumptions are violated in many ways; transitions are more frequent than transversions, base composition is not uniform, and mutation rates show strong context dependence [17]. In a real analysis, we would use probabilistic methods to compare the log likelihood ratio (LLR) of a phylogenetic tree under competing hypotheses of two different rates [8,23,24], so we can deal with real phylogenies and different expected rates of substitutions at different bases.
The relative predicted scaling behaviors are unlikely to change under more realistic simulations. However, for the model to be useful as a rough guide for required genome number under different comparative analysis scenarios, at a minimum we want to know whether the absolute predicted numbers would be substantially different for features evolving under a more realistic evolutionary model, such as the Hasegawa-Kishino-Yano (HKY) model [25], which models nonuniform base composition and transition/tranversion rate bias, and if we analyzed those data with LLR statistics instead of simply counting substitutions.
Therefore, I performed the following computational simulation study. Synthetic “neutral” and “conserved” feature alignments were generated using two HKY models that differed in evolutionary rate by a factor of ω. The rates in the HKY models were parameterized with an AT-biased base composition of 33% A, 17% C, 17% G, and 33% T, and a biased transition/transversion rate ratio of 4.0. A feature alignment was simulated by choosing a random L-mer (using the specified base composition) as the target feature, then generating N homologous features from it with substitutions according to an HKY conditional substitution matrix at distance D. For each dataset, 103 conserved feature alignments and 106 neutral feature alignments were generated. These alignments were then scored under the two HKY models and ranked by LLR score. This was repeated for increasing N until an LLR score threshold existed that could satisfy the chosen FP and FN thresholds. I then reproduced the analyses in Figures 1 and 2 using the HKY/LLR simulation for the 27 highlighted points with ω = 0.2. That is, for the 27 combinations of D = 0.03, 0.19, or 0.31; L = 1, 8, or 50; and (FP, FN) = (0.0001, 0.01), (0.0001, 0.5), or (0.006, 0.5), I determined the minimum number of genomes required to achieve the chosen thresholds.
This analysis showed that the predictions of the simple model's equations and the results of the HKY/LLR simulations are in close agreement. The maximum deviation was 15%. For example, for the [D = 0.19, L = 1] points where the model predicts needing N = 183, 89, and 40 for the different values of FP and FN, the HKY/LLR simulation predicts needing N = 210, 80, and 35; for the [D = 0.19, L = 8] points, the model predicts N = 23, 12, and 5, and the simulation predicts N = 24, 11, and 5; and for the [D = 0.19, L = 50] points the model and simulation both predict N = 4, 2, and 1.
More significant discrepancies appear at larger distances. A simple Jukes-Cantor model has only one substitution rate, so all types of substitutions saturate equally fast. In an HKY model, some substitution rates are faster than others. Intuitively, one expects an HKY model to be able to extract information from slower, less quickly saturated substitutions at longer distances, resulting in more discrimination at large D than the simple model predicts. This effect appears at distances of D > 2.0–3.0 or so: for instance, for [ω = 0.2, L = 8] features, to achieve FP < 0.0001 and FN < 0.01 for distances of D = 1, 2, 3, 4, 5, and 10, the simple model predicts needing N = 8, 8, 11, 17, 26, and 335 genomes, respectively, whereas HKY/LLR simulations predict needing N = 8, 9, 9, 13, 17, and 82 genomes. Thus, the simple model's approximation breaks down somewhat at larger distances, beyond the D < 1 range that is considered here to be reasonable for comparative genomics.
Additionally, nonuniform base composition causes some composition-dependent spreading around the mean N that is not predicted by the simple model. For instance, GC-rich features are more easily detected than AT-rich features when substitution rates are biased towards high AT composition. Additional HKY/LLR simulations, using the same HKY matrices as above but specifically looking at poly-A features versus poly-C target features, show this effect; for instance, for [ω = 0.2, L = 8] features at D = 0.19, to achieve FP < 0.0001 and FN < 0.01, we need at least N = 24 genomes to detect features on average, but specifically we need N = 19 for poly-C/G features and N = 29 for poly-A/T features.
Reasonable Agreement with Available Data
One also wants to see that the model's predictions do not disagree with published results, at least to the extent that it is possible to crudely compare real phylogenies to the abstracted uniform star topology of the model. Three examples follow.
Cooper et al. estimated that the mouse and rat genomes suffice for about 50-nt resolution of human conserved features [26]. The independent branch lengths to human, mouse, and rat are roughly 0.3, 0.3, and 0.1 neutral substitutions/site; the rat is close to the mouse, so this situation is difficult to fit with a single D. However, using either N = 1 and D = 0.6 (pairwise comparison to one rodent using one full pairwise distance for D), or N =2 and D = 0.23 (approximating D as an average of three independent branch lengths), or N = 2 and D = 0.35 (approximating D as one-half the average pairwise distances from human to mouse and rat), the model predicts that 90% of 50-nt features with ω = 0.2 can be detected with a reasonable FP of between 0.0003 and 10−5; but for features just half that size (L = 25), FP collapses to between 0.02 and 0.006 (one false prediction every 50–150 nt).
Boffelli et al., in introducing “phylogenetic shadowing,” used 13–17 primate sequences with a total independent branch length of about 0.6 neutral substitutions/site to analyze conserved sequences smoothed in 50-nt windows, and conserved regions down to 40–70 nt were detected effectively [8]. The model predicts that for N = 15 and D = 0.04 (average independent branch length of 0.6/15), one can detect 90% of 50-nt, ω = 0.2 features with a FP of 10−5; but for 25-nt features, FP collapses to 0.003 (one false prediction per 300 bp).
Kellis et al. and Cliften et al. reported comparative analyses to identify transcription factor binding sites in Saccharomyces cerevisiae using alignments of intergenic regions to three comparative Saccharomyces genomes, with a total independent branch length of about 0.8–0.9 [11,12]. For N = 3 and D = 0.27–0.30 (average independent branch length of 0.8/3 to 0.9/3), the model predicts that binding site–like features (L = 8, ω = 0.2) would be detected with a FP of 0.001–0.002 (about one false prediction per upstream region) and a sensitivity of about 25%, suggesting that these data are barely sufficient to identify individual short conserved features. Indeed, though both research groups showed examples of highly conserved individual sites, both groups analyzed their data primarily at the level of detecting motif consenses, rather than attempting to detect individual features genome-wide. That is, they required that the same motif be found conserved in multiple places upstream of multiple genes. This is a data aggregation strategy, multiplying the effective L by the number of copies of the feature. In this way, even when only a fraction of individual features are identified, the existence of a conserved consensus motif may be inferred from the average conservation of the aggregated data.
Limitations on the Generality of These Conclusions
The model assumes a pure, brute force detection of individual conserved features by comparative analysis. For many particular problems, one can leverage additional information and reduce the number of comparative genomes needed. Data aggregation strategies are one example (for instance, detecting that a particular consensus motif is conserved more often than expected, averaged across all individual occurrences [27]). Another strategy is to combine sequence conservation data with other experimental data (for instance, using microarray data to detect that a marginally conserved motif is also statistically associated with a coordinately regulated set of genes [28]).
Some features are not just conserved, but also show informative patterns of substitution, insertion, and deletion, so we can gain power by using feature-specific evolutionary models instead of a general conservation screen. For instance, coding regions predominately show substitutions in wobble positions and strong selection against insertions/deletions, and those insertions/deletions that remain will generally preserve frame [11]. Conserved structural RNAs reveal their basepaired secondary structure interactions by compensatory basepair mutations [29]. In such analyses it becomes important to see enough evolutionary events to distinguish one kind of conserved feature from other kinds of conserved features, not just to discriminate conserved from neutral. Because different conserved features evolve at different rates, one would generally want to have a range of comparative genomes at different distances, so that for any given conserved feature with its particular relative rate of evolution, one can find alignments in a “sweet spot” with the right amount of divergence.
Finally, there are other important uses of comparative genomics in addition to DNA sequence analysis of conserved elements. For example, evolutionary/developmental studies choose species based on phylogenetic position, and population genetics studies choose multiple individuals within the same species.
Concluding Remarks
The principal results here are two inverse scaling behaviors that provide useful intuitions for planning comparative genome sequencing. All other things being constant, the required number of comparative genomes is inversely proportional to detectable feature size, and at small evolutionary distances, required genome number becomes inversely proportional to the neutral distance to the comparative genomes.
Neither behavior is entirely surprising; the contribution of an abstract model is to see them more clearly. Obviously, it takes more comparative genomes to recognize smaller features, though one may not have predicted a simple inverse relationship between L and N. And it is already common to use total independent neutral branch length as a measure of the strength of a comparative dataset [7,8,9,30], which implies an inverse relationship between genome number and evolutionary distance, a relationship made explicit in a simplified model where total independent branch length is ND.
The model also shows clearly that for two analysis scenarios—identification of small conserved features and the identification of lineage-specific conserved features in closely related genomes—it will be useful to obtain large numbers of comparative genome sequences. Since a small number of comparative genome sequences are already enabling powerful analyses, this may be surprising. Even for simple conservation analyses, we have not begun to exhaust the power of comparative genome analysis.
Materials and Methods
The model was implemented in several ANSI C programs, which can be downloaded at http://www.genetics.wustl.edu/eddy/publications/Eddy05.
This work was motivated by my past service on the National Institutes of Health (NIH) National Human Genome Research Institute (NHGRI) Genome Resources and Sequencing Priorities Panel, which assisted NHGRI in prioritizing genome sequencing decisions. I am grateful for funding support from the NIH NHGRI, the Howard Hughes Medical Institute, and Alvin Goldfarb.
Competing interests. I am affiliated with the Washington University Genome Sequencing Center (St. Louis, Missouri, United States). Because of past service on scientific advisory boards, I hold stock options in two genomics companies, Celera Genomics (Rockville, Maryland, United States) and Orion Genomics (St. Louis, Missouri, United States). Genome centers and genomics companies may be affected financially by funding decisions and priorities in comparative genome sequencing.
Author contributions. SRE conceived and designed the models, analyzed the data, and wrote the paper.
Citation: Eddy SR (2005) A model of the statistical power of comparative genome sequence analysis. PLoS Biol 3(1): e10.
Abbreviations
FPfalse positive probability
FNfalse negative probability
HKYHasegawa-Kishino-Yano
LLRlog likelihood ratio
==== Refs
References
Hardison RC Conserved noncoding sequences are reliable guides to regulatory elements Trends Genet 2000 16 369 372 10973062
Sidow A Sequence first. Ask questions later Cell 2002 111 13 16 12372296
Hardison RC Comparative genomics PLoS Biol 2003 1 e58 14624258
Thomas JW Touchman JW Blakesley RW Bouffard GG Beckstrom-Sternberg SM Comparative analyses of multi-species sequences from targeted genomic regions Nature 2003 424 788 793 12917688
Cliften PF Hillier LW Fulton L Graves T Miner T Surveying Saccharomyces genomes to identify functional elements by comparative DNA sequence analysis Genome Res 2001 11 1175 1186 11435399
Bergman CM Pfeiffer BD Rincon-Limas DE Hoskins RA Gnirke A Assessing the impact of comparative genomic sequence data on the functional annotation of the Drosophila genome Genome Biol 2002 3 RESEARCH0086 12537575
Cooper GM Brudno M Green ED Batzoglou S Sidow A Quantitative estimates of sequence divergence for comparative analyses of mammalian genomes Genome Res 2003 13 813 820 12727901
Boffelli D McAuliffe J Ovcharenko D Lewis KD Ovcharenko I Phylogenetic shadowing of primate sequences to find functional regions of the human genome Science 2003 299 1391 1394 12610304
Boffelli D Nobrega MA Rubin EM Comparative genomics at the vertebrate extremes Nat Rev Genet 2004 5 456 465 15153998
Jukes TH Cantor CR Munro HN Evolution of protein molecules Mammalian protein metabolism 1969 New York Academic Press 21 132
Kellis M Patterson N Endrizzi M Birren B Lander ES Sequencing and comparison of yeast species to identify genes and regulatory elements Nature 2003 423 241 254 12748633
Cliften P Sudarsanam P Desikan A Fulton L Fulton B Finding functional features in Saccharomyces genomes by phylogenetic footprinting Science 2003 301 71 76 12775844
Makalowski W Boguski MS Evolutionary parameters of the transcribed mammalian genome: An analysis of 2,820 orthologous rodent and human sequences Proc Natl Acad Sci U S A 1998 95 9407 9412 9689093
Dermitzakis ET Clark AG Evolution of transcription factor binding sites in mammalian gene regulatory regions: Conservation and turnover Mol Biol Evol 2002 19 1114 1121 12082130
Moses AM Chiang DY Kellis M Lander ES Eisen MB Position specific variation in the rate of evolution in transcription factor binding sites BMC Evol Biol 2003 3 19 12946282
Mouse Genome Sequencing Consortium Initial sequencing and comparative analysis of the mouse genome Nature 2002 420 520 562 12466850
Hwang DG Green P Bayesian Markov chain Monte Carlo sequence analysis reveals varying neutral substitution patterns in mammalian evolution Proc Natl Acad Sci U S A 2004 101 13994 14001 15292512
Ludwig MZ Patel NH Kreitman M Functional analysis of eve stripe 2 enhancer evolution in Drosophila Rules governing conservation and change Development 1998 125 949 958 9449677
Ludwig MZ Bergman C Patel NH Kreitman M Evidence for stabilizing selection in a eukaryotic enhancer element Nature 2000 403 564 567 10676967
Costas J Casares F Vieira J Turnover of binding sites for transcription factors involved in early Drosophila development Gene 2003 31 215 220
Ludwig MZ Functional evolution of noncoding DNA Curr Opin Genet Dev 2002 12 634 639 12433575
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
Anisimova M Bielawski JP Yang Z Accuracy and power of Bayes prediction of amino acid sites under positive selection Mol Biol Evol 2002 19 950 958 12032251
Yang Z Inference of selection from multiple species alignments Curr Opin Genet Dev 2002 12 688 694 12433583
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
Cooper GM Brudno M Stone EA Dubchak I Batzoglou S Characterization of evolutionary rates and constraints in three mammalian genomes Genome Res 2004 14 539 548 15059994
Fairbrother WG Yeo GW Yeh R Goldstein P Mawson M RESCUE-ESE identifies candidate exonic splicing enhancers in vertebrate exons Nucleic Acids Res 2004 32 W187 W190 15215377
Wang T Stormo GD Combining phylogenetic data with co-regulated genes to identify regulatory motifs Bioinformatics 2003 19 2369 2380 14668220
Rivas E Eddy SR Noncoding RNA gene detection using comparative sequence analysis BMC Bioinformatics 2001 2 8 11801179
Cooper GM Sidow A Genomic regulatory regions: Insights from comparative sequence analysis Curr Opin Genet Dev 2003 13 604 610 14638322
| 15660152 | PMC539325 | CC BY | 2021-01-05 08:21:19 | no | PLoS Biol. 2005 Jan 4; 3(1):e10 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030010 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1566015410.1371/journal.pbio.0030013Research ArticleGenetics/Genomics/Gene TherapyPlant SciencePlantsSorghum Genome Sequencing by Methylation Filtration Sorghum Genome SequencingBedell Joseph A [email protected]
1
Budiman Muhammad A
2
Nunberg Andrew
1
Citek Robert W
1
Robbins Dan
1
Jones Joshua
2
Flick Elizabeth
2
Rohlfing Theresa
3
Fries Jason
3
Bradford Kourtney
3
McMenamy Jennifer
3
Smith Michael
4
Holeman Heather
4
Roe Bruce A
5
Wiley Graham
5
Korf Ian F
6
Rabinowicz Pablo D
7
Lakey Nathan
8
McCombie W. Richard
9
Jeddeloh Jeffrey A
4
Martienssen Robert A
9
1Bioinformatics, Orion GenomicsSaint Louis, MissouriUnited States of America2Library Construction, Orion GenomicsSaint Louis, MissouriUnited States of America3Sequencing, Orion GenomicsSaint Louis, MissouriUnited States of America4Biomarkers, Orion GenomicsSaint Louis, MissouriUnited States of America5Department of Chemistry and Biochemistry, University of OklahomaNorman, OklahomaUnited States of America6Genome Center, University of CaliforniaDavis, CaliforniaUnited States of America7The Institute for Genomic Research, RockvilleMarylandUnited States of America8Business, Orion GenomicsSaint Louis, MissouriUnited States of America9Cold Spring Harbor Laboratory, Cold Spring HarborNew YorkUnited States of AmericaDoebley John Academic EditorUniversity of WisconsinUnited States of America1 2005 4 1 2005 4 1 2005 3 1 e1316 8 2004 8 11 2004 Copyright: © 2005 Bedell 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.
Separating Wheat from Chaff in Plant Genomes
Sorghum bicolor is a close relative of maize and is a staple crop in Africa and much of the developing world because of its superior tolerance of arid growth conditions. We have generated sequence from the hypomethylated portion of the sorghum genome by applying methylation filtration (MF) technology. The evidence suggests that 96% of the genes have been sequence tagged, with an average coverage of 65% across their length. Remarkably, this level of gene discovery was accomplished after generating a raw coverage of less than 300 megabases of the 735-megabase genome. MF preferentially captures exons and introns, promoters, microRNAs, and simple sequence repeats, and minimizes interspersed repeats, thus providing a robust view of the functional parts of the genome. The sorghum MF sequence set is beneficial to research on sorghum and is also a powerful resource for comparative genomics among the grasses and across the entire plant kingdom. Thousands of hypothetical gene predictions in rice and Arabidopsis are supported by the sorghum dataset, and genomic similarities highlight evolutionarily conserved regions that will lead to a better understanding of rice and Arabidopsis.
Methylation filtration makes practical the sequencing of large genomes, such as those found in sorghum, by preferentially capturing functionally relevant sequences
==== Body
Introduction
Sorghum bicolor is a vitally important crop in Africa and much of the developing world. It has a remarkable ability to endure both drought conditions and water-logging and it grows well on marginal lands [1]. It is the dietary staple of more than 500 million people in more than 30 countries with only rice, wheat, maize, and potatoes feeding more people than sorghum [1]. Sorghum is in the panicoid grass subfamily and is closely related to maize, millet, and especially sugarcane, and is more distantly related to wheat and rice. Its value as a dietary staple to much of the world and its placement within the grass family make it a valuable target for genome sequencing.
Genome sequencing in most plants is difficult because of the size and complexity of the genomes. Plant genomes range in size from 54 megabases (Mb) for Cardamine amara to 124,000 Mb for a lily (Fritillaria assyriaca) [2]. Although they vary drastically in size, the larger genomes do not correspond to proportionally more genes, but instead to repetitive elements that have blossomed in the plant kingdom [3,4,5,6]. The extremely large genomes of such economically important crops as bread wheat (16,900 Mb), maize (2,600 Mb), soybean (1,100 Mb), and sorghum (735 Mb) [2] make them difficult to tackle with standard methods of genome sequencing such as clone-by-clone [7] and whole-genome shotgun [8]. For example, a whole-genome shotgun project of maize to 8× genome equivalents would require nearly 24 million sequencing reads, and sorghum would require 7.5 million reads. Additionally, the maize and sorghum genomes are more than 75% repetitive [9,10], which would make the final assembly of shotgun sequence extremely difficult [11]. The large-insert clone-by-clone approach solves some of the difficult assembly problems, but it requires a much larger initial investment in resources and is much more expensive. Furthermore, the highly repetitive large-insert clones would still be difficult to assemble.
Evidence has accumulated over the last ten years that many plant genomes are separated into large tracts of methylated repeats and stretches of hypomethylated, low-copy gene–rich space [4,6,12,13,14,15]. On the basis of this knowledge of plant genome architecture, two techniques have been developed to isolate the low-copy or hypomethylated regions of the genome for sequencing. The first technology, high C0
t selection (C0
t is the product of the DNA concentration [C0] and the reassociation time in seconds [t]), allows the separation of low-copy sequences from those of high copy based on annealing rates [16,17]. High C0
t selection has been used successfully to sequence the low-copy, genic regions of maize [18] and has been applied to sorghum [10]. The second technology is methylation filtration (MF), which preferentially clones the hypomethylated fraction of the genome. MF has also been successfully applied in maize to sequence the genic regions [18,19,20]. It appears that MF will be a successful strategy across the plant kingdom, as it has been shown to enrich for genes in all plants tested, from monocots to dicots to gymnosperms, and even in nonvascular plants such as moss (P.D. Rabinowicz, unpublished data).
We have applied MF technology to generate sequence from the hypomethylated portion of the sorghum genome. Successful sequencing of fewer than 550,000 MF reads revealed that approximately 96% of the gene set of sorghum has been sequence-tagged, with an average coverage of 65% across their length. Because MF targets genomic sequence within and around genes, many important components of the genome are represented, including promoters, microRNAs (miRNAs), introns, simple sequence repeats (SSRs), and potentially active transposable elements.
The sorghum gene space is a powerful resource for comparative genomics within the grass family and across the plant kingdom. The MF dataset can be used to confirm hypothetical genes in complete genomes such as rice and Arabidopsis and to identify functional elements conserved across different plant species.
Results/Discussion
The Size of the Genome Space Sampled by MF
To calculate the genome space sampled by MF, two independent methods were used, genome sampling and gene-enrichment. Genome sampling is an empirical calculation based on a modification of the Lander-Waterman equations [21], as used by Whitelaw and colleagues [18]. The reduced genome size is calculated based on the size of the sampled space as judged by the number of times that independent reads overlap. Independent reads will overlap more often when sampling a small region versus a larger region; therefore, one can derive an empirical assessment of the size of the region being sampled [18]. The sampled genome space for the sorghum MF set is 262 Mb.
The gene enrichment method works on the assumption that genes are enriched in the MF libraries in proportion to the reduction in genome size. For example, if the genome is reduced by 3-fold, then gene discovery should occur 3-fold faster in MF versus whole-genome shotgun libraries. The extent to which this number agrees with the genome sampling method is the extent to which the genes reside in the sampled space. We calculate gene enrichment because it can be estimated very early in a sequencing project, whereas the genome sampling method requires at least 0.1× coverage of the sampled space to get an accurate estimate (unpublished data).
The gene enrichment factor is called filter power (FP); we use FP to derive the sampled genome space by dividing it into the size of the whole genome. We calculated the sorghum FP using a subset of our filtered and unfiltered (UF) sequences compared to a curated database of known genes over a range of BLAST Expect values (E-values) (Table 1). The FP is between 3.0 and 3.8 with a median value of 3.15. By dividing this range of FP values into the 735 Mb sorghum genome, the sampled genome is estimated to be between 193 Mb and 245 Mb, with a median of 233 Mb. The median estimate is somewhat lower than the 262 Mb estimation derived by the genome sampling method. However, the result depends critically on genome size estimates, which for S. bicolor range from 735 Mb to 858 Mb [2]. If 858 Mb is used, gene enrichment predicts a 272-Mb gene space, which is slightly higher than the 262 Mb obtained by genome sampling, thus bracketing the genome sampling approximation, depending not only on the range of FP, but on the range of genome size estimates.
Table 1 Gene Enrichment (or FP) of MF Versus UF Sequences
FP was calculated by comparing the MF and UF sequences to a curated set of Arabidopsis proteins, then dividing the proportion of hits in MF by the proportion of hits in UF over a range of BLAST E-values. The median FP is 3.15, with a range of 3.0 to 3.8
Therefore, completely independent estimates of gene space, namely genome sampling and gene enrichment, agree well and are within measurement error. For the purposes of this manuscript, 247 Mb, which is the average of the two methods, will be used as an approximation of the sampled, hypomethylated genome space (Figure 1). The MF dataset consists of a nuclear coverage, after collapsing read pairs, of 285 Mb, which is approximately 1.15× coverage of the sampled space.
Figure 1 Genome Reduction
MF reduces the sorghum genome by 66% in sampling a hypomethylated space of approximately 247 Mb (green) and filtering out 488 Mb (red) of the 735-Mb sorghum genome.
Gene Tagging and Coverage
The purpose of a genome reduction method such as MF is to identify genes in a robust and efficient manner. We assessed the efficiency of gene discovery by calculating the percentage of known genes tagged as a function of read number for MF and compared this value to the rate of gene discovery obtained by expressed sequence tags (ESTs) for sorghum (Figure 2). Additionally, we conducted a simulation in Arabidopsis to assess the expected gene identification rate in a completed plant genome where the level of coverage could be controlled precisely in silico (see Expected Gene Tagging, below). The results of these analyses are summarized in Figure 2.
Figure 2 Gene Discovery Rate
Gene discovery rates for sorghum MF (blue), sorghum ESTs (pink), and an Arabidopsis simulation (dotted black) are shown. The gene discovery rates for the MF and ESTs were calculated based on matches to a set of 137 genes annotated on sorghum BAC clones versus the number of MF and EST reads. The Arabidopsis simulation was calculated based on the fold-coverage of chromosome 1, which contains 7,520 genes. The fold coverage was converted into read numbers as detailed in the Materials and Methods.
To estimate the percentage of genes that have been tagged by MF, we used high-quality sorghum bacterial artificial chromosome (BAC) sequences as a source of gene annotations. At the time of analysis, 14 finished sorghum BACs had been deposited in GenBank (http://www.ncbi.nlm.nih.gov/). Because the GenBank annotations were outdated, we reannotated the BACs through a custom annotation pipeline (see Materials and Methods). We annotated a total of 148 genes on these BACs, then mapped our MF reads to the BACs using stringent BLAST criteria. Of the 148 genes, the MF reads match 133 (90%) of them, with an average nucleotide coverage of 61%.
However, 11 of the 148 annotations are alpha kafirin storage protein genes on BAC AF527808. Ten of them constitute a tandem repeat cluster of nearly identical sequences that could be expected to be methylated [22] and are therefore not recovered efficiently in a MF library. This is indeed the case, as only two out of the 11, or 18%, are recovered in the MF clones. This is far below the 90% average for the whole set, suggesting that the kafirin genes may be at least partially methylated (see Methylated Gene Recovery, below). If we remove these 11 genes from the analysis, 131 (95.6% [Figure 2]) of the remaining 137 genes are tagged across 65% of their nucleotides. We also removed the kafirin genes from the EST analysis in Figure 2.
In addition to tagging 95.6% of the gene set, a majority of the coding sequence (CDS), upstream, and downstream genomic regions are covered. The average coverage of the CDS regions of all 137 genes is 65%, thus providing a tag across more than half of the gene on average. This coverage is consistent with the 67% nucleotide coverage predicted at 1.15× raw sequence coverage [21]. Additionally, we calculated the nucleotide coverage 500 basepairs (bp) upstream (5′) and downstream (3′) of the CDS and found 74% and 69% coverage, respectively. The coverage of the 5′ and 3′ regions is higher than expected, which is at least partly due to the close spacing of sorghum genes in this set, with 16/137 (greater than 10%) having 5′ and/or 3′ regions within 1 kb.
For comparison, the gene tagging ability of the publicly available sorghum EST sequences was assessed. At the time of analysis, there were 161,766 sorghum ESTs deposited in GenBank. Using criteria of 98% identity over at least 50 bp of the CDS, the sorghum ESTs matched 84/137 (61%) of the annotated BAC genes (Figure 2). Notably, the ESTs did not match any of the 11 kafirin genes.
Expected Gene Tagging: An Arabidopsis Simulation
If MF faithfully represents the genic region of sorghum and contains the vast majority of the genes, then the rate of gene tagging should produce results that are similar to whole-genome shotgun coverage [21] at the same level of raw coverage. To test this hypothesis, we simulated a whole-genome shotgun project of the finished Arabidopsis chromosome 1 (see Materials and Methods). We decided to use Arabidopsis for the simulation because it is finished to high quality, the gene predictions are the most robust of any plant species, and Arabidopsis best represents the size of plant genes, which are much smaller on average than animal genes.
The simulation showed that, at 1.1× coverage, 96.4% of the genes are sequence-tagged across 66.8% of their length. These numbers are very similar to the percentages calculated from the MF gene tagging analysis (95.6% of genes covered over 65%). Since the simulation is set to replicate Lander-Waterman whole-genome shotgun conditions, these results mean that MF obeys the mathematics of Lander-Waterman, although it is a highly fragmented sampling space. Furthermore, if the BAC gene set is representative of the genome, this implies that nearly all the genes in the genome are accessible to MF and that all genes are currently covered over an average of 65% of their length. Theoretically, 100% nucleotide coverage will be reached at 6× coverage, which would require less than 2.5 million additional MF reads.
Figure 2 shows the comparison of the gene tagging rates for the Arabidopsis simulation, the MF reads, and the sorghum ESTs. Notice that the gene tagging for the sorghum ESTs and MF are more rapid than the Arabidopsis simulation. Rather than reflecting a real difference in ability to tag genes using MF versus whole-genome shotgun, this higher rate likely reflects the larger average gene length for the sorghum CDS annotations (3 kb) versus Arabidopsis (2.3 kb), making gene tagging more rapid in sorghum. Additionally, the sorghum ESTs show the most rapid gene-tagging rate, but begin to level off at 60% gene tagging and are passed by the sorghum MF after 70,000 reads.
Methylation of Transposons, Repeats, and Pseudogenes
Overall, recognizable repeats constitute 62% of the sorghum genome (Table 2, Unfiltered), which is comparable to maize [18,19]. Retrotransposons are the most abundant class of repetitive DNA sequence, occupying about 1/3 of the genome, followed distantly by DNA transposons at 1/20 of the genome (Table 2, Unfiltered). MF reduces the recovery of ribosomal genes, centromeric repeats, and retrotransposable elements (Table 2, Filtered), so that only 27% of filtered reads match repeats. These results can be described in terms of the total fraction of repeats (R/N, where R is the total length of repeats in the genome and N is the size of the genome), the unmethylated fraction of repeats (r/UM, where r is total length of repeatsin the unmethylated fraction and UM is the size of the unmethylated genome), and the filter power (FP) (N/UM) according to Palmer and others [19]. Given a FP of 3.15 (N/UM), we can calculate the proportion of unmethylated repeats (r/R) as ([r/UM]/[R/N])/(N/UM), or approximately 10%. This is consistent with maize [19], and indicates that a substantial portion of sorghum transposons, especially DNA transposons, are unmethylated and may be capable of transposition. For example, the active sorghum transposon Candystripe1(Cs1) [23] is represented in our dataset across 23% of its length (unpublished data). The lower-than-average percent coverage (23% versus 66%) may be due to some methylation within the element, as has been reported for several maize transposons [12]. Additionally, Cs1 is known to have a low copy number (less than 10) in sorghum, and the redundancy of coverage across the 23% represented suggests that MF is sampling from a single element (unpublished data).
Table 2 Repeat Analysis for MF Versus UF Reads
a Note that the MF sequences sample approximately 1/3 of the genome, so this percentage of repeats reflects 1/3 of the total genomic content
MITES, miniature inverted terminal repeat elements
The majority of methylation in plants occurs at the canonical sites CG and CNG (where N is any nucleotide) [24,25,26,27]. MF uses in vivo restriction via modified cytosine restriction, subunits BC (mcrBC) at the recognition site (A/G) methylated cytosine (mC). The observed versus expected occurrences of mcrBC sites, along with those sites that overlap the canonical methylation sites of CG and CNG, are shown for retrotransposons and genic sequences in Figure 3A and 3B, respectively. Although the mcrBC half-sites ([A/G] C) occur as expected in MF and UF retrotransposons and genes, the sites that overlap canonical methylation sites are significantly reduced in MF versus UF retrotransposons, but not in genic sequence, where, in fact, they occur more frequently in MF than UF (Figure 3). It has been shown previously that CG and CNG nucleotides are suppressed in MF repetitive elements [19], presumably because mCs have been converted over time to thymine by deamination [28]. Our results suggest that such conversion has occurred in transposon sequences, but not in genes, consistent with their differential methylation.
Figure 3 CG and CNG Suppression in MF versus UF Sequences
Sequences were analyzed for their mcrBC half-sites, those that overlap CG dinucleotides, and those that overlap CNG trinucleotides. The ratio of observed to expected sites is graphed for filtered (hatched) and unfiltered (white) for retrotransposons (A) and CDSs (B).
The increased frequency of CG and CNG nucleotides in genic sequences recovered by MF versus UF (Figure 3B) suggests that CDS derived from MF and UF are different. One source of this difference may be the presence of pseudogenes. In plants, most pseudogenes are marked by small insertions and deletions, resulting in frame shift(s) of the coding region, but are otherwise indistinguishable from functional genes [29]. Pseudogenes are likely targets of silencing and are thus probably methylated, excluding them from MF sequences. To test if pseudogenes are more abundant in the UF dataset, sequences from both UF and MF that matched Arabidopsis proteins, and are therefore considered genes, were compared to a database of all plant proteins using BLASTX. Sequences with more than one high-scoring segment pair and with an E-value of 1 × 10−20 or less were analyzed for the presence of a frame shift. The rate of potential frame shifts for UF is 103/530 (19.4%) versus 1,599/17,103 (9.35%) for MF, indicating that pseudogenes are recovered at a higher rate in UF (comparable to the rate of retrotransposons) and are therefore most likely methylated.
Methylated Gene Recovery
Comparison with the BAC sequences revealed that a small number of genes were not represented in the sorghum MF reads. Two explanations were considered: First, these genes may have been missed by chance, as only 97% of sorghum genes were expected to be sampled by this depth of coverage. Second, these genes might be methylated. Two examples were chosen for further analysis: the teosinte branched2 gene (tb2), which was recovered in our dataset, and the kafirin storage protein gene cluster (Figure 4). The kafirin gene cluster was chosen because it is underrepresented in the MF sequences and could be methylated since it is a tandem repeat cluster [22]. We used a real-time PCR technology to assess DNA methylation (see Materials and Methods). As expected, methylation analysis of tb2 (on BAC AF466204) indicates that it is unmethylated (Figure 4A and 4C).
Figure 4 Methylation Status of tb2 and Kafirin Cluster
(A and B) Restriction maps of the tb2 gene (A) and the kafirin consensus sequences (B) are shown. The relevant restriction sites are indicated vertically and the numbers indicate the distances scale in basepairs. Each CDS is depicted as a blue-shaded arrow, and the region assayed is indicated by a black bar. The circles depict sites that are not present in every kafirin gene, and the color represents the number of genes that do not share the site. The orange circle (5′-most HhaI site) is a site conserved in nine of 11 kafirin genes, and the red circle (3′-most PstI site) is a site present in ten of the 11.
(C) Results from a representative methylation analysis of tb2; the inset depicts the template dilution standard curve used to set the threshold for the experiment. Each experiment was performed three times with four on-board replicates per assay point. The results for each of the four differentially treated reactions are depicted with different colors. Red, mock-treated; blue, mcrBC-digested; orange, HhaI-digested; and green, HhaI + mcrBC double-digest. The inset shows the standard dilution control with two replicates at each dilution. The control was used to set the threshold for detection. The specificity of each reaction was confirmed using melt-curve analysis.
(D) Results from a representative methylation analysis of the 11 kafirin genes. The results for each of the six differentially treated reactions are the same as in (C), with the following additional digests: pink, PstI-digested; light blue, PstI + mcrBC double-digest. Notice that the mcrBC with and without PstI yields the same Ct, while HhaI + mcrBC (green) yields a higher Ct on average; suggesting additional cleavage.
For the kafirin gene cluster, only two of 11 genes from BAC clone AF527808 were represented in the MF dataset, suggesting that most or all of them may be methylated. Ten of the genes are tandemly arrayed in a cluster and share an average of 99.1% sequence identity, while the eleventh gene is located 45 kb away and is more diverged (76.2% identity on average). A 247-bp region was selected for PCR close to the 5′ end because of its near identity across all 11 genes and because of the high CG and CNG content (Figure 4B). The methylation results are depicted in Figure 4D. PstI sites are methylated (at CNG), since the PstI-treated sample (Figure 4D, pink) has the same cycle threshold (Ct) as the mock-treated sample (Figure 4D, red). This result is supported by the mcrBC digested sample, which has a significantly higher Ct (Figure 4D, dark blue) than the mock-treated DNA control. All, or almost all, of the PstI sites are methylated, because the double PstI +McrBC digest (Figure 4D, light blue) has the same Ct as mcrBC alone (Figure 4D, dark blue). These results indicate that every gene has CNG methylation covering these sites.
As for CG methylation, the HhaI-digested (orange) sample has the same Ct as the mock-treated control (red); however, the Ct of the HhaI + McrBC double digest (green) is 2.46 cycles greater than the mcrBC alone (dark blue), indicating that some HhaI sites must not be modified. A cycle threshold difference of 2.46 indicates that there is 22.46, or approximately 5.5-fold, less DNA in the HhaI + mcrBC double-digested sample. This suggests that two out of the 11 kafirin genes have some unmethylated HhaI sites.
To determine which kafirin genes might be unmethylated, we sequenced the kafirin PCR products from mcrBC treated and untreated genomic DNA (gDNA). 130 sequences from mcrBC-treated DNA and 126 sequences from the mock-treated sample were analyzed. The kafirin genes fall into “subfamilies” based on six polymorphisms within this highly conserved genomic region (see Materials and Methods). Each of these subfamilies was represented among the sequenced clones, including the orphaned kafirin gene outside the tandem array, indicating that none was completely removed as a consequence of mcrBC treatment. Thus, it is likely that all the kafirin genes contain some level of methylation, and that the genes are displaying nonuniform CG methylation randomly, perhaps on a per-cell basis, across all 11 genes.
Drought Resistance Genes
In order to assess how useful the current low level (approximately 1×) coverage of the gene space is for answering important comparative genomics questions, we chose to analyze genes related to drought resistance. Sorghum's ability to grow in arid conditions makes it an attractive source of genes to enhance drought resistance in other grasses. Part of the drought-responsive pathway in plants involves the activation of dehydration-responsive element binding protein (DREB) transcription factors belonging to the APETALA2 (AP2) family. The overexpression of DREB1-encoding genes can promote drought, freezing, and salinity tolerance in transgenic plants [30].
A screen of the sorghum MF dataset reveals five full-length DREB1-like proteins, based on conservation of the AP2 domain and a conserved C-terminal LWSY motif (see Materials and Methods). A phylogenetic tree constructed from the AP2 domains of the Arabidopsis, rice, and sorghum DREB1-encoding genes suggests that sorghum has expanded the DREB1 family and that SbDREB1–1 and SbDREB1–2 are the closest orthologs to the Arabidopsis DREB1 family (Figure 5). This analysis also suggests that the rice gene OsDREB1D may not belong to the DREB1 family, a hypothesis supported by the fact that OsDREB1D does not contain the conserved LWSY motif and its expression was not detected under drought, freezing, or salt-stress conditions [31]. An expansion of the DREB1 family in sorghum may contribute to the plant's enhanced drought resistance. Certainly the identification of other sorghum genes involved in the drought response regulatory pathway is now possible. This analysis highlights the utility of this dataset in answering fundamental comparative biology questions even at such a low level of gene space coverage.
Figure 5 Phylogenetic Comparison of Sorghum DREB1 Genes
A phylogenetic tree comparing the AP2 domain of the sorghum DREB1 genes to those of Arabidopsis and rice was constructed using CLUSTALX [61]. The genes encoding proteins from Arabidopsis are DREB1A, DREB1B, and DREB1C. Rice genes are OsDREB1A, OsDREB1B, OsDREB1C (nucleotides 142,337–142,981), and OsDREB1D (nucleotides 1,489–2,250). AP2 domains from other Arabidopsis proteins are also included: APETALA2 (R2 domain), AtERF-1, LEAFY PETIOLE, and TINY.
Global Comparisons to Rice and Arabidopsis
In order to assess the utility of the sorghum MF set for cross-genome annotations, we compared the annotation of rice by sorghum MF versus the complete gene set in Arabidopsis. The rice genes were downloaded from The Institute for Genomic Research (TIGR) and contain the genomic sequence of gene predictions, which includes exons and introns. The rice set contains 57,535 genes that we categorized into known (23,115), hypothetical (21,438), and repetitive (12,982), based on the annotation (see Materials and Methods).
The rice sequence was used as the query in searches of sorghum MF and Arabidopsis proteins. A rice gene was considered supported if it had a best match better than or equal to a BLAST E-value of 1 × 10−8. Of the rice gene set, 46,450 (81%) had a match to sorghum MF, while only 38,462 (67%) matched Arabidopsis. The matches can be further broken down by category, with 22,282 (96%) of known rice genes, 13,262 (62%) of hypothetical rice genes, and 10,906 (84%) rice repeats matched by sorghum MF. In comparison, Arabidopsis annotated 20,827 (90%) known, 7,850 (37%) hypothetical, and 9,785 (75%) repeats. Thus, the 1.15× coverage of the closely related sorghum gene space does a much better job of providing supporting evidence for gene predictions in rice than does Arabidopsis. Interestingly, the number of hypothetical genes matched by sorghum MF is almost 2-fold higher than that annotated by Arabidopsis. This may indicate a higher proportion of grass-specific genes in the hypothetical predictions.
To understand how well cross-species gene annotation is accomplished in a low-redundancy MF versus a nearly complete genome, we compared the annotation of Arabidopsis by sorghum MF to that by rice. Such a comparison provides a good test of annotation capacity without being complicated by different evolutionary distances, since Arabidopsis, being dicotyledonous, is expected to be the same evolutionary distance from both sorghum and rice.
An Arabidopsis protein was considered supported if it had a BLAST match less than or equal to an E-value of 1 × 10−8 (Figure 6). In this analysis, 19,700 (84%) of the known and 1,664 (38%) of the hypothetical proteins had a match to sorghum MF, whereas 21,093 (90%) of the known and 1,979 (45%) of the hypothetical proteins had a match to rice. This indicates, as expected, that a complete monocot genome is a better tool for annotating a dicot than is a partial genome; however, the difference is not that big, suggesting that a low level, cost-effective skim of many different genomes for comparative genomics may be more economical than complete sequencing.
Figure 6 Annotation of Arabidopsis by Sorghum MF Versus Rice Gene Sequences
Shown are the number of Arabidopsis proteins that are matched in a TBLASTN comparison to the sorghum MF set (blue) versus the rice gene sequences (yellow). The Arabidopsis proteins, after having known repetitive elements removed (see Materials and Methods), have been categorized as either hypothetical or known based on the definition line. Arabidopsis proteins were considered supported if they matched with an E-value less than or equal to 1 × 10−8. Sb, S. bicolor MF set; Osj:seq, Oryza sativa japonica gene sequences.
Interestingly, although the rice sequences match more Arabidopsis proteins than sorghum, the set is not completely overlapping, and sorghum matches 247 proteins that are not matched by the rice sequences. Since we used rice gene predictions as our database for comparison, it is likely that some of the Arabidopsis proteins are in the genome but are not annotated as genes. To address this possibility, we compared the 247 Arabidopsis proteins to the entire rice genome (Oryza sativa japonica) and found that 59 did indeed match to the bare gDNA versus the annotations, and therefore were not unique to the sorghum-Arabidopsis genomes. That left 188 proteins that may be conserved in sorghum and Arabidopsis, but not in rice. The O. s. japonica genome was sequenced by the BAC-by-BAC method [32], and it is likely that some regions are not represented in the BAC clones. Therefore, we compared these 188 to the O. s. indica genome, which was sequenced by whole-genome shotgun [33] and would have different biases than BAC-by-BAC. Again, a proportion (61) of these were found in the genome under our BLAST criteria, leaving 127 Arabidopsis proteins that are supported by sorghum but either missing from or significantly diverged in the current versions of the O. s. japonica and O. s. indica genomes (Table S1). Laboratory experiments will be needed to confirm that these are truly missing from rice; if they are missing, they represent an interesting set of genes that could highlight previously unknown shared features between sorghum and Arabidopsis to the exclusion of rice. For example, the myb-related protein CAPRICE, a gene involved in root-hair cell development [34,35], was in this set, which may indicate a previously unknown conserved root development pathway in sorghum and Arabidopsis to the exclusion of rice.
MiRNAs
MiRNAs are a class of small RNAs that are important in gene regulation through recognition and cleavage of target mRNA. They are short sequences, usually 18–24 nucleotides in length, that match target genes and gene families, although usually imperfectly. Regulation is achieved through cleavage by the RNAi silencing complex. They are encoded by hairpin precursors that are processed in at least two steps by RNase III-domain ribonucleases related to Dicer. MiRNAs have been found in all eukaryotes surveyed and seem to be well conserved between plant species [36,37,38].
We downloaded 122 and 92 known rice and Arabidopsis miRNAs, respectively [39], and used them in a BLAST search against the sorghum MF set. Of these, 91 (75%) of the rice miRNAs and 44 (48%) of the Arabidopsis miRNAs had exact matches in the sorghum MF set (Table 3). For comparison, the miRNAs were searched against the completed rice genome, sorghum ESTs, and maize MF + HC (high C0
t) assemblies, with 121, 16, and 88 of the rice miRNAs and 52, 10, and 46 of the Arabidopsis miRNAs matching, respectively.
Table 3 MiRNA Content in Sorghum, Rice, and Maize
a Although the 122 miRNAs were reported by Jones-Rhoades and colleagues [39] in the O. sativa japonica genome, we were not able to find a perfect match for MIR395f, although there are several nearly identical matches
To ensure that these were authentic matches and not just due to chance, we performed a test with shuffled miRNA sequences, maintaining the nucleotide composition (see Materials and Methods). None of the shuffled sequences matched any of the databases, indicating that the matches are authentic and not due to the small size or a biased nucleotide composition of the miRNAs. Additionally, precursor sequences surrounding these miRNAs could form hairpins (Figure 7 and unpublished data), and were also matched by rice gDNA, indicating they are likely to encode the corresponding miRNA.
Figure 7 Secondary Structure of Predicted MiRNAs
Predicted hairpin secondary structure of miRNA MIR156a from rice and the newly discovered ortholog from sorghum. The 21-nucleotide MIR156a sequence is highlighted in red.
We do not know a priori how many of the rice miRNAs would be expected to be conserved in sorghum, but we can assume that most, if not all, of the miRNAs conserved between Arabidopsis and rice would also be conserved between Arabidopsis and sorghum. Therefore, given that the rice genome is nearly complete, we expect to find the same 52 Arabidopsis miRNAs in sorghum, and we have identified 44 (85%). The eight that are missing may be present in the data but not identified because of sequencing errors; not yet sampled, as we expect only approximately 66% of the nucleotides to be present at this level of coverage; or some of these eight may represent miRNAs conserved in rice but lost in sorghum.
Simple Sequence Repeats
SSRs are stretches of DNA with simple sequence pattern repetitions, usually in the form of di-, tri-, or tetra-nucleotide expansions such as (CA)n, (CAG)n, or (
GATA)n. These stretches of DNA are useful for genetic marker analysis, because they are unstable and often are polymorphic between closely related individuals [40,41]. Overall, SSRs are enriched in MF sorghum sequences, 22,445 of 417,113 (5.4%), compared to UF, 335 of 17,276 (1.9%), indicating that most SSRs are unmethylated. GC-rich trinucleotide repeat (TNR) SSRs in plants have been shown to be preferentially associated with coding regions [42,43]. We observe an increase in the proportion of GC-rich TNRs to total TNRs in MF sequences, 6,464 of 8,957 (72%), compared with whole-genome shotgun, 63 of 129 (49%). This observation suggests that this collection of sorghum sequences is laden with new and publicly available molecular breeding and genetic mapping tools.
The Sorghum Genome and Comparative Genomics
The sequence of the sorghum gene space provides an excellent tool for comparative genomics [44]. Unlike maize, which it otherwise resembles, sorghum has not undergone recent genome duplications, although there is evidence for ancient duplications in most cereal genomes [45]. For this reason, sorghum and rice share a greater degree of colinearity than maize and rice [40], potentially facilitating mapping of quantitative traits across these three genomes, including drought resistance [46]. Sorghum is also a close relative of sugarcane (Saccharum spp.), whose large and variable chromosome content makes genome sequencing impractical. The availability of a large number of sugarcane EST sequences [47] will enable comparison of these genomes to identify genes of potential agronomic value in this species as well. Such comparisons will extend even to the large collection of microsatellite SSR markers reported [41]. The sequence reported here is an important first step for these comparisons.
The sorghum gene set present in the MF data is very nearly complete, as illustrated by the ability to annotate Arabidopsis nearly as well as the completed rice genome and by the ability to identify 95% of the genes from finished sorghum BACs. This was achieved with a minimal sequencing effort, which brings within reach the prospect of sequencing multiple strains of the same species. Such a feat is of critical importance in maize, in which inbred lines differ substantially in gene order and content [48,49].
Sequencing Large Plant Genomes Using MF
A disadvantage of gene enrichment strategies, whether they are EST sequencing, high C0
t selection, or MF, is that the recovered fragments are not positioned on the genome. Mapping has to be accomplished by either mapping the reads to a physical or genetic map or by combining the gene enrichment with an anchored clone map. MF reads are enriched for SSRs, which make good genetic markers and allow some reads to be placed on a genetic map. If a framework physical map of fingerprinted BAC clones exists, then MF can be easily integrated onto the physical map in three ways: PCR mapping to BAC pools, hybridization to BAC filters, and/or by sequence integration. Sequence integration can be accomplished using either BAC end sequence or shotgun sequence from a representative tiling path of the BAC contigs. While there is no whole genome BAC map of sorghum yet available, a robust map is almost complete in maize [50,51]. It is estimated that a BAC tile of maize will consist of approximately 18,000 BAC clones. Skim-sequencing from these clones at approximately 1× coverage, combined with a deep coverage through gene enrichment, are predicted to generate a high-quality sequence map for a fraction of the cost of whole genome sequencing [48]. BAC sequencing projects are ongoing for sorghum [40], which can use the MF reads in much the same way to enhance the BAC shotgun sequence and speed the completion of the genome.
Materials and Methods
MF library construction
Seeds of S. bicolor ATX623, kindly provided by J. Osborne (NC+ Hybrids, Colwich, Kansas, United States), were germinated and grown in soil under growth chamber conditions. Then gDNA was purified from isolated nuclei of 1-mo-old leaves as described [52], except that OptiPrep (Axis-Shield PoC, Oslo, Norway) was used. Shearing of nuclear DNA was performed using either a nebulizer (Cis-Us, Bedford, Massachusetts, United States) or Hydroshear (GeneMachines, San Carlos, California, United States). Sheared fragments were end-repaired using a variety of enzymes including mungbean nuclease, T4 DNA polymerase, Klenow fragment, and T4 polynucleotide kinase. End-repaired fragments were size-selected on an agarose gel and DNA fragments ranging from 0.7 to 1.5 kb were extracted and ligated to dephosphorylated, HincII-digested pBC SK– vector (Stratagene, La Jolla, California, United States) which was used to construct both MF (GeneThresher technology; Orion Genomics, Saint Louis, Missori, United States) and UF libraries. Ligation reactions were transformed into mcrBC+ and mcrBC– strains of E. coli for generation of MF and UF libraries respectively. Recombinant clones were picked using Genetix Q-bot robot (Research Genetics, Carlsbad, California, United States) and stored individually in 384-well microtiter plates.
Sequence data
Two sources of MF sequencing reads were used. Out of 604,641 attempts at Orion Genomics, 532,150 were successful (accession numbers CL147592–CL197752 and CW020594–CW502582), 514,983 of which are considered of nuclear origin based on comparison with chloroplast, mitochondrial, viral, and bacterial databases. Additionally, we have included 36,825 sorghum MF reads previously generated at Cold Spring Harbor Laboratories (Cold Spring Harbor, New York, United States) (accession numbers CC058553–CC059980, BZ329127–BZ342789, BZ342901–BZ352342, BZ365856–BZ368372, BZ369686–BZ370012, BZ421595–BZ424357, BZ625682–BZ629992, and BZ779555–BZ781928). The sorghum UF sequences also came from two sources: Orion Genomics (1,819 reads) (accession numbers CW512190–CW514008) and the University of Oklahoma (15,889 reads) (NCBI TraceDB accession numbers TI566112507–TI566128395). The average read lengths were 600 bp and 550 bp for each class of reads, giving a total, raw nuclear dataset of approximately 330 Mb (MF) and 10.5 Mb (UF). The MF dataset was further collapsed by assembling overlapping read pairs to generate a set of independent sampling events comprising approximately 285 Mb.
Database curation and FP calculation
We have done a first pass definition-line curation of publicly available sequence databases to eliminate obvious transposon sequences that would hamper subsequent analyses by virtue of inflating the true “gene” content of the given database.
The Arabidopsis protein set, which was used for the gene enrichment calculations and assessment of cross-genome annotation potential, was downloaded from the NCBI (ftp://ftp.ncbi.nih.gov/genomes/Arabidopsis_thaliana/CHR_*/*.faa). The files were dated 23 May 2003 and contained 28,581 sequences (12,112,846 total letters). Repeats were removed from this dataset if the definition line meets both of the following two criteria: (1) Matched the case-insensitive regular expression “/retro|mutator|transpos|reverse transcriptase|polyprotein|\bgag\b|BARE-1|athila/”, and (2) did not match “/\[.*retro.*\]|leucine|WD-repeat|WD repeat|WD40|WD-40| ankyrin|telomere|arm repeat|PPR-repeat|armadillo|tetratricopeptide|TPR-repeat|TPR repeat|Kelch|pentapeptide|C-repeat/”.
This second step was used to replace falsely identified nonrepetitive elements. Removing repeats reduced the database size by 640 sequences to 27,941, which included 4,412 sequences identified as hypothetical by matching the definition line to the case-insensitive regular expression “/hypothetical/.”
The rice sequence set was downloaded from TIGR (ftp://ftp.tigr.org/pub/data/Eukaryotic_Projects/o_sativa/annotation_dbs/pseudomolecules/version_2.0/all_chrs/all.seq). The file was dated 30 April 2004 and contained 57,535 sequences (155,419,428 total bases). The rice sequence set contains the genomic regions for all predicted rice genes, which includes exons, introns, and untranslated regions where good evidence is provided. No sequences were removed from this database, but they were classified as “repeats,” “hypothetical,” and “known” by the following criteria. (1) Sequences were classified as repeats if the definition line matched a case-insensitive regular expression “/retro|transpos|reverse transcriptase|gag|polyprotein|mutator|maggy|rire|gypsy|copia|bare-1/”; (2) the sequences were classified as hypothetical if the definition line matched a case-insensitive regular expression “/hypothetical/”; and (3) the remaining sequences were classified as known. In total, there were 13,008 repeats, 21,441 hypotheticals, and 25,263 known proteins. The rice chromosomal genomic sequences were used for miRNA identification (ftp://ftp.tigr.org/pub/data/Eukaryotic_Projects/o_sativa/annotation_dbs/pseudomolecules/version_1.0/all_chrs/all.con) and dated 05 September 2003. It contains 12 chromosomes, with sequences comprising 358,546,960 bases.
Sorghum ESTs were download from the NCBI (ftp://ftp.ncbi.nih.gov/genbank/*.seq.gz and ftp://ftp.ncbi.nih.gov/genbank/daily-nc/nc*.flat.gz, which was last dated 20 October 2003, and contains 161,766 sequences (83,411,684 total bases). No sequences were removed from this database.
Gene enrichment was calculated by comparing the rate of gene discovery between MF and UF sequences. To ensure high quality, unique sampling events, reads were chosen that contained at least 100 contiguous Phred Q20 bases and only one read per clone. Detection of genes was accomplished by an NCBI-BLASTX search (parameters: -e 0.01; -b 5; -v 5) of the curated Arabidopsis protein database (see Materials and Methods). Aside from the curation of the Arabidopsis database to remove repetitive elements, matches to proteins annotated as hypothetical were not counted. Hypothetical genes are often false gene predictions or unknown repetitive elements. In order to calculate a gene enrichment factor, or FP, the proportion of matches from MF sequences are compared to the proportion of matches in UF sequences over a range of E-values from 1 × 10−5 to 1 × 10−20, such that all matches better than the given E-value are tabulated (Table 1). For sorghum, the genome size is estimated at 735 Mb [2]. Dividing the genome size by the median 3.15 FP provides an estimate of a 233 Mb sampled space.
BAC annotation
There were 14 finished BAC clones at the time of analysis, with the following accession numbers (and GenInfo identifiers). AC120496.1 (GI:20486389), AF010283.1 (GI:2735839), AF061282.1 (GI:4539654), AF114171.1 (GI:4680196), AF124045.1 (GI:5410347), AF369906.1 (GI:19851516), AF466199.1 (GI:18390096), AF466200.1 (GI:18481699), AF466201.1 (GI:18483227), AF466204.1 (GI:18568251), AF503433.1 (GI:21326110), AF527807.1 (GI:22208458), AF527808.1 (GI:22208471), and AF527809.1 (GI:22208503). The BACs were manually annotated, then reads were mapped to the BACs by BLAST to determine the locations of hits relative to the genes.
We analyzed the BACs with several computational tools in addition to manual editing. Repetitive elements were identified using RepeatMasker [53] with the MaskerAid speed enhancement [54] and the TIGR cereal repeat database. The TIGR cereal repeat database, dated 11 July 2003, was downloaded (ftp://ftp.tigr.org/pub/data/TIGR_Plant_Repeats/) and contained 11,043 repeat entries. RepeatMasker was run with the following parameters: “-s; -w; -no_is; -nolow; -lib cereal_repbase.lib”. RepeatMasker parameters also included “-xsmall” to mask in lowercase and “-w” to use the MaskerAid [54] enhancement. To look for known protein-coding genes, we searched each repeat-masked BAC against all plant proteins with WU-BLASTX 2.0MP-Washu (23 May 2003) [55,56] using a serial strategy [57]. The first search used the parameters W=5; V=0; E=1e-5; X=10; nogap; kap; altscore=“* any na”; altscore=“any * na”; wordmask=seg; lcmask. The second search used default parameters. To look for transcript similarities, we searched all plant transcripts with WU-BLASTN using a serial strategy with the following first-round parameters: W=12; V=0; X=7. In the second round we used the parameter: W=9. Both BLASTN searches had these additional parameters: wordmask=seg; lcmask; M=1; N=–1; Q=3; R=3; kap; E=1e-10; hspmax=0. To look for potentially novel genes, we used Fgenesh (http://www.softberry.com/berry.phtml) with monocot parameters, Genscan [58] with Arabidopsis parameters, and SNAP [59] with Arabidopsis parameters.
In order to annotate the locations of genes in each BAC, we loaded all the computational results into the ACEDB viewer (http://www.acedb.org) and edited gene structures by hand. One of the challenges was how to determine when the tools had identified pseudogenes. These are often marked by adjacent repeats, BLASTX alignments containing stop codons, or gene predictions with tiny introns to compensate for frame-shifts. Another challenge was how to use cross-species alignments. Alignments that are nearly identical to genomic sequence are useful for delimiting exon boundaries, but inexact matches pose problems because alignments may terminate because of real exon boundaries or differences between the sequences. Since most of the alignments were from plants other than S. bicolor, we did not employ any programs that align a protein or transcript directly to a genome. Instead, we assigned the position of the splice sites in part by consulting exon predictions, since gene finding algorithms contain probabilistic models of splice sites. We did not report any raw gene predictions. However, some genes do contain exons with no overlapping evidence and are included in the gene structure because they complete an otherwise incomplete gene structure and in some cases are necessary to maintain the reading frame. The BAC annotations are available in GFF format with the supplemental online data.
The sorghum MF sequences were compared to the collection of 14 sorghum BACs using NCBI-BLASTN (parameters: -p blastn; -F ‘m D'; -e 0.01; -b 14; -v 14). A sequence was considered mapped to a BAC if the match was over 90% of the read with 98% identity or higher. A single read was mapped to only one location. A gene was considered tagged if one or more MF sequence(s) overlapped the CDS region by 50 bases or more. The set of S. bicolor ESTs were mapped to the BACs using the same BLAST parameters, but a gene was considered tagged using less stringent criteria, since genomic introns will not align. A gene is tagged by an EST if it aligns at 98% identity over at least 50 bp, but there was no requirement for the percentage of the EST that needed to be aligned.
Arabidopsis simulation
A computational simulation of shotgun sequencing Arabidopsis chromosome 1 was compared to the empirical gene tagging results in sorghum. The sequence and annotation of Arabidopsis chromosome 1 was downloaded from TIGR (ftp://ftp.tigr.org/pub/data/a_thaliana/ath1) on 20 February 2004. The chromosome is approximately 30 Mb long with 7,520 genes annotated. The median gene size is 1,960 bp.
Computationally generated “reads” of 700 bp in length were created from the chromosome for different levels of raw coverage from 0.5× up to 3.5×. The reads were then mapped back to the chromosome annotation to determine the percentage of the 7,520 genes that were tagged at each level of raw coverage (results shown in Figure 2). The percent gene tagging was calculated on a fold-coverage basis (e.g., 0.5×, 1.0×, etc.), so in order to convert it to a meaningful read number basis for Figure 2, we converted the fold-coverage to a number of reads by using the estimated genome space (247 Mb) divided by the average sorghum read size (604 bp), resulting in approximately 409,000 reads per 1× coverage.
MiRNA analysis
The A. thaliana and O. sativa miRNAs were downloaded from the supplementary online material for Jones-Rhoades and colleagues [39]. This dataset contains 122 and 92 computationally predicted and experimentally confirmed miRNAs for Oryza sativa and Arabidopsis thaliana, respectively. The miRNAs are grouped into 18 and 22 families for rice and Arabidopsis, respectively. These sequences were used in a WU-BLASTN [55] search of the MF sorghum set (parameters: -W 18; -M 1; -N −4; -Q 1; -R 1; -wordmask=seg; -warnings). A match was scored if the miRNA matched at 100% identity over its complete length. The same parameters were used for the rice genome, sorghum ESTs, and maize MF + HC databases. The maize MF + HC database is release 4.0 of the Zea mays MF and HC combined assembly from TIGR (http://www.tigr.org/tdb/tgi/maize/).
In order to test the specificity of these miRNA matches, we generated shuffled sequences for the 122 rice and 92 Arabidopsis miRNAs. The shuffling maintains the nucleotide composition of each while scrambling the order [60]. The shuffled sequences were used in WU-BLAST searches against all the databases with the same parameters as above. None of the shuffled sequences had an identical match to any database. These results indicate that the miRNAs are not matching simply because of their small size and nucleotide composition, but probably represent authentic evolutionarily conserved units.
Comparison with rice and Arabidopsis.
The rice sequences were compared to the sorghum MF using NCBI-BLASTN with the rice sequences as the query and the sorghum MF reads as the database (parameters: -p blastn; -b 1500; -v 1500; -r 1; -q -1; -G 2; -E 1; -F ‘mD'; -e 1e-5). We counted a rice gene as hit if the E-value was less than or equal to 1 × 10−8, which corresponds to a bit score of approximately 61. The rice hits were then counted and categorized.
To assess how well rice is annotated by a dicot, the rice sequences were also searched against the Arabidopsis protein set using NCBI-BLASTX (parameters: -p blastx; -e 1e-5; -F ‘mS'). We counted a rice gene as hit if the E-value was less than or equal to 1 × 10−8, which corresponds to a bit score of approximately 51. The rice genes hit were counted and categorized.
The Arabidopsis protein set was compared the sorghum MF dataset using NCBI-TBLASTN (parameters: -p tblastn; -e 1e-5; -F ‘mS'). We counted an Arabidopsis protein as hit if the E-value was less than or equal to 1 × 10−8, which corresponded to a bit score of approximately 57. The Arabidopsis hits were then counted and categorized as shown in Figure 6.
The Arabidopsis protein set was also compared to the rice gene sequence dataset using NCBI-TBLASTN (parameters: -p tblastn; -e 1e-5; -F ‘mS'). We counted an Arabidopsis protein as hit if the E-value was less than or equal to 1 × 10−8, which corresponded to a bit score of at least 57. The Arabidopsis hits were then counted and categorized as shown in Figure 6.
There were 247 Arabidopsis proteins that were annotated by sorghum MF but not rice sequence. These 247 proteins were then compared to the entire rice genome using NCBI-TBLASTN (parameters: -p tblastn; -e 1e-5; -F ‘mS'). From that set of 247 we removed any Arabidopsis proteins if the E-value was less than or equal to 1 × 10−8, which corresponded to a bit score of at least 59. The remaining 188 proteins were then compared to the entire genome from the O. s. indica cultivar of rice [33] using NCBI-TBLASTN (parameters: -p tblastn; -e 1e-5; -F ‘mS'). From that set of 188 we removed any Arabidopsis proteins if the E-value was less than or equal to 1 × 10−8, which corresponded to a bit score of approximately 60. The resulting set contained 127 Arabidopsis proteins that were supported by sorghum MF but not found in the rice genomes.
Methylation analysis
Methylation was assessed using MethylScreen analysis, which is a real-time PCR technique that reports DNA methylation occupancy information for genomic markers through enzymatic interrogation. MethylScreen analysis compares the cycle thresholds (Cts) of gDNA that has been subjected to various treatments and infers 5′ methylated cytosine (5 mC) occupancy through the changes in Ct mediated by the treatments. The Ct of any locus is a function of the number of copies present within the assay tube. MethylScreen analysis relies upon the simple formula that total gene copies = number of methylated copies + number of unmethylated copies in every sample. Typical sample assays utilize four sample subportions, the first portion of a sample is mock-digested, reporting total copies present. A second (and equal) portion is treated with a methylation-sensitive restriction enzyme (MSRE), reporting the number of gene copies that are methylated. The third portion is digested with a methylation-dependent restriction enzyme (MDRE) such as mcrBC, reporting the number of copies that are unmethylated. The fourth reaction is doubly-digested with both the MSRE and MDRE. When working with relatively pure samples, methylated loci have a Ct from MSRE that is the same as the untreated control, and the Ct obtained from the MDRE is greater. Conversely, unmethylated loci have a MDRE Ct that is identical to untreated and a greater Ct in the MSRE reaction.
The gene tb2 targeted a 263 bp region from the 5′ end of the gene for the assay. There are four HhaI restriction sites and more than 25 possible mcrBC half-sites (5′-RC-3′) in this region (Figure 4A). We developed a SYBR green real-time PCR assay using the Dynamo Kit from MJ Research (Boston, Massachusetts, United States). The forward primer used was 5′-
GCCGCCGCCGACGCCAGCTTTCAC-3′, and the reverse primer was 5′-
ATCCCGGGCGCGGTGCATATCTTGCTGTG-3′. The cycling parameters were 95 °C for 3 min, followed by 50 cycles of two-step PCR: 95 °C for 30 s and 70 °C for 30 s. We utilized both a low-temperature (70 °C) and a high-temperature (82 °C) plate read. 2 μg of gDNA was added to a 200-μl reaction cocktail for digestion using the conditions specified by NEB (Beverly, Massachusetts, United States). Half of the sample was digested with 40 U of HhaI overnight, while the other half remained mock-digested. Both “digests” were subsequently split in two, and to each new digestion, cocktails with NEB2, BSA, and 2×GTP were prepared using a final volume of 100 μl. 40 U of mcrBC was added to one of the mock-digested samples and to one of the HhaI-digested samples. All four reactions were incubated overnight at 37 °C. The PCR assays utilized approximately 40 ng from each of the digests. All amplifications were performed in quadruplicate. A standard dilution curve of S. bicolor gDNA in 1× NEB2 was used to ensure linearity of the system. All reactions were verified using melt-curve analysis. Three replicate analyses were performed (digestions and cycling).
Each of the 11 genes was broken into approximately 1.5-kb pieces, which were aligned to create a consensus kafirin assembly (Figure 4B). The consensus kafirin sequence was examined and a 247-bp region was selected. The forward primer was 5′-
CTCCTTGCGCTCC
TTGCTCTTTC-3′, (where
GCGC is a HhaI restriction site) and the reverse primer was 5′-
GCTGCGCTGCGATGGTCTGT-3′. We used the same SYBR green real-time PCR assay with the Dynamo Kit (MJ Research), as mentioned above for the tb2 gene. Cycling parameters were 95 °C for 3 min, followed by 50 cycles of two-step PCR: 95 °C for 30 s and 56 °C for 30 s. We utilized both a low-temperature (70 °C) and a high-temperature (82 °C) plate read. The input of gDNA was cut to 10 ng per reaction. All amplifications were performed in quadruplicate. Three replicate analyses were performed (digestions and cycling). The threshold was set using a template dilution standard control. For the kafirin genes, the average difference in Ct between the mcrBC single and the HhaI + McrBC double-digests is 2.46 cycles (22.08 ± 0.34 HhaI + McrBC - 19.62 ± 0.19 McrBC).
PCR products from the kafirin cycling reactions were cloned using the topoisomerase-assisted method (Invitrogen, Carlsbad, California, United States). Libraries of insert-bearing clones were generated using standard techniques. From each library, 200 lacZ-negative clones were selected for characterization. The clones were sequenced with a single read using the M13 priming site on the pCR2.1 plasmid (Invitrogen). All seven subfamilies were discovered from both the treated and untreated genomic samples (unpublished data), indicating that all 11 genes were amplified and recoverable, even in the mcrBC-digested fractions.
Identification of DREB1 orthologs in the sorghum dataset
The five Arabidopsis DREB genes DREB1A, DREB1B, DREB1C, DREB2A, and DREB2B were used in a TBLASTN search of an assembly the sorghum dataset using WU-BLAST (parameters: E =e-5; matrix=BLOSUM80; topcomboN=1; wordmask=seg+xnu). Matches to the sorghum assembly with an E-value of 1 × 10−8 or less were analyzed with FGENESH (monocot) to select assemblies with a full-length protein. Out of 67 full length proteins identified in this manner, five sorghum proteins were identified as DREB1 genes based on conservation of the AP2 domain and a conserved C-terminal motif, LWSY [31].
Supporting Information
Table S1 Arabidopsis Proteins with Homologs in Sorghum but Not Rice
Shown is a list of 127 Arabidopsis proteins that have matches to the sorghum MF set at a TBLASTN E-value less than or equal to 1 × 10−8, but are not found in the O. s. japonica or O. s. indica genomes at the same cutoff.
(120 KB DOC).
Click here for additional data file.
Accession Numbers
The sorghum MF sequence set is deposited in the Genome Survey Sequence division of GenBank (http://www.ncbi.nlm.nih.gov/). On 6 January 2004, Orion deposited 50,161 of the sequences under accession numbers CL147592–CL197752. The 36,825 Cold Spring Harbor Laboratories MF sequences were previously deposited in GenBank under the accession numbers CC058553–CC059980, BZ329127–BZ342789, BZ342901–BZ352342, BZ365856–BZ368372, BZ369686–BZ370012, BZ421595–BZ424357, BZ625682–BZ629992, and BZ779555–BZ78192. The remaining 481,989 MF sequences from Orion Genomics are deposited in GenBank under accession numbers CL147592–CL197752 and CW020594–CW502582. The Orion UF sequences are deposited in GenBank's Genome Survey Sequence under accession numbers CW512190–CW514008. The University of Oklahoma UF sequences are deposited in the NCBI trace archive under accession numbers TI566112507–TI566128395. GenBank accession numbers for other genes discussed in this paper are sorghum Cs1 (AF206660); Arabidopsis DREB genes DREB1A (Q9M0L0), DREB1B (P93835), DREB1C (Q9SYS6), DREB2A (O82132), and DREB2B (O82133). Genbank accession numbers for BAC clones (with GenInfo identifiers) are AC120496.1 (GI:20486389), AF010283.1 (GI:2735839), AF061282.1 (GI:4539654), AF114171.1 (GI:4680196), AF124045.1 (GI:5410347), AF369906.1 (GI:19851516), AF466199.1 (GI:18390096), AF466200.1 (GI:18481699), AF466201.1 (GI:18483227), AF466204.1 (GI:18568251), AF503433.1 (GI:21326110), AF527807.1 (GI:22208458), AF527808.1 (GI:22208471), and AF527809.1 (GI:22208503). The GenBank accession number for the protein CAPRICE is NP_182164. Accession numbers for genes used in phylogenetic analysis of sorghum DREB are as follows. Rice genes are OsDREB1A (AF300970), OsDREB1B (AF300972), OsDREB1C (AP001168, nucleotides 142337–142981), and OsDREB1D (AB023482, nucleotides1489–2250); AP2 domains from other Arabidopsis proteins are also included: APETALA2 (R2 domain, accession number P47927), AtERF-1 (BAA32418), LEAFY PETIOLE (AAF32292), and TINY (Q39127).
We would like to thank our collaborators J. McLaren, PhD from Solvigen and J. Osborne, PhD and M. Lenz, PhD from NC+ Hybrids. The sorghum MF sequence set was generated as part of a Department of Energy matching funds grant (DE-fc36–01ID14212) to Orion Genomics and United States Department of Agriculture IFAFS grant 2001–52100-11331 to RAM and WRM. The sorghum whole-genome shotgun sequences were generated as part of a National Science Foundation Experimental Program to Stimulate Competitive Research grant to BAR. Finally, we would like to thank the Free Software/Open Source community for creating wonderful software and for fielding our questions, especially OpenOffice.org, ImageMagick, BioPerl, and the multitude of Linux groups.
Competing interests. Orion Genomics markets the GeneThresher methylation filtering technology.
Author contributions. JAB, MAB, AN, RWC, IFK, PDR, NL, WRM, JAJ, and RAM conceived and designed the experiments. JAB, MAB, AN, RWC, JJ, EF, TR, JF, KB, JM, MS, HH, PDR, and JAJ performed the experiments. JAB, AN, RWC, DR, HH, IFK, PDR, JAJ, and RAM analyzed the data. JAB, BAR, GW, IFK, and PDR contributed reagents/materials/analysis tools. JAB, AN, RWC, IFK, PDR, JAJ, and RAM wrote the paper.
Citation: Bedell JA, Budiman MA, Nunberg A, Citek RW, Robbins D, et al. (2005) Sorghum genome sequencing by methylation filtration. PLoS Biol 3(1): e13.
Abbreviations
5 mC5′ methylated cytosine
BACbacterial artificial chromosome
bpbasepairs
CDScoding sequence
Ctcycle threshold
DREBdehydration-responsive element binding protein
ESTexpressed sequence tag
E-valueBLAST Expect value
FPfilter power
gDNAgenomic DNA
HChigh C0
t
kbkilobase
Mbmegabase
mcrBCmodified cytosine restriction
MDREmethylation-dependent restriction enzyme
MFmethylation filtration/filtering/filtered
miRNAmicroRNA
MSREmethylation-sensitive restriction enzyme
TNRtrinucleotide repeat
UFunfiltered
==== Refs
References
Board on Science and Technology for International Development, National Research Council Lost crops of Africa 1996 Washington, DC National Academy Press 386
Bennett MD Leitch IJ Plant DNA C-values database (release 2.0, January 2003) 2003 http://www.rbgkew.org.uk/cval/homepage.html , accessed in September 2004
Bennetzen JL The evolution of grass genome organisation and function Symp Soc Exp Biol 1998 51 123 126 10645434
Bennetzen JL Schrick K Springer PS Brown WE SanMiguel P Active maize genes are unmodified and flanked by diverse classes of modified, highly repetitive DNA Genome 1994 37 565 576 7958822
SanMiguel P Gaut BS Tikhonov A Nakajima Y Bennetzen JL The paleontology of intergene retrotransposons of maize Nat Genet 1998 20 43 45 9731528
SanMiguel P Tikhonov A Jin YK Motchoulskaia N Zakharov D Nested retrotransposons in the intergenic regions of the maize genome Science 1996 274 765 768 8864112
Lander ES Linton LM Birren B Nusbaum C Zody MC Initial sequencing and analysis of the human genome Nature 2001 409 860 921 11237011
Venter JC Adams MD Myers EW Li PW Mural RJ The sequence of the human genome Science 2001 291 1304 1351 11181995
Meyers BC Tingey SV Morgante M Abundance, distribution, and transcriptional activity of repetitive elements in the maize genome Genome Res 2001 11 1660 1676 11591643
Peterson DG Schulze SR Sciara EB Lee SA Bowers JE Integration of Cot analysis, DNA cloning, and high-throughput sequencing facilitates genome characterization and gene discovery Genome Res 2002 12 795 807 11997346
Pevzner PA Tang H Waterman MS An Eulerian path approach to DNA fragment assembly Proc Natl Acad Sci U S A 2001 98 9748 9753 11504945
Martienssen R Transposons, DNA methylation and gene control Trends Genet 1998 14 263 264 9676527
Moore G Abbo S Cheung W Foote T Gale M Key features of cereal genome organization as revealed by the use of cytosine methylation-sensitive restriction endonucleases Genomics 1993 15 472 482 8468041
White SE Habera LF Wessler SR Retrotransposons in the flanking regions of normal plant genes: A role for copia-like elements in the evolution of gene structure and expression Proc Natl Acad Sci U S A 1994 91 11792 11796 7991537
Colot V Rossignol JL Eukaryotic DNA methylation as an evolutionary device Bioessays 1999 21 402 411 10376011
Britten RJ Davidson EH DNA sequence arrangement and preliminary evidence on its evolution Fed Proc 1976 35 2151 2157 780132
Britten RJ Kohne DE Repeated sequences in DNA. Hundreds of thousands of copies of DNA sequences have been incorporated into the genomes of higher organisms Science 1968 161 529 540 4874239
Whitelaw CA Barbazuk WB Pertea G Chan AP Cheung F Enrichment of gene-coding sequences in maize by genome filtration Science 2003 302 2118 2120 14684821
Palmer LE Rabinowicz PD O'Shaughnessy AL Balija VS Nascimento LU Maize genome sequencing by methylation filtration Science 2003 302 2115 2117 14684820
Rabinowicz PD Schutz K Dedhia N Yordan C Parnell LD Differential methylation of genes and retrotransposons facilitates shotgun sequencing of the maize genome Nat Genet 1999 23 305 308 10545948
Lander ES Waterman MS Genomic mapping by fingerprinting random clones: A mathematical analysis Genomics 1988 2 231 239 3294162
Martienssen RA Maintenance of heterochromatin by RNA interference of tandem repeats Nat Genet 2003 35 213 214 14593407
Chopra S Brendel V Zhang J Axtell JD Peterson T Molecular characterization of a mutable pigmentation phenotype and isolation of the first active transposable element from Sorghum bicolor
Proc Natl Acad Sci U S A 1999 96 15330 15335 10611384
Montero LM Filipski J Gil P Capel J Martinez-Zapater JM The distribution of 5-methylcytosine in the nuclear genome of plants Nucleic Acids Res 1992 20 3207 3210 1620618
Kovarik A Matyasek R Leitch A Gazdova B Fulnecek J Variability in CpNpG methylation in higher plant genomes Gene 1997 204 25 33 9434162
Gruenbaum Y Naveh-Many T Cedar H Razin A Sequence specificity of methylation in higher plant DNA Nature 1981 292 860 862 6267477
Jeddeloh JA Richards EJ mCCG methylation in angiosperms Plant J 1996 9 579 586 8653110
Holliday R Grigg GW DNA methylation and mutation Mutat Res 1993 285 61 67 7678134
Arabidopsis Genome Initiative Analysis of the genome sequence of the flowering plant Arabidopsis thaliana
Nature 2000 408 796 815 11130711
Shinozaki K Yamaguchi-Shinozaki K Seki M Regulatory network of gene expression in the drought and cold stress responses Curr Opin Plant Biol 2003 6 410 417 12972040
Dubouzet JG Sakuma Y Ito Y Kasuga M Dubouzet EG OsDREB genes in rice, Oryza sativa L., encode transcription activators that function in drought-, high-salt- and cold-responsive gene expression Plant J 2003 33 751 763 12609047
Goff SA Ricke D Lan TH Presting G Wang R A draft sequence of the rice genome (Oryza sativa L. ssp. japonica ) Science 2002 296 92 100 11935018
Yu J Hu S Wang J Wong GK Li S A draft sequence of the rice genome (Oryza sativa L. ssp. indica ) Science 2002 296 79 92 11935017
Wada T Kurata T Tominaga R Koshino-Kimura Y Tachibana T Role of a positive regulator of root hair development, CAPRICE, in Arabidopsis root epidermal cell differentiation Development 2002 129 5409 5419 12403712
Schellmann S Schnittger A Kirik V Wada T Okada K TRIPTYCHON and CAPRICE mediate lateral inhibition during trichome and root hair patterning in Arabidopsis
EMBO J 2002 21 5036 5046 12356720
Rhoades MW Reinhart BJ Lim LP Burge CB Bartel B Prediction of plant microRNA targets Cell 2002 110 513 520 12202040
Reinhart BJ Weinstein EG Rhoades MW Bartel B Bartel DP MicroRNAs in plants Genes Dev 2002 16 1616 1626 12101121
Floyd SK Bowman JL Gene regulation: Ancient microRNA target sequences in plants Nature 2004 428 485 486 15057819
Jones-Rhoades MW Bartel DP Computational identification of plant microRNAs and their targets, including a stress-induced miRNA Mol Cell 2004 14 787 799 15200956
Klein PE Klein RR Vrebalov J Mullet JE Sequence-based alignment of sorghum chromosome 3 and rice chromosome 1 reveals extensive conservation of gene order and one major chromosomal rearrangement Plant J 2003 34 605 621 12787243
Cordeiro GM Casu R McIntyre CL Manners JM Henry RJ Microsatellite markers from sugarcane (Saccharum spp.) ESTs cross transferable to erianthus and sorghum Plant Sci 2001 160 1115 1123 11337068
Morgante M Hanafey M Powell W Microsatellites are preferentially associated with nonrepetitive DNA in plant genomes Nat Genet 2002 30 194 200 11799393
McCouch SR Teytelman L Xu Y Lobos KB Clare K Development and mapping of 2240 new SSR markers for rice (Oryza sativa L) DNA Res 2002 9 199 207 12597276
Bennetzen JL Ma J The genetic colinearity of rice and other cereals on the basis of genomic sequence analysis Curr Opin Plant Biol 2003 6 128 133 12667868
Paterson AH Bowers JE Chapman BA Ancient polyploidization predating divergence of the cereals, and its consequences for comparative genomics Proc Natl Acad Sci U S A 2004 101 9903 9908 15161969
Sanchez AC Subudhi PK Rosenow DT Nguyen HT Mapping QTLs associated with drought resistance in sorghum (Sorghum bicolor L. Moench ) Plant Mol Biol 2002 48 713 726 11999845
Vettore AL da Silva FR Kemper EL Souza GM da Silva AM Analysis and functional annotation of an expressed sequence tag collection for tropical crop sugarcane Genome Res 2003 13 2725 2735 14613979
Martienssen RA Rabinowicz PD O'Shaughnessy A McCombie WR Sequencing the maize genome Curr Opin Plant Biol 2004 7 102 107 15003207
Fu H Zheng Z Dooner HK Recombination rates between adjacent genic and retrotransposon regions in maize vary by 2 orders of magnitude Proc Natl Acad Sci U S A 2002 99 1082 1087 11792865
Fang Z Cone K Sanchez-Villeda H Polacco M McMullen M iMap: A database-driven utility to integrate and access the genetic and physical maps of maize Bioinformatics 2003 19 2105 2111 14594716
Cone KC McMullen MD Bi IV Davis GL Yim YS Genetic, physical, and informatics resources for maize. On the road to an integrated map Plant Physiol 2002 130 1598 1605 12481043
Peterson DG Boehm KS Stack SM Isolation of milligram quantities of nuclear DNA from tomato (Lycopersicon esculentum ), a plant containing high levels of polyphenolic compounds Plant Mol Biol Rep 1997 15 148 153
Smit A Green P RepeatMasker 2002 version 20020713 [computer program]
Bedell JA Korf I Gish W MaskerAid: A performance enhancement to RepeatMasker Bioinformatics 2000 16 1040 1041 11159316
Gish W BLASTN 1996–2004 version 2.0MP-WashU (03-Mar-2004) [computer program]
Gish W States DJ Identification of protein coding regions by database similarity search Nat Genet 1993 3 266 272 8485583
Korf I Serial BLAST searching Bioinformatics 2003 19 1492 1496 12912829
Burge C Karlin S Prediction of complete gene structures in human genomic DNA J Mol Biol 1997 268 78 94 9149143
Korf I Gene finding in novel genomes BMC Bioinformatics 2004 5 59 15144565
Korf I Yandell M Bedell J BLAST 2003 Sebastopol, California O'Reilly 357
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
| 15660154 | PMC539327 | CC BY | 2021-01-05 08:21:19 | no | PLoS Biol. 2005 Jan 4; 3(1):e13 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030013 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1566015510.1371/journal.pbio.0030014Research ArticleBotanyEcologyEvolutionPlant SciencePlantsRelaxed Molecular Clock Provides Evidence for Long-Distance Dispersal of Nothofagus (Southern Beech) Journey of the Southern BeechKnapp Michael
1
Stöckler Karen
1
Havell David
1
2
Delsuc Frédéric
1
3
Sebastiani Federico
4
Lockhart Peter J [email protected]
1
1Allan Wilson Centre for Molecular Ecology and Evolution, Institute of Molecular BioSciencesMassey University, Palmerston NorthNew Zealand2Universal College of LearningPalmerston NorthNew Zealand3Institut des Sciences de l'EvolutionUniversité Montpellier IIFrance4Università degli Studi di Firenze, Dipartimento di Biotecnologie Agrarie LaboratoryGenexpress, Polo Scientifico, Sesto FiorentinoItalyMoritz Craig Academic EditorUniversity of California at BerkeleyUnited States of America1 2005 4 1 2005 4 1 2005 3 1 e143 5 2004 10 11 2004 Copyright: © 2005 Knapp 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.
Dispersal or Drift? More to Plant Biodiversity Than Meets the Eye
Nothofagus (southern beech), with an 80-million-year-old fossil record, has become iconic as a plant genus whose ancient Gondwanan relationships reach back into the Cretaceous era. Closely associated with Wegener's theory of “Kontinentaldrift”, Nothofagus has been regarded as the “key genus in plant biogeography”. This paradigm has the New Zealand species as passengers on a Moa's Ark that rafted away from other landmasses following the breakup of Gondwana. An alternative explanation for the current transoceanic distribution of species seems almost inconceivable given that Nothofagus seeds are generally thought to be poorly suited for dispersal across large distances or oceans. Here we test the Moa's Ark hypothesis using relaxed molecular clock methods in the analysis of a 7.2-kb fragment of the chloroplast genome. Our analyses provide the first unequivocal molecular clock evidence that, whilst some Nothofagus transoceanic distributions are consistent with vicariance, trans-Tasman Sea distributions can only be explained by long-distance dispersal. Thus, our analyses support the interpretation of an absence of Lophozonia and Fuscospora pollen types in the New Zealand Cretaceous fossil record as evidence for Tertiary dispersals of Nothofagus to New Zealand. Our findings contradict those from recent cladistic analyses of biogeographic data that have concluded transoceanic Nothofagus distributions can only be explained by vicariance events and subsequent extinction. They indicate that the biogeographic history of Nothofagus is more complex than envisaged under opposing polarised views expressed in the ongoing controversy over the relevance of dispersal and vicariance for explaining plant biodiversity. They provide motivation and justification for developing more complex hypotheses that seek to explain the origins of Southern Hemisphere biota.
A phylogenetic analysis of Nothofagus species provides evidence for their transoceanic dispersal during the Tertiary, and helps resolve the debate about the origins of plant biodiversity in the Southern Hemisphere
==== Body
Introduction
An important principle of evolutionary inference is that explanations for the past require an understanding of mechanisms and processes applicable in the present [1]. It is perhaps this sticking point more than any other that has polarised views over the relative importance of vicariance and dispersal for explaining extant plant biodiversity. In 1915, Alfred Wegener put forward a testable hypothesis and mechanism that could explain the transoceanic distribution of animal and plant species. In the 21st century, with many DNA studies now implicating the importance of long-distance dispersal for explaining plant biodiversity [2,3,4,5], it is disconcerting that there is currently a very poor understanding of the mechanisms of transoceanic dispersal (but see [6,7,8,9,10]). Indeed, the inference that the seeds of extant Nothofagus species are not suited for dispersal across large distances has played a major role in motivating the hypothesis that transoceanic distributions of Nothofagus (Figure 1) can only be explained by vicariance [11,12,13,14,15]. This hypothesis posits that following the Cretaceous breakup of Gondwana, Nothofagus rafted and evolved in situ upon different Southern Hemisphere lands. Whilst very attractive, this hypothesis fits somewhat uncomfortably with the findings from analyses of morphological and molecular data. In particular, whilst earlier molecular data have been insufficient for rigorous molecular clock analyses, their interpretation has favoured hypotheses of transoceanic dispersal [16,17,18].
Figure 1 Southern Hemisphere Maps and Present-Day Nothofagus Distribution
(A) Transoceanic distribution of Nothofagus subspecies Lophozonia and Fuscospora and South American species N. nitida (subgenus Nothofagus). Map adapted from Swenson et al. [43]. ASE, Australia; NCA, New Caledonia; NGU, New Guinea; NZE, New Zealand; SAM, South America; TAS, Tasmania.
(B) Relationship of Australia, New Zealand, and South America 65 Myr and 35 Myr before present, reconstructed from http://www.odsn.de/ (link “Plate Tectonic Reconstructions”).
Based on the sequence of Gondwana breakup, a hypothesis of vicariance most parsimoniously predicts that Australian Nothofagus species should be most closely related to South American species. This follows since South America and Australia were connected via Antarctica until approximately 35 million years (Myr) ago (Figure 1). In contrast, New Zealand is thought to have separated from Australia 80 Myr ago [19,20]. Thus to explain the close relationship between Australian and New Zealand species by vicariance, it is necessary to argue that extinction of Australian and/or closely related South American species has occurred [12]. Whilst this explanation is ad hoc, the fossil record does provide evidence for numerous Nothofagus extinctions in Australia, South America, and New Zealand [21,22,23].
However, the fossil record has also been interpreted as indicating multiple events of transoceanic dispersal of Nothofagus from Australia to New Zealand. Whilst the extinct “ancestral” Nothofagus pollen type occurred in New Zealand prior to the breakup of Gondwana, Fuscospora pollen first appeared in New Zealand during the Palaeocene (65 Myr ago) and Lophozonia pollen first appeared during the late Eocene (50 Myr ago; [24]). Sixty-five Myr ago the Tasman Sea had already reached its present-day size [19,20]. Hence it is possible that extant New Zealand Nothofagus subgenera did not have the opportunity to reach New Zealand via overland migration. Hill [25] has also described the species Nothofagus cethanica, which first appeared in Oligocene macrofossils from Tasmania. This species shares unique features with extant N. fusca and N. truncata from New Zealand and may share a sister relationship with these species explained by trans-Tasman Sea dispersal [26].
A contribution to the debate over the relative importance of vicariance and dispersal can be made by estimating the divergence times of extant species. However, DNA sequences of insufficient length have prevented robust molecular clock analyses from being undertaken. For this reason, we report the sequencing of a 7.2-kb chloroplast genome fragment covering the gene regions (trnL–trnF and atpB–psaI; see Table 1 for accession numbers) for 11 species of three Nothofagus subgenera (Lophozonia, Fuscospora, and Nothofagus). Our aim has been to date divergence of extant species in the subgenera Lophozonia and Fuscospora. We have carried out relaxed molecular clock analyses using the methods of Sanderson [27,28] and Thorne et al. [29]. Our findings are that, whilst vicariance is likely to explain some transoceanic relationships amongst Nothofagus species, phylogenetic relationships between trans-Tasman species in both Lophozonia and Fuscospora can only be explained by mid- to late-Tertiary transoceanic dispersal.
Table 1 Origin of Nothofagus Samples and Sequence Accession Numbers
Results
Figure 2 shows an optimal maximum-likelihood reconstruction of phylogenetic relationships for chloroplast DNA sequences (7.2-kb comprising the atpB–psaI region and the trnL–trnF region; 7,269 nucleotide sites) for Nothofagus (subgenera or pollen groups (a) Lophozonia, (b) Fuscospora, and (c) Nothofagus) and outgroup Castanea sativa (not shown). In a sensitivity analysis of 60 substitution models, the tree shown in Figure 2 was always recovered with very little difference in branch lengths regardless of the substitution model used. Of all substitution models evaluated, K81uf+G was identified as the best fitting one for the data based on hierarchical likelihood ratio tests and the Akaike Information Criterion. This substitution model and also the F84+ Γ8 model were used for further analyses. The latter was included because the Bayesian relaxed molecular clock (BRMC) approach as implemented in the program MULTIDIVTIME (see Materials and Methods) only allows the use of the JC and the F84 models. Thus analysis with the F84+ Γ8 model allowed a comparison of date estimates to be obtained using different relaxed molecular clock methods. All nodes of the optimal ML tree recovered in the sensitivity analysis received nonparametric bootstrap support greater than 97%, with the only exception being the grouping of N. cunninghamii with N. moorei, which received 72% support.
Figure 2 ML Tree Indicating Evolutionary Relationships for Nothofagus Species Based on the atpB–psaI and trnL–trnF Region of the Chloroplast Genome (7,269 bp)
Divergence dates (in Myr) were obtained with an F84+ Γ8 substitution model using the BRMC approach of Thorne et al. [29]. For the dates indicated, the age of the root node and that of F/N1 were constrained to 70–80 Myr; L2 was also constrained in accordance with fossil data [26] at 20 Myr. Violet numbers show bootstrap values. The pollen grains represent the first appearance of the respective pollen type in the New Zealand fossil record. Plio, Pliocene; Oligo, Oligocene; Palaeo, Palaeocene; Ma, Maastrichian; Campan, Campanian. L1–L4, Lophozonia 1–4; F1–F2, Fuscospora 1–2; F/N1, Fuscospora/Nothofagus 1.
Divergence times for the nodes in this tree (Figure 2) were estimated using the penalized likelihood (PL) method [27] and BRMC method [29,30,31]. For these analyses, a period of 70–80 Myr was used to calibrate the divergence between the three fossil pollen groups representing subgenera Lophozonia, Nothofagus, and Fuscospora. These four pollen groups all first appeared in the fossil record approximately 75 Myr ago [32]. A second constraint of a minimum of 20 Myr for the divergence of N. cunninghamii and N. moorei was also used. This constraint was based on observations reported by Hill [26] that 20-Myr-old fossils intermediate between N. moorei and N. cunninghamii were recorded from Tasmania and that fossils closely resembling N. moorei were also present at that time. The inferred ages for the remaining nodes of the tree, obtained under the F84+ Γ8 model of substitution are given in Table 2 and graphically illustrated on Figure 2. The variance on these estimates was low and the values were little influenced by the choice of substitution model (Table 3). The robustness of the estimates to calibration error was tested by constraining the divergence of Australian and New Zealand sister taxa to 65 Myr (the time before present when the Tasman Sea reached its present position; thus this date provided us with a lower bound for divergence times of trans-Tasman Nothofagus disjunctions that might be explained by vicariance). Constraining these two nodes in this way produced unrealistic age estimates for all basal nodes. For example, using the BRMC method, which additionally required a prior expectation to be specified for the age of the root node (which we specified at 75 Myr—the time of appearance of all four extant pollen types), we estimated a more likely age for the root node at 191 Myr. For the PL approach, which does not require specification of a prior, we estimated the age of the root node at 634 Myr. Other basal nodes in both the Fuscospora and Lophozonia lineages were also much older than reasonably expected (see Table 2).
Table 2 Estimated Divergence Dates and Standard Deviations (in Brackets) of Different Nothofagus Clades
The numbers in bold are all the nodes that were estimated without constraints
Dates are based on different calibration dates and estimation approaches and are given in Myr before present
a Node fixed
b Node constrained
DOI: 10.1371/journal.pbio.0030014.t002
Table 3 Variation of Estimated Divergence Times (in Myr) under 60 Symmetrical Models of DNA Substitution
Dates estimated using PL approach
a Node constrained
Discussion
Our findings from molecular clock analyses using five independent calibrations (for four nodes), suggest that the sister relationships of the Australasian (Australia and New Zealand) species within both Lophozonia and Fuscospora lineages are too young to be explained by continental drift (as indicated by the inferred ages of nodes F1 and L3). Transoceanic dispersal appears the most likely explanation for the trans-Tasman sister relationships indicated in Figures 1 and 2. In contrast, the age inferred for node F2, using both relaxed clock methods is compatible with a hypothesis of continental drift as an explanation for the sister relationship between South American and Australasian Fuscospora lineages. The age for node L4, which separates Australasian and South American Lophozonia, may also be consistent with vicariance. The BRMC method dates it at 34 Myr before present. However, the PL method estimates this node to be only 25 Myr old, an age too recent to be consistent with vicariance. Thus we regard our results for node L4 as equivocal. Nevertheless, southern beeches are likely to have been present in Antarctica 25 myr ago [33], and thus long-distance dispersal across the young southern ocean between South America and Australia via Antarctica may be conceivable.
The robustness of our phylogenetic inferences has been investigated by varying the substitution model (60 symmetric models were used), estimating the variance of age estimates, and evaluating the influence of calibrations on divergence times. With the exception of the root node, the PL method consistently gave more recent age estimates than did the BRMC method. Both methods showed sensitivity to the number of calibration points used, a finding consistent with recent observations on the performance of relaxed molecular clock methods [34]. In general, the date estimates produced by the BRMC approach were more consistent with the fossil record [26]. A relevant question is whether or not additional calibration points could make date estimates older and thus change our conclusion of trans-Tasman dispersal. We suggest that this may be unlikely, given the observation that constraining a minimum age for trans-Tasman sister species to 65 Myr leads to greatly inflated and unrealistic age estimates for all basal nodes. Hence to explain this finding we would need to invoke a further hypothesis of a dramatic and independent slowing in the rate of evolution in Lophozonia, Fuscospora, and Nothofagus lineages.
Thus the hypothesis that present-day distribution patterns of Nothofagus can be explained by continental drift following the breakup of Gondwana and subsequent extinction of some species [24] can be rejected on the basis of the divergence dates that we have estimated. These dates also indicate that present-day Nothofagus species in New Zealand are not the direct descendants of the Fuscospora and Lophozonia southern beeches that reached New Zealand in the Palaeocene and Eocene eras, respectively [24]. This finding highlights the caution that needs to be taken when interpreting fossil evidence for the apparent first appearance of extant taxa. Fossils that identify specific evolutionary lineages may not necessarily indicate the origins for extant taxa or suggest a continuous presence for these taxa. Similar concerns follow from the findings of molecular analyses for Ascarina and Laurelia in New Zealand [2,4].
The strength of our molecular analyses highlights the importance of future research into potential mechanisms of long-distance dispersal, and in particular reinvestigation of the transoceanic dispersal properties of Nothofagus seeds. For the reasons that we outline in our introduction, it seems likely that only once the mechanisms of long-distance dispersal are understood will hypotheses based on DNA divergence time estimates be truly convincing. DNA sequence analyses have also suggested that long-distance dispersal and continental drift are both important for explaining distributions of the conifer Agathis (Araucariaceae) in the South Pacific [35]. Although the molecular evidence for Agathis is not as strong as it is for Nothofagus, the findings from the molecular studies on these genera highlight the importance of considering more complex hypotheses of relationship in debates concerning the relative importance of dispersal and vicariance.
Materials and Methods
Sequence data
Chloroplast DNA sequences (7.2 kb comprising the atpB–psaI region and the trnL–trnF region) were determined for each of 11 accessions of Nothofagus (subgenera or pollen groups Lophozonia, Fuscospora, and Nothofagus) sampled in South America, Australia, and New Zealand (see Table 1). These genome regions were also determined for C. sativa (an outgroup taxon from Fagaceae) and aligned using progressive multiple-sequence alignment: ClustalX version 1.81 [36]. This resulted in an unambiguous alignment of 7,269 nucleotide sites. Data are missing for approximately 250 bp of the atpB gene and atpB–rbcL intergene region of Nothofagus.
Tree building
Phylogenetic analyses were conducted with PAUP* version 4.0b10 [37] under the ML criterion. A model sensitivity test was conducted, investigating a range of 60 symmetrical models of DNA substitution corresponding to the 56 implemented in MODELTEST version 3.06 [38] (http://darwin.uvigo.es/software/modeltest.html) plus F84, F84+I, F84+Γ8, and F84+I+Γ8
. ML parameters of these models were estimated by PAUP* following the approach used in MODELTEST. These parameters were then used to conduct 60 individual ML heuristic searches in PAUP* with tree bisection-reconnection branch swapping and a neighbour-joining starting tree. ML bootstrap proportions were obtained after 100 replications, using the same search strategy and ML parameters as for the analysis of the original dataset.
Molecular dating: The PL method
Divergence dates were obtained using the PL method of Sanderson [27] as implemented in the program r8s, version 1.60 [28] (http://ginger.ucdavis.edu/r8s/) with the TN algorithm. The outgroup was excluded using the “prune” command. The degree of autocorrelation within lineages was estimated using cross-validation as suggested by Sanderson [27], and the correcting smoothing parameter λ defined accordingly. Divergence dates were estimated on the 60 ML phylograms recovered in the phylogenetic model sensitivity analysis. Ages for each node across the 60-ML trees were summarized using the “profile” command. Confidence limits on dating estimates were computed by using nonparametric bootstrapping of the original dataset as suggested by Sanderson and Doyle [39]. The program SEQBOOT of the PHYLIP 3.6 package [40] was used to generate 100 bootstrap resampled datasets of 7,269 sites in length. ML branch lengths of the optimal topology were then estimated under the F84+ Γ8 model for each of the bootstrap resampled datasets using PAUP*. Divergence estimates were then calculated for each of the 100 bootstrap replicates using r8s to obtain standard deviations on each node by the “profile” command and the settings described above.
Molecular dating: The BRMC method
The BRMC approach was applied using the program MULTIDIVTIME as implemented in the Thornian Time Traveller (T3) package [41]. First, the program BASEML of the PAML package version 3.13 [42] (http://abacus.gene.ucl.ac.uk/software/paml.html) was used to estimate the ML parameters of the F84+ Γ8 substitution model, using the ML topology previously identified. Second, the program ESTBNEW (ftp://abacus.gene.ucl.ac.uk/pub/T3/) was used to estimate branch lengths of the ML topology and the corresponding variance–covariance matrix. Finally, the program MULTIDIVTIME was used to run a Markov chain Monte Carlo for estimating mean posterior divergence times on nodes with associated standard deviations from the variance–covariance matrix produced by ESTBNEW. The Markov chain was sampled 10,000 times every 100 cycles after a burn-in stage of 100,000 cycles. We used a 75-Myr (SD = 37.5 Myr) prior [32] for the expected number of time units between tip and root and a prior of 200 Myr for the highest possible number of time units between tip and root. Other priors for gamma distribution of the rate at root node and the Brownian motion constant describing the rate variation (i.e., the degree of rate autocorrelation along the descending branches of the tree) were derived from the median branch length. As for the PL method, the outgroup was not included in this analysis.
Supporting Information
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/) accession numbers for the sequences discussed in this paper are given in Table 1.
We thank Bruce Christie, Peter Seemann, Andrew Rozefelds, and Martin Roberts for help with obtaining tissue samples, and also Trish McLenachan, Tim White, and Michael Heads for valuable discussions. Leon Perrie, Heidi Meudt, Robert Hill, Matt McGlone, Peter Raven and an anonymous reviewer provided helpful and critical comments on our manuscript that were much appreciated. We would like to thank the New Zealand Marsden Fund and the Alexander von Humboldt Foundation for funding our study. MK is supported by doctoral scholarships from Massey University and the German Academic Exchange Service.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MK, KS, DH, FD, and PJL conceived and designed the experiments. MK and KS performed the experiments. MK, KS, FD, and PJL analyzed the data. DH and FS contributed reagents/materials/analysis tools. MK, DH, and PJL wrote the paper.
Citation: Knapp M, Stöckler K, Havell D, Delsuc F, Sebastiani F, et al. (2005) Relaxed molecular clock provides evidence for long-distance dispersal of Nothofagus (southern beech). PLoS Biol 3(1): e14.
Abbreviations
BRMCBayesian relaxed molecular clock
MLmaximum likelihood
Myrmillion years
PLpenalized likelihood
==== Refs
References
Lyell C Principles of geology: Being an attempt to explain the former changes of the earth's surface by reference to causes now in operation 1970 3 v New York S-H Service Agency
Renner SS Foreman DB Murray D Timing transantarctic disjunctions in the Atherospermataceae (Laurales): Evidence from coding and noncoding chloroplast sequences Syst Biol 2000 49 579 591 12116428
Winkworth RC Wagstaff SJ Glenny D Lockhart PJ Plant dispersal N.E.W.S from New Zealand Trends Ecol Evol 2002 17 514 520
Zhang L-B Renner S The deepest splits in Chloranthaceae as resolved by chloroplast sequences Int J Plant Sci 2003 164 383 392
Givnish TJ Millam KC Evans TM Hall JC Pires JC Ancient vicariance or recent long-distance dispersal? Inferences about phylogeny and South American–African disjunctions in Rapateaceae and Bromeliaceae based on ndhF sequence data Int J Plant Sci 2004 165 S35 S54
Close RC Moar NT Tomlinson AI Lowe AD Aerial dispersal of biological material from Australia to New Zealand Int J Biometeorol 1978 22 1 19
Carlquist SJ Keast A Miller SE Plant dispersal and the origin of Pacific island floras The origin and evolution of Pacific island biotas, New Guinea to Eastern Polynesia: Patterns and processes 1996 Amsterdam Academic 153 164
Hurr K Lockhart PJ Heenan PB Penny D Dispersal of the Edwardsia section of Sophora (Leguminosae) around the southern oceans: Molecular evidence J Biogeogr 1999 26 565 577
Muñoz J Felicísimo ÁM Cabezas F Burgaz AR Martínez I Wind as a long-distance dispersal vehicle in the Southern Hemisphere Science 2004 304 1144 1147 15155945
Renner S Plant dispersal across the tropical Atlantic by wind and sea currents Int J Plant Sci 2004 165 S23 S33
Preest DS Gressit JL A note on the dispersal characteristics of the seed of the New Zealand podocarps and beeches, and their biogeographical significance Pacific Basin biogeography, a symposium 1963 Honolulu Bishop Museum Press 415 424
Linder HP Crisp MD
Nothofagus and Pacific biogeography Cladistics 1995 11 5 32
Craw RC Gehan JR Heads MJ Panbiogeography: Tracking the history of life 1999 Oxford Oxford University Press
McCarthy D The trans-Pacific zipper effect: Disjunct sister taxa and matching geological outlines that link the Pacific margins J Biogeogr 2003 30 1545 1561
Sanmartín I Ronquist F Southern Hemisphere biogeography inferred by event-based models: Plant versus animal patterns Syst Biol 2004 53 216 243 15205050
Martin PG Dowd JM Using sequences of rbcL to study phylogeny and biogeography of Nothofagus species Aust J Bot 1993 6 441 447
Hill RS Jordan GJ The evolutionary history of Nothofagus (Nothofagaceae) Aust J Bot 1993 6 111 126
Manos P Systematics of Nothofagus (Nothofagaceae) based on rDNA spacer sequences (ITS): Taxonomic congruence with morphology and plastid sequences Am J Bot 1997 84 1137 1150 21708668
Wilford GE Brown PJ Hill RS Maps of late Mesozoic–Cenozoic Gondwana breakup: Some palaeogeographical implications History of the Australian vegetation: Cretaceous to recent 1994 Cambridge Cambridge University Press 5 13
McLoughlin S The breakup history of Gondwana and its impact on pre Cenozoic floristic provincialism Aust J Bot 2001 49 271 300
Hill RS Jordan GJ Macphail MK Veblen TT Hill RS Read J History and paleoecology of Australian Nothofagus forests The ecology and biogeography of Nothofagus forests 1996 New Haven (Connecticut) Yale University Press 182 199
Markgraf V Romero E Villagrán C Veblen TT Hill RS Read J History and palaeoecology of South American Nothofagus forests The ecology and biogeography of Nothofagus forests 1996 New Haven (Connecticut) Yale University Press 354 386
McGlone MS Mildenhall DC Pole MS Veblen TT Hill RS Read J History and palaeoecology of New Zealand Nothofagus forests The ecology and biogeography of Nothofagus forests 1996 New Haven (Connecticut) Yale University Press 83 130
Hill RS Biogeography, evolution and palaeoecology of Nothofagus (Nothofagaceae): The contribution of the fossil record Aust J Bot 2001 49 321 332
Hill RS Tertiary Nothofagus macrofossils from Cethana, Tasmania Alcheringia 1984 8 81 86
Hill RS Tertiary Nothofagus (Fagaceae) macrofossils from Tasmania and Antarctica and their bearing on the evolution of the genus Bot J Linn Soc 1991 105 73 112
Sanderson MJ Estimating absolute rates of molecular evolution and divergence times: A penalized likelihood approach Mol Biol Evol 2002 19 101 109 11752195
Sanderson MJ r8s: Inferring absolute rates of molecular evolution and divergence dates in the absence of a molecular clock Bioinformatics 2003 19 301 302 12538260
Thorne JL Kishino H Painter IS Estimating the rate of evolution of the rate of molecular evolution Mol Biol Evol 1998 15 1647 1657 9866200
Kishino H Thorne JL Bruno WJ Performance of a divergence time estimation method under a probabilistic model of rate evolution Mol Biol Evol 2001 18 352 361 11230536
Thorne JL Kishino H Divergence time and evolutionary rate estimation with multilocus data Syst Biol 2002 51 689 702 12396584
Dettmann ME Pocknall D Romero EJ Zamboloa M
Nothofagidites Erdman ex Potonie 1960: A catalogue of species with notes on palaeogeographic distribution of Nothofagus Bl. (southern beech) NZ Geol Surv Palaeontol Bull 1990 60 1 79
Hill RS Truswell EM Kennett JP Warnke DA
Nothofagus fossils in the Sirius Group, Transantarctic Mountains: Leaves and pollen and their climatic implications Antarctic research series, Volume 60. The Antarctic palaeoenvironment: A perspective on global change 1993 Washington, DC American Geophysical Union 67 73
Pérez-Losada M Høeg JT Crandall KA Unraveling the evolutionary radiation of the thoracian barnacles using molecular and morphological evidence: A comparison of several divergence time estimation approaches Syst Biol 2004 53 244 264 15205051
Stöckler K Daniel IL Lockhart PJ New Zealand Kauri (Agathis australis (D.Don) Lindl., Araucariaceae) survives Oligocene drowning Syst Biol 2002 51 827 832 12396595
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG The ClustalX windows interface: Flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res 24: 4876–4882 (Program available: ftp-igbmc.u-strasbg.fr 1997 Accessed 16 November 2004.)
Swofford DL PAUP*. Phylogenetic analysis using parsimony (*and other methods), version 4 [computer program] 2003 Sunderland (Massachusetts) Sinauer
Posada D Crandall K MODELTEST: Testing the model of DNA substitution Bioinformatics 1998 14 817 818 9918953
Sanderson MJ Doyle JA Sources of error and confidence intervals in estimating the age of angiosperms from rbcL and 18S rDNA data Am J Bot 2001 88 1499 1516 21669683
Felsenstein J PHYLIP (PHYLogeny inference package), version 3.6 [computer program]. Available: http://evolution.genetics.washington.edu/phylip.html
2002 Accessed 18 November 2004
Yang Z T3 “Thornian time traveller” [computer program]. Available: http://abacus.gene.ucl.ac.uk/
2003 Accessed 16 November 2004
Yang Z PAML: A program package for phylogenetic analysis by maximum likelihood Comput Appl Biosci 1997 13 555 556 9367129
Swenson U Backlund A McLoughlin S Hill RS
Nothofagus biogeography revisited with special emphasis on the enigmatic distribution of subgenus Brassospora in New Caledonia Cladistics 2001 17 28 47
| 15660155 | PMC539330 | CC BY | 2021-01-05 08:21:18 | no | PLoS Biol. 2005 Jan 4; 3(1):e14 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030014 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1566015610.1371/journal.pbio.0030015Research ArticleGenetics/Genomics/Gene TherapyInfectious DiseasesMicrobiologyEubacteriaMajor Structural Differences and Novel Potential Virulence Mechanisms from the Genomes of Multiple Campylobacter Species Sequencing of Four Campylobacter SpeciesFouts Derrick E [email protected]
1
Mongodin Emmanuel F
1
Mandrell Robert E
2
Miller William G
2
Rasko David A
1
Ravel Jacques
1
Brinkac Lauren M
1
DeBoy Robert T
1
Parker Craig T
2
Daugherty Sean C
1
Dodson Robert J
1
Durkin A. Scott
1
Madupu Ramana
1
Sullivan Steven A
1
Shetty Jyoti U
1
Ayodeji Mobolanle A
1
Shvartsbeyn Alla
1
Schatz Michael C
1
Badger Jonathan H
1
Fraser Claire M
1
Nelson Karen E
1
1The Institute for Genomic Research, RockvilleMarylandUnited States of America2Produce Safety and Microbiology Research Unit, Agricultural Research ServiceUnited States Department of Agriculture, Albany, CaliforniaUnited States of AmericaRelman David A. Academic EditorStanford UniversityUnited States of America1 2005 4 1 2005 4 1 2005 3 1 e152 7 2004 11 11 2004 Copyright: © 2005 Fouts 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.
Multiple Campylobacter Genomes Sequenced
Sequencing and comparative genome analysis of four strains of Campylobacter including C. lari RM2100, C. upsaliensis RM3195, and C. coli RM2228 has revealed major structural differences that are associated with the insertion of phage- and plasmid-like genomic islands, as well as major variations in the lipooligosaccharide complex. Poly G tracts are longer, are greater in number, and show greater variability in C. upsaliensis than in the other species. Many genes involved in host colonization, including racR/S, cadF, cdt, ciaB, and flagellin genes, are conserved across the species, but variations that appear to be species specific are evident for a lipooligosaccharide locus, a capsular (extracellular) polysaccharide locus, and a novel Campylobacter putative licABCD virulence locus. The strains also vary in their metabolic profiles, as well as their resistance profiles to a range of antibiotics. It is evident that the newly identified hypothetical and conserved hypothetical proteins, as well as uncharacterized two-component regulatory systems and membrane proteins, may hold additional significant information on the major differences in virulence among the species, as well as the specificity of the strains for particular hosts.
Although the single genome sequence of the human Campylobacter jejuni yielded some insight, comparison of multiple species of the same genus elucidated greater understanding of virulence mechanisms
==== Body
Introduction
The Gram-negative, spiral-shaped bacterium Campylobacter jejuni is commensal in cattle, swine, and birds [1]. Campylobacter species, however, are the major cause of human bacterial gastroenteritis, and may be responsible for as many as 400–500 million cases worldwide each year [2]. Although the genus Campylobacter is composed of 16 described species [3], human illness is associated primarily with C. jejuni and C. coli and infrequently with C. upsaliensis, C. lari, and C. fetus. Filtration-based isolation techniques have revealed C. upsaliensis to be associated with human disease more than previously known [4]. The majority of C. jejuni infections result in uncomplicated gastroenteritis, but the development of the peripheral neuropathies, Guillain-Barré and Miller-Fisher syndromes is often associated with prior C. jejuni infection [5,6].
All clinically relevant Campylobacter spp. are considered to be thermotolerant in nature. C. jejuni, C. coli, C. lari, and C. upsaliensis also grow readily under microaerophilic conditions (5% oxygen) at 37 °C, and the majority of strains from these species will also grow at 42 °C. The thermotolerant Campylobacter spp. can also be distinguished by their host range. C. jejuni and C. coli are commensal in cattle, swine, and birds [1]; however, C. jejuni is often the predominant species in poultry, and C. coli in swine [4,7]. C. lari is prevalent in birds (seagulls in particular) [8], but has also been isolated from dogs and swine [9,10]. C. upsaliensis has frequently been isolated from domestic dogs and cats [11,12,13,14,15].
The main route of C. jejuni and C. coli human infection is through improperly handled or undercooked poultry, although illnesses caused by the consumption of livestock meat, unpasteurized milk, and contaminated water have also been reported [1]. C. lari has been isolated infrequently from poultry, ox and pork livers [16,17,18], and produce [19], in contrast to frequent isolation at moderate to high levels from fresh water, seawater, and shellfish [20,21]. C. upsaliensis has been isolated infrequently from poultry, ducks, and shellfish, and not from other food sources [4,22,23]. The main reservoir of C. upsaliensis appears to be dogs and cats, with reports of transmission of C. upsaliensis from animal to person [24,25] or person to person [26,27]. Human illness caused by C. lari and C. upsaliensis, unlike C. jejuni and C. coli, may be due to proximity to water and shellfish, and handling of pets, livestock, or livestock carcasses.
The genome sequence of C. jejuni strain NCTC 11168 [28], a human clinical isolate, provided a starting point for studying the proteins involved in outer surface structures and glycosylation [29], and the expression of contingency gene products such as glycosyl transferases and restriction enzymes. However, in contrast to the current understanding of the pathophysiology of other enteric bacteria, that of Campylobacter species remains poorly understood.
The genome of one C. jejuni strain is insufficient to provide a complete picture of the major aspects of Campylobacter biology, including the colonization of reservoir hosts [30], variation in lipooligosaccharide (LOS) and capsule, and potential adaptations of Campylobacter in poultry production and processing environments. In addition, information on the basis of Campylobacter virulence and potential targets for drug and vaccine design is still lacking. Therefore, we sequenced and finished the genome of C. jejuni strain RM1221 (
ATCC BAA-1062), and compared it with the genomes of C. coli strain RM2228 (
ATCC BAA-1061), C. lari strain RM2100 (
ATCC BAA-1060), and C. upsaliensis strain RM3195 (
ATCC BAA-1059) sequenced to at least 8-fold coverage. Strain RM1221 was sequenced because it was isolated from a chicken carcass and minimally passaged [31]. In addition, experimental work with this isolate has identified a number of unique features not present in the previously sequenced C. jejuni strain NCTC 11168, including the colonization of chicken skin and ceca, invasion of Caco-2 cells [31], unique LOS and capsule loci, and other unique open reading frames (ORFs) (unpublished data). C. coli RM2228 was sequenced because it is a multi-drug-resistant chicken isolate. Both C. lari RM2100 (CDC strain D67, “case 6” [32]) and C. upsaliensis RM3195 were selected for sequencing because they are clinical isolates. C. upsaliensis RM3195 was isolated from a patient with Guillain-Barré syndrome, using a filtration-based method of selection [33], and may have been responsible for this disease.
Results/Discussion
Comparative Genome Features
The genome of C. jejuni RM1221 is a single circular chromosome, 1,777,831 bp in length, with an average G+C content of 30.31%. There are a total of 1,884 predicted coding regions in the genome with an average ORF length of 885 bp. Ninety-four percent of the genome represents coding sequence. Putative role assignments could be made for 1,124 of the ORFs (60%) (Table 1; Figure S1). The bacterium was found to belong to multilocus sequence type (MLST) 354 and FlaA short variable region (SVR) 33, which belongs to clonal complex 354, whose members are associated with human disease or chickens/chicken meat (Table 1) [34]. The genome features for the unfinished Campylobacter genomes were based on automated analysis and are presented in Table 1. The average coverage of the unfinished genomes was found to be 8.5-fold for C. coli RM2228, 16.5-fold for C. lari RM2100, and 9.0-fold for C. upsaliensis RM3195 for those contigs used to construct the pseudomolecules. The ambiguity rate (number of consensus-altering ambiguities per basepair) was determined to be between 1:54,000 and 1:93,000 for these unedited, unfinished genomes at 8-fold depth of coverage. The genomic structure of C. jejuni RM1221 is syntenic with the genome of C. jejuni NCTC 11168, and is disrupted by inserted prophages/genomic islands in RM1221 (see below), and ORFs within the capsular (extracellular) polysaccharide (EP) loci in NCTC 11168 (Figures 1A and S2). The C. coli RM2228 genomic structure also has a considerable amount of synteny with C. jejuni RM1221, sharing similar breakpoints, as observed in the C. jejuni comparisons, but displaying evidence of rearrangements about the oriC, as described for other bacterial genomes [35]. In contrast, C. lari and C. upsaliensis possess little if any synteny with C. jejuni RM1221.
Figure 1 Whole-Genome Comparison of Five Campylobacter Strains
Line figures depict the results of PROmer analysis. Colored lines denote percent identity of protein translations and are plotted according to the location in the reference (C. jejuni RM1221, x-axis) and query genomes (C. jejuni NCTC 11168 [upper y-axis] and C. coli RM2228 [lower y-axis]) (A). The Venn diagrams show the number of proteins shared (black) or unique (red) within a particular relationship for all five Campylobacter strains (B) and for members of the sequenced ɛ-Proteobacteria compared in this study (C). Protein sequences binned as “unique” are unique within the context of the genomes plotted and the cutoffs used to parse the BLASTP data. The pie charts plot the number of protein sequences by main functional role categories for C. jejuni RM1221 ORFs. A frequency distribution of protein percent identity (D) was computed: specifically, the number of protein sequences within class intervals of 5% amino acid identity from 35% to 100% that match C. jejuni RM1221 reference sequences were plotted.
Table 1 Genome Features of Five Campylobacter Genomes
a MLST and complex designations follow the PubMLST Web site (http://pubmlst.org/) [101]
b FlaA SVR (http://phoenix.medawar.ox.ac.uk/flaA/) [34,102]
c Estimate (number based on manual inspection of only a subset of genes)
d From [28]
e Disrupted ORF
f Contingency gene present
g Based on TIGR role category
h See Table S10 for gene lists
MOMP, major outer membrane protein
Comparison of C. jejuni RM1221 protein sequences with those of other fully sequenced members of the ɛ-Proteobacteria revealed 540 shared protein sequences, many of which are proposed to have “house-keeping” functions (Figure 1C). Of the 1084 protein sequences shared by all the Campylobacter species in this study, 46 had no match to any other organism in the database (p-value cutoff ≤ 10−5) (Figure 1B). Eleven of these were assigned functions related to cell envelope biosynthesis, or fatty acid and phospholipids metabolism. Further analysis revealed 44 proteins considered C. jejuni-specific, of which 12 mostly hypothetical proteins were truly novel, having no match to other organisms in the database. Of the 300 C. jejuni RM1221-specific protein sequences, only 95 were not in phage or genomic island regions.
To quantify relatedness among the sequenced ɛ-Proteobacteria, the average protein percent identity was computed for all proteins matching the reference strain C. jejuni RM1221 with a p-value less than or equal to 10−5, identity of 35% or more, and match lengths of at least 75% of the length of both query and subject sequence. Not surprisingly, C. jejuni NCTC 11168 had the highest average protein percent identity (1,468 proteins averaging 98.41% identity) with C. jejuni RM1221 proteins. C. coli RM2228 was second, with 1,399 proteins averaging 85.81% identity. Surprisingly, C. upsaliensis RM3195 had the third highest average protein percent identity with C. jejuni RM1221 (1,261 proteins; 74.72% average identity), followed by C. lari RM2100 with 1,251 proteins having 68.91% average identity. This was surprising since a 16S rRNA tree depicts C. upsaliensis to be more dissimilar to C. jejuni, C. coli, and C. lari [3]. Wollinela succinogenes DSMZ1740 was next, with 838 proteins averaging 53.77% identity, followed by Helicobacter hepaticus
ATCC 51449 (770 proteins; 53.66% average identity), H. pylori 26695 (675 proteins; 52.39% average identity), and H. pylori J99 (682 proteins; 52.28% average identity).
Phylogenetic Comparisons
To resolve the apparent discrepancy regarding the relatedness of the ɛ-Proteobacteria between the results of average protein percent identities from this study and the previously published 16S rRNA tree based on percent sequence similarity [3], a consensus boot-strapped maximum-likelihood tree was generated based on trimmed alignments with gaps removed (Figure 2A). One of the advantages of generating whole-genome sequence is the magnitude of information available for resolving differences between closely related organisms. To better resolve the Campylobacter species, we took advantage of the wealth of sequence information to construct a maximum-likelihood concatenated protein tree using a set of 12 conserved protein sequences that have been previously shown to be reliable markers for phylogenetic analysis (Figure 2B) [36,37]. A frequency distribution of protein percent identity was plotted with 5% class intervals to visualize the similarities of these genomes at the protein level (see Figure 1D). The 16S rRNA tree of sequenced members of the ɛ-Proteobacteria suggests that C. jejuni RM1221 is more closely related to C. coli RM2228 than to the other C. jejuni strain, NCTC 11168. However, the concatenated protein tree of these same organisms showed the two C. jejuni strains to be more closely related to each other than either is to C. coli RM2228, agreeing with the distributions of protein percent identities (see Figure 1D). Both trees indicate that W. succinogenes is more closely related to Helicobacter than to Campylobacter. Most likely, the protein tree is more accurate and the rRNA tree is incorrect because the 16S rRNA does not have enough variation to resolve these close relationships [37]. Whole-genome sequencing of more members of the ɛ-Proteobacteria will enable a clearer resolution of the evolutionary relationships within this group of related organisms.
Figure 2 Phylogenetic Analysis and Frequency Distribution of Protein Percent Identity
Concensus maximum-likelihood trees are depicted using multiple alignments of 16S rRNA (A) or 12 concatenated protein datasets (B). The numbers along the branches denote percent occurrence of nodes among 100 bootstrap replicates. The scale bar represents the number of nucleotide (A) or amino acid (B) substitutions.
Phages/Genomic Islands
The major difference between the C. jejuni NCTC 11168 and C. jejuni RM1221 genomes is the presence within the strain RM1221 genome of four large integrated elements (Figures 3 and S3). This characteristic has been observed in whole-genome intra-species comparisons of both Gram-positive and Gram-negative microorganisms [38,39,40,41,42]. The first element, Campylobacter Mu-like phage (CMLP1) (30.5% G+C content), located upstream of argC (CJE0275), encodes several proteins with similarity to bacteriophage Mu and other Mu-like prophage proteins [43], including putative MuA and MuB transposase homologs. Another feature consistent with the identification of CMLP1 as a novel Mu-like prophage is the presence of terminal 5′-TG-3′ dinucleotides flanked by a five-base direct repeat (
TATGC). Preliminary results suggest that this prophage is inducible with mitomycin C and that other C. jejuni strains harbor a related prophage (unpublished data). Genetic manipulation of this phage could yield useful molecular tools analogous to the Mu derivatives for the construction of random gene fusions or mini-Mu elements for in vivo cloning. Although this Mu-like prophage contains no characterized virulence determinants, it could potentially alter pathogenicity or other phenotypes via insertional inactivation.
Figure 3 Linear Representations of Prophage Regions
Regions are (from top to bottom): CMLP1, CJIE2, CJIE4, CLIE1, and CUIE1. Colors of ORFs are indicated in the key by putative phage function. Connecting lines represent those ORFs whose protein sequences match at a BLASTP of 30% identity or better. These lines do not indicate the coordinates of match, merely that there is a match.
In contrast to CMLP1, C. jejuni RM1221 integrated elements 2 and 4 (CJIE2 and CJIE4) have integrated into the 3′ end of arginyl- and methionyl-tRNA genes, respectively. Several ORFs predicted to encode phage-related endonucleases, methylases, or repressors are present within these elements; however, unlike CMLP1, few ORFs encoding phage structural proteins were identified within CJIE4. CJIE4 is similar to a putative prophage contained within the C. lari RM2100 genome (C. lari integrated element 1 [CLIE1]); 66% (35/53) of predicted proteins have BLASTP matches (p-value ≤ 10−5; identity ≥ 30%) (Figure 3). CLIE1 is integrated into a leucinyl-tRNA. The inability to identify matches to major capsid, portal, and scaffold protease proteins within CJIE2 or C. upsaliensis RM3195 integrated element 1 (CUIE1) suggests that they represent either intact prophages with novel head morphogenesis proteins, satellite phages, or nonfunctional prophages or genomic islands.
The absence of any phage-related ORFs within CJIE3 (located within an arginyl-tRNA), suggests that CJIE3 is not a prophage but rather a genomic island or integrated plasmid. Seventy-three percent (45/62) of the CJIE3 predicted proteins are similar to predicted proteins encoded on the C. coli RM2228 megaplasmid (pCC178) (Figure S4; see below), suggesting that CJIE3 was plasmid-derived. However, the observed lack of synteny between CJIE3 and the C. coli RM2228 megaplasmid suggests that CJIE3 was not derived from pCC178 but possibly from a related Campylobacter megaplasmid. Although most of the ORFs contained within CJIE3 encode hypothetical proteins (23% 14/62), many are similar to proteins encoded within the 71-kb H. hepaticus
ATCC 51449 genomic island (HHGI1), suggesting this genomic island could also be plasmid-derived [44]. Furthermore, 33% (23/70) of HHGI1 proteins match pCC178-encoded proteins.
Bacteriophages are vehicles for the lateral or horizontal movement of genes that can increase bacterial fitness [45,46]. Additionally, it has been demonstrated that bacteriophage-carried genes can play a role in many aspects of bacterial virulence (adhesion, invasion, host evasion, and toxin production) [47]. Though only one of the Campylobacter prophages (CMLP1) has been shown to be inducible, we cannot predict whether the other putative prophages or plasmid-like element can be excised. Because the majority of ORFs that lie within prophage regions are hypothetical proteins, we are unable to deduce any putative functions from them; however, we cannot rule out possible functions that either directly impact virulence or increase the fitness of the host in a particular environment.
Plasmids
C. coli RM2228 and C. lari RM2100 each contain a single plasmid (pCC178; approximately 178 kb, and pCL46, approximately 46 kb, respectively), whereas C. upsaliensis RM3195 contains two plasmids (pCU3, approximately 3.1 kb, and pCU110, approximately 110 kb; Tables 1 and S1). In the current study, neither C. jejuni isolate harbors a plasmid; however, a C. jejuni virulence plasmid, pVir from C. jejuni strain 81–176, was previously sequenced and shown to play a role in pathogenesis [48]. The coding regions of pVir are entirely in one orientation except for a single coding region, which is uncharacteristic for a plasmid of this size. The coding regions of pCU110 and pCL46, like pVir, show a similar coding strand bias. In pCC178, the lack of coding region bias may be explained by the presence of antibiotic resistance genes (Tables 2 and S2) flanked by putative mobile genetic elements. Only the 3.1-kb plasmid of C. upsaliensis RM3195 (pCU3) has a defined plasmid replication region. The single-stranded binding (Ssb) proteins are conserved among all of the plasmids, alluding to a common evolutionary origin; however, the nickase proteins on the plasmids are not conserved, suggesting that nickase may be specific to the plasmid or strain.
Table 2 Relevant Drug Resistance Profiles
a Resistance mechanism: CCOA0067/CCOA0068—aminoglycoside 3′-phosphotransferase from pCC178
b Sensitivity likely due to fragmentation of a class D β-lactamase
c Resistance mechanism: 23S rRNA (A2122G), corresponding to position 2,143 of H. pylori sequence [84]
d Resistance mechanism: T86V mutation in gyrA (CLA1521)
e Resistance mechanism: CCOA0206—tetracycline resistance protein (tetO) from pCC178
I, intermediate resistance; R, resistant; S, susceptible
One conserved feature of all of the large Campylobacter plasmids is the presence of a Type IV secretion system (T4SS), possibly involved in conjugative plasmid transfer or secretion of virulence factors [49] (Figure S4). The plasmid-encoded T4SSs in the non–C. jejuni species are most similar to each other based on synteny and amino acid identity; however, they share only synteny with the T4SS encoded by pVir or the Agrobacterium tumefaciens Ti plasmid [50]. The non–C. jejuni plasmid T4SSs may be involved in conjugation rather than secretion of virulence factors because they are more similar to T4SSs known to mobilize DNA than to T4SSs that secrete effectors [50] (Figure S4). Unlike pVir, the other Campylobacter plasmids encode proteins similar to VirB2 of the Ti plasmid, which is responsible for pilus formation [49] (Figure S4) and has recently been shown to be essential for DNA transfer, further hinting at a role in DNA mobility [51]. Additionally, pCU110 appears to contain a number of other proteins that are similar to conjugal transfer proteins of other plasmids, which may function independently or in concert with the T4SS to transfer plasmid DNA to donor cells.
Transposable Elements
Both C. jejuni NCTC 11168 and C. jejuni RM1221 are notable for the apparent absence of intact insertion sequence (IS) elements. With the exception of one copy of a degenerate transposase resembling IS605, located between the tonB gene and a 5S rRNA gene, their genomes are devoid of IS elements. In contrast, C. coli RM2228 contains five copies of an IS element (ISCco1 of the IS605 family) at three positions in the chromosome and at least two positions in the megaplasmid pCC178, hinting at recent acquisition and transposition competence. Both the C. upsaliensis RM3195 and C. lari RM2100 pseudomolecules lack the tonB–5S rRNA locus; however, since these are not closed genomes, we cannot accurately assess the status of the IS605 family in these genomes.
CRISPR Analysis
The chromosomes of all five Campylobacter strains in this study were examined for the presence or absence of clustered regularly interspaced short palindromic repeats (CRISPRs) in intergenic regions. A strain was considered CRISPR-positive if it contained two or more direct repeats of a 21-bp or larger DNA segment separated by unique spacer sequences of a similar size. We identified CRISPR elements in only C. jejuni NCTC 11168 and C. jejuni RM1221. However, a previous study found that CRISPR elements are sometimes detectable in C. coli [52]. Also consistent with the previous study, the two strains of C. jejuni examined here can be differentiated by both the unique sequence of the spacer sequences (Figure S5) and the number of CRISPR repeats in the element (five in C. jejuni NCTC 11168 and four in C. jejuni RM1221). It is noteworthy that the previous study did not include C. lari or C. upsaliensis, which appear not to contain CRISPR elements, unless they are in a different region of the genome from the C. jejuni CRISPRs and are in unsequenced areas. This further demonstrates the limited utility of CRISPRs in genotyping studies of Campylobacter species.
Restriction–Modification Systems
The Type I restriction–modification (RM) loci from 65 C. jejuni strains have been characterized previously [53]. In contrast to the C. jejuni, C. coli, and C. lari strains sequenced in this study, the C. upsaliensis RM3195 genome is predicted to contain at least three Type I RM loci (Table S3). C. upsaliensis RM3195 also contains a putative fourth locus where the hsdR gene is absent. The sequenced genomes of the C. jejuni strains NCTC 11168 and RM1221, C. coli RM2228, and C. lari RM2100 encode few Type II or Type III RM systems. C. upsaliensis RM3195 encodes one putative Type II and two putative Type III restriction enzymes. In addition, C. upsaliensis RM3195 encodes 15 putative adenine- or cytosine-specific DNA methyltransferases. It is noteworthy that the sequenced genome of H. hepaticus
ATCC 51449, like C. jejuni RM1221, C. coli RM2228, and C. lari RM2100, has a paucity of RM loci [44] and would therefore be considered “Campylobacter-like” whereas C. upsaliensis RM3195 would be considered “Helicobacter
Pylori-like” with respect to RM systems. At least four of the C. upsaliensis RM3195 RM systems lie within regions of atypical nucleotide composition, suggesting recent horizontal transfer as selfish mobile elements [54].
Diversity within the Campylobacter RM systems has implications for Campylobacter biology, specifically DNA uptake and phage infection. Campylobacter spp. are naturally competent [55], and horizontal gene transfer via natural transformation is thought to play an important role in the evolution of C. jejuni [56]. Natural competence, as well as experimental introduction of DNA by electroporation, would be influenced presumably by host RM systems. Indeed, strain-specific differences in competence have been noted in Campylobacter [1,57]. RM system variation would also impact infection by both lytic and lysogenic bacteriophages. Future studies will be able to determine the functional status of the RM systems and their role in natural competence and phage restriction.
Campylobacter Metabolism
There have been relatively few studies of the metabolic capabilities of Campylobacter spp., but they are known to have a respiratory type of metabolism, with some strains growing under both aerobic and anaerobic conditions [58,59]. Carbohydrates in general are not utilized. Comparative analysis of the genomes of C. jejuni RM1221, C. coli RM2228, C. lari RM2100, and C. upsaliensis RM3195 revealed that these species have very similar metabolic profiles, with the main variation being the presence of a complete or partial tricarboxylic acid cycle (Figure S6). In C. jejuni RM1221, the tricarboxylic acid cycle appears to be intact and most likely serves a dual role of generating biosynthetic compounds and providing intermediates that feed into electron transport. C. coli RM2228, C. upsaliensis RM3195, and C. lari RM2100 apparently lack a succinate dehydrogenase, and none of the strains appear to encode SucAB (oxoglutarate dehydrogenase). All four sequenced strains have pathways for the metabolism and biosynthesis of a number of amino acids (Figure S6), and acetate, formate, and lactate appear to be the main end products of carbon metabolism. Preliminary Biolog data demonstrate differences in substrate utilization patterns across the Campylobacter strains in this study. C. jejuni RM1221, C. coli RM2228, and C. lari RM2100 all respire in the presence of arabinose, fucose, and formic and lactic acid. In addition, C. jejuni RM1221 respires in the presence of fructose, mannose, hydroxybutyric acid, asparagine, and aspartic acid, in contrast to the other species. These observed phenotypic differences from the preliminary Biolog data may be a reflection either of the conditions under which the substrates were tested or of C. jejuni having pathways that are lost in the other strains. Because of the lack of complete genomes from the other strains, we cannot say with confidence what the reason is for the observed differences, but variable patterns in substrate utilization by Campylobacter species have previously been described [60]. Some of these substrate utilization differences might stem from strain- and species-specific ORFs present in these isolates, or from simple gene mutations that cannot be detected at the genome level. In C. jejuni NCTC 11168, for example, the inability to grow on sugars that are added to the growth medium is felt to be a reflection of the missing phosphofructokinase that is necessary for glycolysis [28]. Interestingly, for all the ɛ-Proteobacteria included in this study, no phosphofructokinase could be identified except for W. succinogenes, enabling Wolinella to metabolize a wider range of carbohydrates than Campylobacter.
Chromosomally Encoded Protein Secretion Systems
The five Campylobacter strains analyzed in this study have the Sec-dependent and Sec-independent (twin-arginine translocation “TAT”) protein export pathways for the secretion of proteins across the inner/periplasmic membrane. In addition, Campylobacter has the signal recognition particle pathway. We have found no evidence for chromosomally encoded lol, Type III, or Type IV secretion systems other than the flagellar export apparatus [61]. In all five strains, there are putative proteins that comprise components of a transformation system with similarity to Type II secretion systems [62]. A putative pre-pilin peptidase and several putative pseudopilins have been identified based on BLASTP similarity or the presence of an N-terminal pre-pilin peptidase cleavage signal (Table S4). The two-partner secretion/single accessory pathway [63] is used by Gram-negative bacteria to secrete adhesins and cytolysins [63]. There are undisrupted copies of putative pore-forming single accessory factors (generically termed TpsB homologs) in C. coli RM2228 (CCO0190), C. lari RM2100 (CLA0150), and C. jejuni NCTC 11168 (Cj0975); however, CCO1305 in C. coli and CJE0841–CJE0843 and CJE1056 in C. jejuni RM1221 are disrupted (Figure S7). It is unclear whether these disruptions are real in the unfinished genomes or whether there would be any consequence for the disruption in C. jejuni RM1221.
Virulence
The pathogenic mechanisms responsible for acute intestinal infections by Campylobacter, although still poorly understood, are thought to involve adherence, cellular invasion, and toxin production, but not all clinical isolates of C. jejuni are able to invade cultured human cells or produce defined toxins [64]. However, a common feature of Campylobacter infectious enterocolitis is a localized acute inflammatory response that can lead to tissue damage and may be responsible for many of the clinical symptoms [64].
Motility is the major factor that has been implicated directly in intestinal colonization [65]. Of the 580 ORFs conserved between the Campylobacter and Helicobacter species included in this study (see Figure 1C), 27 ORFs involved in flagellar biosynthesis and function were conserved between Campylobacter and Helicobacter. Another set of 18 ORFs involved in chemotaxis and motility was found to be conserved across the Campylobacter strains, but with no bidirectional match in Helicobacter (criteria: p-value ≤ 10−5, identity ≥ 35%, match lengths of at least 75% of the length of both query and subject sequence), emphasizing the importance of bacterial motility and adhesion for virulence [66].
Two-component regulatory (TCR) systems are used commonly by bacteria to respond to specific environmental signals. We identified five TCR systems (pairs of adjacent histidine kinase and response regulator genes) that appear to be conserved across the Campylobacter spp.: CJE0968–CJE0969, CJE1357–CJE1358, CJE1361–CJE1362, racR–racS (CJE1397– CJE1398), and CJE1664–CJE1665. In addition, another four putative response regulator genes (CJE0746, CJE0404, CJE1168, and CJE1780) and one putative histidine kinase gene (CJE0884) could be found in the finished C. jejuni genomes. Brás et al. [67] showed that the RacR–RacS system is involved in a temperature-dependent signaling pathway and is required for the organism to colonize the chicken intestinal tract. The high degree of conservation of these ORFs suggests an importance in the Campylobacter pathogenicity, not surprising given the likely exposure of the bacteria to temperature stress during the infectious process.
Adherence of C. jejuni to epithelial cells is mediated by multiple adhesins, including CadF (CJE1651), PEB1 (CJE0997–CJE1000), JlpA (CJE1065), and a 43-kDa major outer membrane protein (CJE1395). Fibronectin (FN) has been implicated in C. jejuni adherence to epithelial cells via the protein CadF [68]. In addition to CadF, we found two putative FN-binding proteins (CJE1415 and CJE1538) that are conserved across the five Campylobacter strains. The FN host cell-surface receptor is the α5β1 integrin. In intact epithelia, α5β1 integrins are restricted to the basolateral membrane and thus are not available for interaction with luminally positioned microbial pathogens [69]. However, Monteville et al. showed that adherence and internalization of C. jejuni were greatly increased by exposure of cellular basolateral surfaces, and that FN was the receptor [70]. This suggests that C. jejuni invasion preferentially occurs via a paracellular route, rather than via an intracellular route. Additionally, inspection of loci adjacent to putative TpsB proteins revealed two intact filamentous hemagglutinin (FHA)–like adhesions: in C. lari RM2100, CLA0151, and in C. coli RM2228, CCO1312. The regions upstream of the remaining TpsB-like proteins have fragmented adhesion-like ORFs (Table 1; Figure S7). Only C. lari RM2100 has both an undisrupted TpsB-like transporter (CLA0150) and an adjacent putative FHA-like adhesion (CLA0151), which, if functional, could enable C. lari RM2100 to attach to cell surfaces.
Cytolethal distending toxins from enteropathogenic Escherichia coli have been shown to disrupt the barrier function of host intestinal epithelial tight junctions [71]. The three cytolethal distending toxins A, B, and C (CJE0075, CJE0074, and CJE0073) were conserved across the five Campylobacter strains. In addition, C. lari RM2100 encodes a single peptide (CLAA0034) in pCL46 that is similar to the Yersinia invasin proteins that enable Yersinia to penetrate host cells [72], suggesting that this C. lari strain might also have the ability to penetrate host cells.
Identification of a Novel Campylobacter Putative Virulence Locus
Examination of the C. upsaliensis RM3195 sequence revealed a putative licABCD (CUP0277–CUP0274) locus with varying, but significant, identity to genes present in Haemophilus influenzae [73], commensal Neisseria species [74], and Streptococcus pneumoniae [75]. licABCD genes in these microorganisms encode proteins involved in the acquisition of choline (licB, CUP0276), synthesis of phosphorylcholine (PCho) (licA, CUP0277; licC, CUP0275), and transfer of PCho (licD, CUP0274) to LOS or teichoic/lipoteichoic acids to facilitate attachment to host cells [74]. Preliminary studies indicate that other strains of C. upsaliensis from South Africa also contain licA (unpublished data). It is noteworthy that licA expression in Haem. influenzae is regulated by variation in the number of intragenic tandem tetranucleotide repeats (
CAAT) at the 5′ end, resulting in translational on/off synthesis of PCho and expression on LOS [76]. A poly G tract within the licA gene (bp 132–146) of C. upsaliensis RM3195 probably regulates synthesis of PCho and decoration of LOS by a similar mechanism.
Hypervariable Homopolymeric Tracks
The presence of the homopolymeric repeat sequences in the genome of C. jejuni NCTC 11168 has been described [28]. However, in comparing these five Campylobacter strains, a number of other phenomena related to these repetitive regions were observed. First, when a homopolymeric repeat region was associated with a potential coding region, the base mostly included in the repeated region on the coding strand was G, resulting in poly-glycine, not poly-proline, in the peptide. Secondly, the C. upsaliensis RM3195 genome contains nearly three times as many variable homopolymeric repeats (22) as C. jejuni RM1221 (8), seven times as many as C. lari RM2100 (3), and 22 times as many as C. coli RM2228 (1) (Table 1). These varied C. upsaliensis RM3195 poly G:C tracts come from a pool of almost five times as many total poly G:C tracts (Table 1) as C. jejuni RM1221 and C. coli RM2228, and nearly ten times as many total poly G:C tracts as C. lari RM2100. Of these 22 varied poly G:C tracts, 11 (50%) are strain-specific (Tables S5 and S6). It appears that excess variable poly G:C tracts are due to the presence of unique ORFs; however, it is unclear as to why C. upsaliensis RM3195 contains so many more total homopolymeric repeated regions, since only 61 of the 209 regions are within unique ORFs. These variable regions encode a combination of hypothetical, cell envelope, and virulence-associated ORFs (Table S6), which in other pathogenic bacteria has been shown to be the molecular basis of lipopolysaccharide phase variation [77], has been used to identify novel virulence genes in Haem. influenzae [78], and has been speculated to have a similar role in C. jejuni [28]. However, these observed differences could be the result of different culturing conditions prior to library construction.
LOS and EP Biosynthesis
LOSs and EPs are important surface structures in C. jejuni that function in the interactions of the organism with the environment. Interesting aspects of C. jejuni LOSs are their molecular mimicry of host gangliosides and their presumed roles in evasion of host immune responses and autoimmunity [79], decreased immunogenicity [80], and attachment and invasion [48]. The capsule of C. jejuni 81–176 has been reported to have a role in increasing serum resistance, invasion of cell lines, and surface hydrophilicity [81].
The LOS biosynthesis loci of all sequenced Campylobacter spp. are organized as previously observed in other C. jejuni strains [82]. At either end of the loci are the heptosyltransferase genes, waaC and waaF, that surround regions exhibiting significant variation in ORF content. Thus, these organisms likely synthesize novel LOS structures [82]. In particular, the LOS of C. jejuni RM1221 is distinct from the LOS of NCTC 11168, as seen on polyacrylamide gels, in that it possesses three LOS bands while NCTC 11168 possesses only one (unpublished data). Two LOS genes from C. jejuni RM1221 possess homopolymeric G:C tracts that may explain the additional bands. Comparison of the LOS genes from the sequenced Campylobacter spp. with those from C. jejuni strains that produce ganglioside mimics [29] demonstrates that these four strains do not possess the genes involved in the synthesis of N-acetylneuramic (sialic) acid or the associated sialic acid transferase, and are not likely to produce ganglioside mimics. Within the LOS loci of C. lari RM2100 and C. upsaliensis RM3195, there are ORF clusters that have homologs in NCTC 11168 that are unrelated to LOS biosynthesis. It is unclear what role this genomic reorganization plays in the biosynthesis of LOS.
C. jejuni RM1221, C. coli RM2228, and C. lari RM2100 possess kps orthologs like the EP locus of C. jejuni NCTC 11168 that are involved in polysaccharide export; however, many putative EP biosynthesis genes from C. jejuni RM1221 and C. coli RM2228 are unique to these strains. The kps orthologs are present in C. upsaliensis RM3195, but they are not clustered with other polysaccharide biosynthetic genes as observed in the other strains. Specifically, there are three clusters of EP genes: CUP0615–CUP0619, CUP1248–CUP1270, and CUP1328–CUP1329. The second cluster contains many ORFs that are unique to C. upsaliensis (Table S5), including two of the three copies of a putative GDP-fucose synthetase (CUP1255, CUP1257, and CUP1258). Only C. jejuni strains (Cj1428c and CJE1612) and C. upsaliensis RM3195 encode this enzyme. Of these GDP-fucose synthetases, only CUP1257 was shown to contain variable poly G tracts (Table S6).
Antibiotic Resistance
The sequenced Campylobacter strains have adapted or acquired many mechanisms of antibiotic resistance (Tables 2 and S2). All strains are resistant to cloxacillin, nafcillin, oxacillin, sulfamethoxazole/Tm, trimethoprim, and vancomycin, and this resistance is likely inherent to all Campylobacter spp. (Table S2). Every strain but C. upsaliensis RM3195 is resistant to most β-lactam antibiotics. This general lack of resistance to β-lactam antibiotics for RM3195 is likely due to the disruption of a class D β-lactamase matching GenBank accession AAT01092 (CUP0345), which was found as an intact single copy in NCTC 11168 (Cj0299), RM1221 (CJE0344), and RM2100 (CLA0304). The corresponding sequence in C. coli RM2228 may reside in unsequenced regions. Only C. lari RM2100 was resistant to a broad range of quinolone/fluoroquinolone antibiotics (Table 2). This broad quinolone/fluoroquinolone resistance is most likely the result of adaptation via a mutation of DNA gyrase (gyrA) that changed codon 86 from threonine to valine [83]. The macrolide antibiotics azithromycin, clindamycin, erythromycin, and tilmicosin were effective against all but C. coli RM2228. This is likely due to a mutation in all three copies of the 23S rRNA (A2122G), corresponding to position 2,143 of the H. pylori sequence [84]. C. coli RM2228 has acquired resistance to the aminoglycosides kanamycin and neomycin, tetracycline, oxytetracycline, minocycline, and presumably hygromycin B (but not gentamicin) from the megaplasmid pCC178 (Table 2). It is possible that C. coli has acquired resistance to macrolides and tetracyclines as a result of the application of these drugs during poultry production. The resistance of C. upsaliensis RM3195 to oxytetracycline and its intermediate resistance to tetracycline may be due to the action of multi-drug efflux pumps or a novel mechanism, since there is no evidence for tetracycline resistance genes [85], and there are no known mutations in the 16S rRNA [86]. Similarly, no known mutations in gyrA or gyrB were found in C. upsaliensis RM3195 to explain the resistance to nalidixic acid [83] and novobiocin [87]. There were no obvious known mutations of dihydropteroate synthase (folP) [88] to explain the observed variable resistance to sulfonamide-class drugs (Table 2). Rifampin resistance was observed in all strains but C. lari RM2100, but was not due to the classic mutations in the β subunit of RNA polymerase [89].
Conclusions
The comparison of five sequenced Campylobacter genomes has provided the core genetic blueprint of the genus. Although the blueprint reveals obvious differences in genome structure and content, additional epidemiological data are needed to correlate these differences, and other, more elusive differences (e.g. differences in regulation and point mutations), with differences in virulence. Some obvious differences were the presence of drug resistance genes that may have been the result of adaptation in the animal production environment, where antibiotics are frequently used to eliminate bacterial infections. It is anticipated that the analysis of the Campylobacter genomes presented here will lay the foundation for the development of systems for fingerprinting strains for phylogenetics, epidemiology, and source tracking, as well as the development of alternative treatments for controlling Campylobacter in food production and in human infection.
Materials and Methods
Strain isolation and propagation
C. jejuni strain RM1221 (
ATCC BAA-1062) was isolated from the skin of a retail chicken using methods modified from those described previously for isolation of Campylobacter from chicken products [31]. C. coli strain RM2228 (
ATCC BAA-1061) was isolated from a chicken carcass obtained from an inspected slaughter plant. A rinse sample was streaked on 5% sheep blood agar plates, and the plates were incubated at 37 °C for 48 h under an atmosphere of 5% O2, 10% CO2, and balance N2. An isolated single colony was picked and maintained on sheep blood agar plates. Three rounds of mixing and sonication of single colony picks were done as described [31]. C. lari strain RM2100 (
ATCC BAA-1060) is a human isolate obtained from the Centers for Disease Control and Prevention, Atlanta, Georgia, United States (CDC strain D67, “case 6” [32]). The strain was maintained on Brucella agar amended with 5% (v/v) laked horse blood (Hema Resource and Supply, Aurora, Oregon, United States). Three rounds of mixing and sonication of single colony picks were done as described [31]. C. upsaliensis strain RM3195 (
ATCC BAA-1059) was obtained from the feces of a 4-y-old boy confirmed clinically to have Guillain-Barré syndrome. The isolation procedure involved a filtration method with selection of Campylobacter cells in diluted feces by their migration through a 0.6-μm membrane filter and subsequent growth on nonselective medium [33].
Genome sequencing
The four species of Campylobacter were sequenced by the random shotgun method [38]. The genome of C. jejuni RM1221 was sequenced to closure, whereas the genomes of strains C. lari RM2100, C. coli RM2228, and C. upsaliensis RM3195 were sequenced to 8-fold coverage of an estimated 1.8-Mbp genome. Briefly, one small insert plasmid library (1.5–2.5 kb) and one medium insert plasmid library (10–12 kb) were constructed for each strain (except RM1221, which had only a small insert library) by random nebulization and cloning of genomic DNA. In the random sequencing phase, 8-fold sequence coverage was achieved from the two libraries (sequenced to 5-fold and 3-fold coverage, respectively). The sequences from the respective strains were assembled separately using TIGR Assembler [90] or Celera Assembler [91]. All sequence and physical gaps for C. jejuni RM1221 were closed by editing the ends of sequence traces, primer walking or transposon-primed sequencing [92] on plasmid clones, and combinatorial PCR followed by sequencing of the PCR product. The correct nucleotide sequences for repetitive regions greater than the maximum insert size of 2.5 kb (i.e., rRNA operons) for C. jejuni RM1221 were confirmed by sequencing PCR products that spanned each repeat unit. Pseudomolecules for the draft sequences were constructed using NUCmer [93] and BAMBUS [38,94] as previously described [38].
Ambiguity rate
The ambiguity rate for the unfinished genomes was determined using the following procedure. First, the consensus of the contigs was recalled using the consensus caller included in the AutoEditor package (http://www.tigr.org/software/autoeditor/) [95] by executing “autoEditor—noedit” on the final contigs. This step was necessary because the contigs as produced by the Celera Assembler were made with a consensus caller which does not assign ambiguity codes, but instead assigns a base call arbitrarily in the event of a tie or near tie situation. The AutoEditor consensus caller recomputes the consensus at each position and assigns an ambiguity code if there is sufficient conflicting information. Using a custom script, a count was made of both the overall number of positions and the number of ambiguous positions with at least the specified depth of coverage. This was necessary because the depth of coverage in the assemblies is not uniform, but directly influences the ambiguity rate. For example, under the AutoEditor ambiguity model, there are no ambiguous positions at 1-fold coverage. The ambiguity rate is then reported as the ratio of the two counts, as a close approximation to the error rate of the true consensus sequence.
Annotation
An initial set of ORFs that likely encode proteins was identified using GLIMMER [96], and those shorter than 90 bp or those with overlaps were eliminated. ORFs were searched against a nonredundant protein database; frameshifts and point mutations were processed only for C. jejuni RM1221 [38]. Two sets of hidden Markov models were used to determine ORF membership in families and superfamilies [38].
Comparative genomics
For the identification of species-specific (Table S7) and strain-specific (Table S5) ORFs, all predicted proteins (excluding pseudogenes) from the four TIGR-sequenced Campylobacter genomes and C. jejuni NCTC 11168 [28] were searched against an in-house database composed of 734,467 protein sequences encoded by 19 archaeal, 192 bacterial, 146 eukaryotic, three phage, and 17 virus chromosomes, as well as 145 plasmid, 29 mitochondrial, 17 plastid, and three nucleomorph genomes, using WU-BLASTP (http://blast.wustl.edu) [97]. To identify genus-specific ORFs, the protein sequences from the above five Campylobacter genomes plus three Helicobacter genomes (H. pylori 26695 [98], H. pylori J99 [99], and H. hepaticus
ATCC 51449 [44]) and the genome of W. succinogenes DSMZ1740 [100] were compared. Specifically, only bidirectional best matches that met the following prerequisites were scored: a p-value less than or equal to 10−5, identity of 35% or more, and match lengths of at least 75% of the length of both query and subject sequence. Match tables were created that were later used to generate the Venn diagrams (Tables S8 and S9). Novel ORFs encoded proteins that had no WU-BLASTP match. Regions of synteny were identified by first finding the maximum unique matches with a minimum length of five amino acids using PROmer, followed by visualization of the data using MUMmerplot (http://www.tigr.org) and Gnuplot version 4.0 (http://www.gnuplot.info/).
MLST and FlaA SVR typing
The MLST of C. jejuni RM1221 was determined by searching the nucleotide sequences of aspartate ammonia-lyase (aspA, CJE0082), glutamine synthetase type I (glnA, CJE0798), citrate synthase (gltA, CJE1851), serine hydroxymethyltransferase (glyA, CJE0451), phosphoglucosamine mutase (pgm/glmM, CJE0409), transketolase (tkt, CJE1817), and ATP synthase F1 alpha subunit (uncA/atpA, CJE0100) on the PubMLST Web site (http://pubmlst.org/) [101]. The sequence of the C. jejuni RM1221 FlaA SVR was found by searching the flaA (CJE1528) nucleotide sequence using the sequence of primers FLA242FU and FLA625RU [34]. This nucleotide sequence was used to query the flaA allele database (http://phoenix.medawar.ox.ac.uk/flaA/) to elucidate the FlaA SVR type [34,102].
Phylogenetic analysis
The programs SEQBOOT, DNAML, PROML, and CONSENSE are part of the PHYLIP version 3.62 package (http://evolution.genetics.washington.edu/phylip.html, http://fink.sourceforge.net/) [103]. Both the 16S rRNA and concatenated protein trees were rooted to the δ-Proteobacterium Desulfovibrio vulgaris subsp. vulgaris strain Hildenborough sequences [104]. One hundred bootstrapped datasets were generated using the SEQBOOT program, and consensus trees were determined using CONSENSE. The final trees with preserved branch lengths were computed with the user tree option of DNAML and PROML.
16S rRNA trees were generated by first creating a multiple alignment using the “PHYLIP Interface” option of the Ribosomal Database Project release 8.1 (http://35.8.164.52/cgis/phylip.cgi, which aligns user-supplied 16S rRNA sequences against the Ribosomal Database Project alignment. The produced alignment was trimmed and gaps removed using an in-house PERL (http://www.perl.org) script. Maximum-likelihood trees were generated using DNAML (R = gamma-distributed rate of variation [coefficient of variation, 1.41; four hidden Markov model rate categories] and S = NO).
Protein trees were generated from concatenated multiple alignments of 12 conserved proteins (initiation factor 2 [InfB]; elongation factors G [FusA] and Tu [Tuf]; ribosomal proteins L2 [RplB], S5 [RpsE], S8 [RpsH], and S11 [RpsK]; DNA topoisomerase I [TopA]; signal recognition particle protein [Ffh] [36]; DNA gyrase B subunit [GyrB]; GTP-binding protein LepA; and CTP synthase [PyrG] [37]). Each protein was aligned separately using CLUSTALW version 1.82 [105], using the slow, more accurate option. The alignments were trimmed to remove gaps using BELVU version 2.16 (http://www.cgb.ki.se/cgb/groups/sonnhammer/Belvu.html). Each organism's aligned sequences were concatenated using an in-house PERL script. Maximum-likelihood trees were generated using PROML (P = Jones-Taylor-Thornton model of change between amino acids, R = gamma-distributed rate of variation [coefficient of variation, 1.41; four hidden Markov model rate categories], and S = NO).
Hypervariable homopolymeric G or C tracts
Hypervariable homopolymeric G or C tracts were identified by analyzing the underlying sequences for each nucleotide within a tract of six or more G or C nucleotides. A hypervariable tract was considered of high quality if its underlying sequence comprised at least three sequencing reads with an average Phred score greater than 30 [106].
Supporting Information
Figure S1 Circular Representation of the Closed C. jejuni RM1221 Genome
Each concentric circle represents genomic data and is numbered from the outermost to the innermost circle. Refer to the key for details on color representations. The first and second circles represent predicted ORFs on the plus and minus strands, respectively. The third circle shows the GC-skew. The fourth circle depicts genetic loci with characteristics or functions of interest: CRISPRs, DNA competence, EP, LOS, prophage and genomic island regions, motility, repeats, and Type I restriction/modification regions. The fifth circle demarcates C. jejuni–specific and C. jejuni RM1221–specific ORFs. The sixth circle plots atypical regions (χ2 value). The seventh circle denotes tRNA, rRNA, and sRNA (tmRNA and 4.5S RNA) loci.
(2.6 MB EPS).
Click here for additional data file.
Figure S2 Linear Illustration of C. jejuni Genome Comparisons
(274 KB PDF).
Click here for additional data file.
Figure S3 Comparison of Plasmid-Like Genomic Islands of C. jejuni RM1221
CJIE3 (top linear figure) and H. hepaticus
ATCC 51449 HHGI1 (bottom line) against pCC178 megaplasmid of C. coli RM2228 (middle line). Colors of ORFs are indicated in the key by putative function. Connecting lines represent those ORFs whose protein sequences match at a BLASTP of 30% identity or better. These lines do not indicate the coordinates of match, merely that there is a match.
(76 KB PDF).
Click here for additional data file.
Figure S4 T4SS Is Shared among the Large Campylobacter Species Plasmids but Is Not the Same as C. jejuni T4SS
(A) shows a conceptual diagram indicating where each of the proteins thought to be involved in the T4SS interact. Each corresponding loci is color-coded in each of the plasmids.
(B) The T4SS in each of the plasmids demonstrates that a number of the core proteins are conserved in all of the Campylobacter plasmids; however, the non–C. jejuni plasmids contain a structure that is more similar to the Agrobacterium tumefaciens T4SS. (In the Campylobacter plasmids, black ORFs are those not directly involved in the T4SS; however, many are similar to plasmid transfer proteins).
(5.3 MB EPS).
Click here for additional data file.
Figure S5 DNA Sequences of the CRISPR Elements Found in the Two Strains of C. jejuni, RM1221 and NCTC 11168
The characters in italics indicate the 32-bp spacer sequences that are unique to the two strains; the spacer sequences for NCTC 11168 are 1 bp longer than presented by others [52]. The bold characters represent the CRISPR repeat region in RM1221 (n = 4) and NCTC 11168 (n = 5). The characters in roman typeface indicate regions flanking the repeat region that are identical in the two strains.
(20 KB DOC).
Click here for additional data file.
Figure S6 Main Pathways for Metabolism Derived from an Analysis of Five Campylobacter Genomes
The tricarboxylic (TCA) cycle has major variations based on comparative analysis across the strains (please refer to text). Differences in substrate respiration based on an analysis of Biolog data and species-specific pathways are also presented in the text.
(51 KB PPT).
Click here for additional data file.
Figure S7 Putative Two-Partner/Single Accessory Secretion Loci
FhaC, the single accessory protein that secretes the Bordetella pertussis FHA across the outer membrane, was used as the query for BLASTP searches against a database containing Campylobacter protein sequences. Fragments of single accessory proteins were found as matches in the Campylobacter match table (see Table S8). Putative single accessory protein/TpsB family proteins (teal) and putative FHAs/hemolysins (red) are noted, as well as putative proteins with weak matches to metacaspases or toxins (tan). The small red ORFs suggest fragmentation of a larger, full-length ORF.
(1.6 MB EPS).
Click here for additional data file.
Table S1 Comparison of Campylobacter Species Plasmids
(19 KB XLS).
Click here for additional data file.
Table S2 Antibiotic Susceptibility Profiles
(22 KB XLS).
Click here for additional data file.
Table S3
C. jejuni, C. coli, C. lari, and C. upsaliensis Restriction-Modification
(22 KB XLS).
Click here for additional data file.
Table S4 Putative DNA Competence Genes
(16 KB XLS).
Click here for additional data file.
Table S5 Strain-Specific Genes with Annotations
(238 KB XLS).
Click here for additional data file.
Table S6 Hypervariable Homopolymeric Sequences Found in Campylobacter Genomes
(57 KB XLS).
Click here for additional data file.
Table S7
C. jejuni–Specific Genes with Annotations
(31 KB XLS).
Click here for additional data file.
Table S8 Match Table Depicting Bidirectional Best Matches of Campylobacter Species
(647 KB XLS).
Click here for additional data file.
Table S9 Match Table Depicting Bidirectional Best Matches of Sequenced ɛ-Proteobacteria
(894 KB XLS).
Click here for additional data file.
Table S10 Arg-Gly-Asp, Lipoprotein, Outer Membrane Protein Signal, Secretion Signal, and Transmembrane Motif Results
(155 KB XLS).
Click here for additional data file.
Accession Numbers
The nucleotide sequence for the closed genome of C. jejuni RM1221 has been deposited at the DNA Data Bank of Japan (DDBJ; http://www.ddbj.nig.ac.jp/, the European Molecular Biology Laboratory Nucleotide Sequence Database (EMBL; http://www.ebi.ac.uk/embl/, and GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) under accession number CP000025. The whole-genome shotgun projects for the genomes of C. lari RM2100, C. coli RM2228, and C. upsaliensis RM3195 that were sequenced to at least 8-fold coverage were deposited at DDBJ, EMBL, and GenBank under accession numbers AAFK00000000, AAFL00000000 and AAFJ00000000, respectively. The versions described in this paper are the first versions, AAFK01000000, AAFL01000000 and AAFJ01000000, respectively. Additionally, all sequence traces and assemblies were deposited at the National Center for Biotechnology Information assembly archive (http://www.ncbi.nlm.nih.gov/Traces/assembly). The contig separator that was used to create the pseudomolecules for the unfinished genomes is NNNNN
TTAATTAATTAANNNNN.
The authors thank T. Feldblyum, T. Utterback, S. Van Aken, J. Kolonay, H. Koo, K. Saeed, W. Nelson, D. Haft, L. Zhou, M. Heaney, S. Lo, and M. Brown at TIGR for support with various aspects of this project. Special thanks go to Jonathan Eisen for help with phylogenetic analysis and Martin Wu for allowing use of his ComboDB. C. coli strain RM2228 was isolated by M. Englen and P. Cray (Agricultural Research Service, Athens, Georgia, United States) as part of the National Antibiotic Resistance Monitoring System project. C. lari strain RM2100 (CDC strain D67) was provided by M. Nicholson, formerly of the Centers for Disease Control and Prevention. C. upsaliensis strain RM3195 was isolated by A. Lastovica (University of Cape Town, Cape Town, South Africa). Serotyping of C. jejuni RM1221 and C. coli RM2228 was performed generously by David Woodward at the National Microbiology Laboratory, Winnipeg, Canada. This project was funded by the United States Department of Agriculture, Agricultural Research Service Agreement # 58–1935-0–004. 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 US Department of Agriculture.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DEF, EFM, REM, WGM, MCS, JHB, and KEN conceived and designed the experiments. DEF, EFM, WGM, JUS, MAA, and AS performed the experiments. DEF, EFM, WGH, DAR, JR, LMB, RTD, CTP, SCD, RJD, ASD, RM, SAS, JUS, MCS, JHB, and KEN analyzed the data. DEF, EFM, REM, WGM, DAR, JR, LMB, RTD, CTP, JUS, MAA, AS, MCS, JHB, CMF, and KEN contributed reagents/materials/analysis tools. DEF, EFM, REM, WGM, DAR, RTD, CTP, MCS, JHB, and KEN wrote the paper.
Citation: Fouts DE, Mongodin EF, Mandrell RE, Miller WG, Rasko DA, et al. (2005) Major structural differences and novel potential virulence mechanisms from the genomes of multiple Campylobacter species. PLoS Biol 3(1): e15.
Abbreviations
CJIE
Campylobacter jejuni RM1221 integrated element
CLIE
Campylobacter lari RM2100 integrated element
CMLP1
Campylobacter Mu-like phage
CRISPRclustered regularly interspaced short palindromic repeat
CUIE
Campylobacter upsaliensis RM3195 integrated element
EPcapsular (extracellular) polysaccharide
FHAfilamentous hemagglutinin
FNfibronectin
HHGI1
Helicobacter hepaticus
ATCC 51449 genomic island
ISinsertion sequence
LOSlipooligosaccharide
MLSTmultilocus sequence type
ORFopen reading frame
PChophosphorylcholine
RMrestriction–modification
SVRshort variable region
T4SSType IV secretion system
TCRtwo-component regulatory
==== Refs
References
Miller WG Mandrell RE Ketley J Konkel ME
Campylobacter in the food and water supply: Prevalence, outbreaks, isolation, and detection Campylobacter jejuni : New perspectives in molecular and cellular biology 2004 Norfolk (United Kingdom) Horizon Scientific Press 109 163
Friedman CR Neimann J Wegener HC Tauxe RV Nachamkin I Blaser MJ Epidemiology of Campylobacter jejuni infections in the United States and other industrialized nations Campylobacter 2000 Washington (DC) ASM Press 121 138
Vandamme P Nachamkin I Blaser MJ Taxonomy of the family Campylobacteraceae
Campylobacter 2000 Washington (DC) ASM Press 3 26
Lastovica AJ Skirrow MB Nachamkin I Blaser MJ Clinical significance of Campylobacter and related species other than Campylobacter jejuni and C. coli
Campylobacter 2000 Washington (DC) ASM Press 89 120
Nachamkin I Allos BM Ho T
Campylobacter species and Guillain-Barre syndrome Clin Microbiol Rev 1998 11 555 567 9665983
Willison HJ O'Hanlon GM The immunopathogenesis of Miller Fisher syndrome J Neuroimmunol 1999 100 3 12 10695710
Kist M Bereswill S
Campylobacter jejuni
Contrib Microbiol 2001 8 150 165 11764732
Glunder G Petermann S [The occurrence and characterization of Campylobacter spp. in silver gulls (Larus argentatus ), three-toed gulls (Rissa tridactyla) and house sparrows (Passer domesticus )] Zentralbl Veterinarmed B 1989 36 123 130 2750361
Hald B Pedersen K Waino M Jorgensen JC Madsen M Longitudinal study of the excretion patterns of thermophilic Campylobacter spp. in young pet dogs in Denmark J Clin Microbiol 2004 42 2003 2012 15131162
Harvey RB Young CR Ziprin RL Hume ME Genovese KJ Prevalence of Campylobacter spp isolated from the intestinal tract of pigs raised in an integrated swine production system J Am Vet Med Assoc 1999 215 1601 1604 14567422
Engvall EO Brandstrom B Andersson L Baverud V Trowald-Wigh G Isolation and identification of thermophilic Campylobacter species in faecal samples from Swedish dogs Scand J Infect Dis 2003 35 713 718 14606609
Hald B Madsen M Healthy puppies and kittens as carriers of Campylobacter spp., with special reference to Campylobacter upsaliensis
J Clin Microbiol 1997 35 3351 3352 9399557
Moser I Rieksneuwohner B Lentzsch P Schwerk P Wieler LH Genomic heterogeneity and O-antigenic diversity of Campylobacter upsaliensis and Campylobacter helveticus strains isolated from dogs and cats in Germany J Clin Microbiol 2001 39 2548 2557 11427567
Sandstedt K Ursing J Walder M Thermotolerant Campylobacter with no or weak catalase activity isolated from dogs Curr Microbiol 1983 8 209 213
Shen Z Feng Y Dewhirst FE Fox JG Coinfection of enteric Helicobacter spp. and Campylobacter spp. in cats J Clin Microbiol 2001 39 2166 2172 11376052
Tresierra-Ayala A Bendayan ME Bernuy A Pereyra G Fernandez H Chicken as potential contamination source of Campylobacter lari in Iquitos, Peru Rev Inst Med Trop Sao Paulo 1994 36 497 499 7569622
Moore JE Madden RH Occurrence of thermophilic Campylobacter spp. in porcine liver in Northern Ireland J Food Prot 1998 61 409 413 9709202
Kramer JM Frost JA Bolton FJ Wareing DR
Campylobacter contamination of raw meat and poultry at retail sale: Identification of multiple types and comparison with isolates from human infection J Food Prot 2000 63 1654 1659 11131886
Park CE Sanders GW Occurrence of thermotolerant campylobacters in fresh vegetables sold at farmers' outdoor markets and supermarkets Can J Microbiol 1992 38 313 316 1611556
Endtz HP Vliegenthart JS Vandamme P Weverink HW van den Braak NP Genotypic diversity of Campylobacter lari isolated from mussels and oysters in the Netherlands Int J Food Microbiol 1997 34 79 88 9029257
Rosef O Rettedal G Lageide L Thermophilic campylobacters in surface water: A potential risk of campylobacteriosis Int J Environ Health Res 2001 11 321 327 11798419
Bourke B Chan VL Sherman P
Campylobacter upsaliensis Waiting in the wings Clin Microbiol Rev 1998 11 440 449 9665977
Atanassova V Ring C Prevalence of Campylobacter spp. in poultry and poultry meat in Germany Int J Food Microbiol 1999 51 187 190 10574094
Goossens H Vlaes L Butzler JP Adnet A Hanicq P
Campylobacter upsaliensis enteritis associated with canine infections Lancet 1991 337 1486 1487
Gurgan T Diker KS Abortion associated with Campylobacter upsaliensis
J Clin Microbiol 1994 32 3093 3094 7883913
Walmsley SL Karmali MA Direct isolation of atypical thermophilic Campylobacter species from human feces on selective agar medium J Clin Microbiol 1989 27 668 670 2723035
Goossens H Giesendorf BA Vandamme P Vlaes L Van den Borre C Investigation of an outbreak of Campylobacter upsaliensis in day care centers in Brussels: Analysis of relationships among isolates by phenotypic and genotypic typing methods J Infect Dis 1995 172 1298 1305 7594667
Parkhill J Wren BW Mungall K Ketley JM Churcher C The genome sequence of the food-borne pathogen Campylobacter jejuni reveals hypervariable sequences Nature 2000 403 665 668 10688204
Gilbert M Brisson JR Karwaski MF Michniewicz J Cunningham AM Biosynthesis of ganglioside mimics in Campylobacter jejuni OH4384. Identification of the glycosyltransferase genes, enzymatic synthesis of model compounds, and characterization of nanomole amounts by 600-mhz (1)h and (13)c NMR analysis J Biol Chem 2000 275 3896 3906 10660542
Ahmed IH Manning G Wassenaar TM Cawthraw S Newell DG Identification of genetic differences between two Campylobacter jejuni strains with different colonization potentials Microbiology 2002 148 1203 1212 11932464
Miller WG Bates AH Horn ST Brandl MT Wachtel MR Detection on surfaces and in Caco-2 cells of Campylobacter jejuni cells transformed with new gfp, yfp and cfp marker plasmids Appl Environ Microbiol 2000 66 5426 5436 11097924
Tauxe RV Patton CM Edmonds P Barrett TJ Brenner DJ Illness associated with Campylobacter laridis a newly recognized Campylobacter species J Clin Microbiol 1985 21 222 225 3972989
le Roux E Lastovica AJ Lastovica AJ Newell DG Lastovica EE The Cape Town Protocol: How to isolate the most campylobacters for your dollar, pound, franc, yen, etc Campylobacter, Helicobacter and related organisms 1998 Cape Town Institute of Child Health 30 33
Dingle KE Colles FM Ure R Wagenaar JA Duim B Molecular characterization of Campylobacter jejuni clones: A basis for epidemiological investigation Emerg Infect Dis 2002 8 949 955 12194772
Eisen JA Heidelberg JF White O Salzberg SL Evidence for symmetric chromosomal inversions around the replication origin in bacteria Genome Biol 2000 1 RESEARCH0011 11178265
Brown JR Douady CJ Italia MJ Marshall WE Stanhope MJ Universal trees based on large combined protein sequence data sets Nat Genet 2001 28 281 285 11431701
Santos SR Ochman H Identification and phylogenetic sorting of bacterial lineages with universally conserved genes and proteins Environ Microbiol 2004 6 754 759 15186354
Nelson KE Fouts DE Mongodin EF Ravel J DeBoy RT Whole genome comparisons of serotype 4b and 1/2a strains of the food-borne pathogen Listeria monocytogenes reveal new insights into the core genome components of this species Nucleic Acids Res 2004 32 2386 2395 15115801
Kuroda M Ohta T Uchiyama I Baba T Yuzawa H Whole genome sequencing of meticillin-resistant Staphylococcus aureus
Lancet 2001 357 1225 1240 11418146
Baba T Takeuchi F Kuroda M Yuzawa H Aoki K Genome and virulence determinants of high virulence community-acquired MRSA Lancet 2002 359 1819 1827 12044378
Beres SB Sylva GL Barbian KD Lei B Hoff JS Genome sequence of a serotype M3 strain of group A Streptococcus : Phage-encoded toxins, the high-virulence phenotype, and clone emergence Proc Natl Acad Sci U S A 2002 99 10078 10083 12122206
Perna NT Plunkett G Burland V Mau B Glasner JD Genome sequence of enterohaemorrhagic Escherichia coli O157:H7 Nature 2001 409 529 533 11206551
Morgan GJ Hatfull GF Casjens S Hendrix RW Bacteriophage Mu genome sequence: Analysis and comparison with Mu-like prophages in Haemophilus , Neisseria and Deinococcus
J Mol Biol 2002 317 337 359 11922669
Suerbaum S Josenhans C Sterzenbach T Drescher B Brandt P The complete genome sequence of the carcinogenic bacterium Helicobacter hepaticus
Proc Natl Acad Sci U S A 2003 100 7901 7906 12810954
Hendrix RW Lawrence JG Hatfull GF Casjens S The origins and ongoing evolution of viruses Trends Microbiol 2000 8 504 508 11121760
Desiere F McShan WM van Sinderen D Ferretti JJ Brüssow H Comparative genomics reveals close genetic relationships between phages from dairy bacteria and pathogenic Streptococci Evolutionary implications for prophage–host interactions Virology 2001 288 325 341 11601904
Wagner PL Waldor MK Bacteriophage control of bacterial virulence Infect Immun 2002 70 3985 3993 12117903
Bacon DJ Alm RA Hu L Hickey TE Ewing CP DNA sequence and mutational analyses of the pVir plasmid of Campylobacter jejuni 81–176 Infect Immun 2002 70 6242 6250 12379703
Cascales E Christie PJ The versatile bacterial type IV secretion systems Nat Rev Microbiol 2003 1 137 149 15035043
Ding Z Atmakuri K Christie PJ The outs and ins of bacterial type IV secretion substrates Trends Microbiol 2003 11 527 535 14607070
Cascales E Christie PJ Definition of a bacterial type IV secretion pathway for a DNA substrate Science 2004 304 1170 1173 15155952
Schouls LM Reulen S Duim B Wagenaar JA Willems RJ Comparative genotyping of Campylobacter jejuni by amplified fragment length polymorphism, multilocus sequence typing, and short repeat sequencing: Strain diversity, host range, and recombination J Clin Microbiol 2003 41 15 26 12517820
Miller WG Keech AM Pearson BM Wells JM Kapitonov VV Diversity of Campylobacter jejuni Type I restriction-modification loci: Induction of hsdS by exogenous DNA Microbiology 2005 In press
Kobayashi I Behavior of restriction-modification systems as selfish mobile elements and their impact on genome evolution Nucleic Acids Res 2001 29 3742 3756 11557807
Wang Y Taylor DE Natural transformation in Campylobacter species J Bacteriol 1990 172 949 955 2404960
Suerbaum S Lohrengel M Sonnevend A Ruberg F Kist M Allelic diversity and recombination in Campylobacter jejuni
J Bacteriol 2001 183 2553 2559 11274115
Wassenaar TM Fry BN van der Zeijst BA Genetic manipulation of Campylobacter : Evaluation of natural transformation and electro-transformation Gene 1993 132 131 135 8406035
Carlone GM Lascelles J Aerobic and anaerobic respiratory systems in Campylobacter fetus subsp. jejuni grown in atmospheres containing hydrogen J Bacteriol 1982 152 306 314 6288661
Smibert RM Holt JG Krieg NR
Campylobacter
Bergey's manual of systematic bacteriology 1984 Baltimore Williams and Wilkins 111 118
Mohammed KA Miles RJ Halablab MA The pattern and kinetics of substrate metabolism of Campylobacter jejuni and Campylobacter coli
Lett Appl Microbiol 2004 39 261 266 15287872
Konkel ME Klena JD Rivera-Amill V Monteville MR Biswas D Secretion of virulence proteins from Campylobacter jejuni is dependent on a functional flagellar export apparatus J Bacteriol 2004 186 3296 3303 15150214
Wiesner RS Hendrixson DR DiRita VJ Natural transformation of Campylobacter jejuni requires components of a type II secretion system J Bacteriol 2003 185 5408 5418 12949093
Jacob-Dubuisson F Locht C Antoine R Two-partner secretion in Gram-negative bacteria: A thrifty, specific pathway for large virulence proteins Mol Microbiol 2001 40 306 313 11309114
Ketley JM Pathogenesis of enteric infection by Campylobacter
Microbiology 1997 143 (Pt 1) 5 21 9025274
Wassenaar TM van der Zeijst BA Ayling R Newell DG Colonization of chicks by motility mutants of Campylobacter jejuni demonstrates the importance of flagellin A expression J Gen Microbiol 1993 139 1171 1175 8360610
Yao R Burr DH Doig P Trust TJ Niu H Isolation of motile and non-motile insertional mutants of Campylobacter jejuni The role of motility in adherence and invasion of eukaryotic cells Mol Microbiol 1994 14 883 893 7715450
Brás AM Chatterjee S Wren BW Newell DG Ketley JM A novel Campylobacter jejuni two-component regulatory system important for temperature-dependent growth and colonization J Bacteriol 1999 181 3298 3302 10322038
Konkel ME Garvis SG Tipton SL Anderson DE Cieplak W Identification and molecular cloning of a gene encoding a fibronectin-binding protein (CadF) from Campylobacter jejuni
Mol Microbiol 1997 24 953 963 9220003
McCormick BA Nusrat A Parkos CA D'Andrea L Hofman PM Unmasking of intestinal epithelial lateral membrane beta1 integrin consequent to transepithelial neutrophil migration in vitro facilitates inv -mediated invasion by Yersinia pseudotuberculosis
Infect Immun 1997 65 1414 1421 9119482
Monteville MR Konkel ME Fibronectin-facilitated invasion of T84 eukaryotic cells by Campylobacter jejuni occurs preferentially at the basolateral cell surface Infect Immun 2002 70 6665 6671 12438340
Muza-Moons MM Koutsouris A Hecht G Disruption of cell polarity by enteropathogenic Escherichia coli enables basolateral membrane proteins to migrate apically and to potentiate physiological consequences Infect Immun 2003 71 7069 7078 14638797
Isberg RR Voorhis DL Falkow S Identification of invasin: A protein that allows enteric bacteria to penetrate cultured mammalian cells Cell 1987 50 769 778 3304658
Weiser JN Lindberg AA Manning EJ Hansen EJ Moxon ER Identification of a chromosomal locus for expression of lipopolysaccharide epitopes in Haemophilus influenzae
Infect Immun 1989 57 3045 3052 2476397
Serino L Virji M Genetic and functional analysis of the phosphorylcholine moiety of commensal Neisseria lipopolysaccharide Mol Microbiol 2002 43 437 448 11985720
Zhang JR Idanpaan-Heikkila I Fischer W Tuomanen EI Pneumococcal licD2 gene is involved in phosphorylcholine metabolism Mol Microbiol 1999 31 1477 1488 10200966
Weiser JN Shchepetov M Chong ST Decoration of lipopolysaccharide with phosphorylcholine: A phase-variable characteristic of Haemophilus influenzae
Infect Immun 1997 65 943 950 9038301
Jennings MP Srikhanta YN Moxon ER Kramer M Poolman JT The genetic basis of the phase variation repertoire of lipopolysaccharide immunotypes in Neisseria meningitidis
Microbiology 1999 145 3013 3021 10589709
Hood DW Deadman ME Jennings MP Bisercic M Fleischmann RD DNA repeats identify novel virulence genes in Haemophilus influenzae
Proc Natl Acad Sci U S A 1996 93 11121 11125 8855319
Moran AP Appelmelk BJ Aspinall GO Molecular mimicry of host structures by lipopolysaccharides of Campylobacter and Helicobacter spp: Implications in pathogenesis J Endotoxin Res 1996 3 521 531
Guerry P Ewing CP Hickey TE Prendergast MM Moran AP Sialylation of lipooligosaccharide cores affects immunogenicity and serum resistance of Campylobacter jejuni
Infect Immun 2000 68 6656 6662 11083778
Bacon DJ Szymanski CM Burr DH Silver RP Alm RA A phase-variable capsule is involved in virulence of Campylobacter jejuni 81–176 Mol Microbiol 2001 40 769 777 11359581
Gilbert M Godschalk PCR Parker CT Endtz HP Wakarchuk WW Ketley J Konkel ME Genetic basis for the variation in the lipooligosaccharide outer core of Campylobacter jejuni and possible association of glycosyltransferase genes with post-infectious neurophathies Campylobacter jejuni : New perspectives in molecular and cellular biology 2004 Norfolk (United Kingdom) Horizon Scientific Press 219 248
Piddock LJ Ricci V Pumbwe L Everett MJ Griggs DJ Fluoroquinolone resistance in Campylobacter species from man and animals: Detection of mutations in topoisomerase genes J Antimicrob Chemother 2003 51 19 26 12493783
Taylor DE Ge Z Purych D Lo T Hiratsuka K Cloning and sequence analysis of two copies of a 23S rRNA gene from Helicobacter pylori and association of clarithromycin resistance with 23S rRNA mutations Antimicrob Agents Chemother 1997 41 2621 2628 9420030
Chopra I Roberts M Tetracycline antibiotics: Mode of action, applications, molecular biology, and epidemiology of bacterial resistance Microbiol Mol Biol Rev 2001 65 232 260 11381101
Trieber CA Taylor DE Mutations in the 16S rRNA genes of Helicobacter pylori mediate resistance to tetracycline J Bacteriol 2002 184 2131 2140 11914344
Gross CH Parsons JD Grossman TH Charifson PS Bellon S Active-site residues of Escherichia coli DNA gyrase required in coupling ATP hydrolysis to DNA supercoiling and amino acid substitutions leading to novobiocin resistance Antimicrob Agents Chemother 2003 47 1037 1046 12604539
Gibreel A Skold O Sulfonamide resistance in clinical isolates of Campylobacter jejuni Mutational changes in the chromosomal dihydropteroate synthase Antimicrob Agents Chemother 1999 43 2156 2160 10471557
Heep M Beck D Bayerdorffer E Lehn N Rifampin and rifabutin resistance mechanism in Helicobacter pylori
Antimicrob Agents Chemother 1999 43 1497 1499 10348780
Sutton GG White O Adams MD Kerlavage AR TIGR Assembler: A new tool for assembling large shotgun sequencing projects Genome Sequencing Tech 1995 1 9 19
Myers EW Sutton GG Delcher AL Dew IM Fasulo DP A whole-genome assembly of Drosophila
Science 2000 287 2196 2204 10731133
Biery MC Stewart FJ Stellwagen AE Raleigh EA Craig NL A simple in vitro Tn7-based transposition system with low target site selectivity for genome and gene analysis Nucleic Acids Res 2000 28 1067 1077 10666445
Delcher AL Phillippy A Carlton J Salzberg SL Fast algorithms for large-scale genome alignment and comparison Nucleic Acids Res 2002 30 2478 2483 12034836
Pop M Kosack DS Salzberg SL Hierarchical scaffolding with Bambus Genome Res 2004 14 149 159 14707177
Gajer P Schatz M Salzberg SL Automated correction of genome sequence errors Nucleic Acids Res 2004 32 562 569 14744981
Delcher AL Harmon D Kasif S White O Salzberg SL Improved microbial gene identification with GLIMMER Nucleic Acids Res 1999 27 4636 4641 10556321
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool J Mol Biol 1990 215 403 410 2231712
Tomb JF White O Kerlavage AR Clayton RA Sutton GG The complete genome sequence of the gastric pathogen Helicobacter pylori
Nature 1997 388 539 547 9252185
Alm RA Ling LS Moir DT King BL Brown ED Genomic-sequence comparison of two unrelated isolates of the human gastric pathogen Helicobacter pylori
Nature 1999 397 176 180 9923682
Baar C Eppinger M Raddatz G Simon J Lanz C Complete genome sequence and analysis of Wolinella succinogenes
Proc Natl Acad Sci U S A 2003 100 11690 11695 14500908
Dingle KE Colles FM Wareing DR Ure R Fox AJ Multilocus sequence typing system for Campylobacter jejuni
J Clin Microbiol 2001 39 14 23 11136741
Meinersmann RJ Helsel LO Fields PI Hiett KL Discrimination of Campylobacter jejuni isolates by fla gene sequencing J Clin Microbiol 1997 35 2810 2814 9350739
Felsenstein J PHYLIP, version 3.6 [computer progam]. Seattle: Department of Genome Sciences, University of Washington, Seattle. Available: http://evolution.genetics.washington.edu/phylip.html
2004 Accessed 16 November 2004
Heidelberg JF Seshadri R Haveman SA Hemme CL Paulsen IT The genome sequence of the anaerobic, sulfate-reducing bacterium Desulfovibrio vulgaris Hildenborough Nat Biotechnol 2004 22 554 559 15077118
Thompson JD Higgins DG Gibson TJ CLUSTALW: 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
Ewing B Hillier L Wendl MC Green P Base-calling of automated sequencer traces using phred. I. Accuracy assessment Genome Res 1998 8 175 185 9521921
| 15660156 | PMC539331 | CC BY | 2021-01-05 08:21:19 | no | PLoS Biol. 2005 Jan 4; 3(1):e15 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030015 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1566016310.1371/journal.pbio.0030016Research ArticleCell BiologyMicrobiologyBiochemistryEubacteriaFlux Analysis Uncovers Key Role of Functional Redundancy in Formaldehyde Metabolism Flux Analysis of Redundant ModulesMarx Christopher J
1
¤Van Dien Stephen J
2
Lidstrom Mary E [email protected]
1
3
1Department of Microbiology, University of WashingtonSeattle, WashingtonUnited States of America2United MetabolicsSeattle, WashingtonUnited States of America3Department of Chemical Engineering, University of WashingtonSeattle, WashingtonUnited States of AmericaMatthews Rowena G. Academic EditorUniversity of MichiganUnited States of America2 2005 4 1 2005 4 1 2005 3 2 e1619 7 2004 11 11 2004 Copyright: © 2005 Marx 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.
How Bacteria Stopped Worrying and Learned to Love...Formaldehyde
Genome-scale analysis of predicted metabolic pathways has revealed the common occurrence of apparent redundancy for specific functional units, or metabolic modules. In many cases, mutation analysis does not resolve function, and instead, direct experimental analysis of metabolic flux under changing conditions is necessary. In order to use genome sequences to build models of cellular function, it is important to define function for such apparently redundant systems. Here we describe direct flux measurements to determine the role of redundancy in three modules involved in formaldehyde assimilation and dissimilation in a bacterium growing on methanol. A combination of deuterium and 14C labeling was used to measure the flux through each of the branches of metabolism for growth on methanol during transitions into and out of methylotrophy. The cells were found to differentially partition formaldehyde among the three modules depending on the flux of methanol into the cell. A dynamic mathematical model demonstrated that the kinetic constants of the enzymes involved are sufficient to account for this phenomenon. We demonstrate the role of redundancy in formaldehyde metabolism and have uncovered a new paradigm for coping with toxic, high-flux metabolic intermediates: a dynamic, interconnected metabolic loop.
By feeding bacteria methanol labeled with deuterium, the relative contributions of two distinct pathways of formaldehyde metabolism have been demonstrated
==== Body
Introduction
The availability of large numbers of genome sequences has facilitated metabolic reconstruction based on predicted gene function, in essence, a prediction of the metabolic blueprint of a cell. Such metabolic reconstructions [1,2,3] can be grouped in functional segments, or metabolic modules [4,5], and the compilation of metabolic modules can be used to predict interactions between the different elements of the metabolic network in a cell. However, a major difficulty with this approach is the common occurrence of apparently redundant functional modules. It is often not possible to assign roles to these metabolic segments, which have been referred to as the “gray areas of the genome” [6]. Expression profiling, either of transcripts or proteins, holds the promise to gain more insight into the function of redundant metabolic modules, but the presence of a transcript or protein does not necessarily correlate with module function, due to posttranslational effects on metabolic flux. In order to determine the true function of such metabolic modules, it is necessary to measure the flux of metabolites through each functional module during relevant physiological changes.
One system that has proved amenable to a modular approach to metabolism is the ability to grow on one-carbon (C1) compounds, or methylotrophy [7]. The availability of a gapped genome sequence for a model methylotrophic bacterium, Methylobacterium extorquens AM1, has accelerated the definition of methylotrophy modules, and a reasonably complete metabolic reconstruction is available for this bacterium [7]. However, these analyses coupled to genetic and physiological studies [8,9,10,11,12,13] have raised a series of fundamental questions that can only be answered through direct flux measurements.
As in other such aerobic methylotrophic bacteria, M. extorquens AM1 oxidizes C1 substrates to formaldehyde and is essentially growing on formaldehyde for both carbon and energy metabolism [14] (Figure 1). It is not yet understood how the toxic central metabolite formaldehyde is efficiently and dynamically partitioned between assimilatory and dissimilatory metabolism, without toxic buildup. Therefore, this system represents both a key problem of methylotrophy and a paradigm for how toxic metabolites are managed in high-flux conditions. Genomic predictions and mutant analyses have identified three functional modules that direct formaldehyde into two outputs: assimilatory or dissimilatory metabolism (Figure 1). The first module consists of the apparently nonenzymatic condensation reaction between formaldehyde and tetrahydrofolate (H4F) [9,15] to generate methylene-H4F directly, which is the C1 donor for assimilation via the serine cycle. The second module is initiated by an enzyme-catalyzed reaction [9] of formaldehyde with a folate compound found in methanogenic Archaea, tetrahydromethanopterin (H4MPT). The resulting methylene-H4MPT is subsequently oxidized through a series of reactions to formate [8,16,17], which can ultimately be dissimilated to CO2 via the activity of multiple formate dehydrogenases [18]. Finally, a third module involves interconversion of methylene-H4F and formate via a familiar set of H4F-dependent reactions found in most organisms [11,19,20]. Mutant analysis has shown that both the H4MPT and H4F modules are required for growth on C1 compounds [8,9,10,11,12,13,19].
Figure 1 Formaldehyde Metabolism of M. extorquens AM1
Three modules work to provide two cellular outputs: formaldehyde assimilation and dissimilation. The direct condensation of formaldehyde with H4F is shown in green. A second proposed route for generating methylene-tetrahydrofolate (methylene-H4F), the consecutive action of the H4MPT and H4F modules is shown in blue. Fae, formaldehyde activating enzyme; Fch, methenyl H4F cyclohydrolase; FDH, formate dehydrogenase; Fhc, formyltransferase/hydrolase complex; FtfL, formyl H4F ligase; H4MPT, tetrahydromethanopterin; Mch, methenyl H4MPT cyclohydrolase; MDH, methanol dehydrogenase; MtdA, methylene H4F/H4MPT dehydrogenase; MtdB, methylene H4MPT dehydrogenase. Spontaneous and reversible reactions are indicated.
Two distinct models exist to explain the necessity of both the H4MPT and H4F modules in methylotrophy, predicting opposite directions for the net flux through the H4F module. It was suggested over 20 y ago that the H4F module functions in formaldehyde oxidation [21]. This predicts that the H4MPT and H4F modules are parallel, redundant formaldehyde oxidation systems. Recent genetic and biochemical evidence [11,12,13], however, suggest that the H4F module is not functionally redundant to the H4MPT module for formaldehyde oxidation. An alternative hypothesis suggests that the H4F module functions in the reductive direction, generating methylene-H4F from formate [11,16,17]. This model suggests a single dissimilatory module (H4MPT module) and two, redundant assimilatory modules: the H4F module and the direct condensation of methylene-H4F from formaldehyde (Figure 1, green arrows). This model predicts two routes for generating the key assimilatory intermediate methylene H4F from formaldehyde: one we will term “direct,” involving the direct condensation step, and one we will term “long,” involving the consecutive action of the H4MPT and H4F modules. Although the direct route (Figure 1, green arrows) requires flux through a nonenzymatic reaction, assimilation via the proposed long route (Figure 1, blue arrows) involving the action of the H4MPT and H4F modules is energetically costly due to a net expenditure of one ATP per C1 unit. If this hypothesis is correct, the H4MPT module would play a role in both dissimilatory and assimilatory metabolism, in much the same way that the tricarboxylic acid cycle plays a dual role in growth on multicarbon compounds.
Clearly, this is an example in which metabolic reconstruction is not sufficient to predict the roles of the central metabolic modules involved in carbon partitioning. In addition, it provides a test case for how cells cope with a high-flux toxic metabolic intermediate. In order to address this problem, we have used a combination of stable isotope- and radioisotope-labeling approaches, which has allowed the complete determination of flux through every branch of methylotrophy. The results provide a dynamic picture of the response of M. extorquens AM1 during transitions in and out of methylotrophy. Furthermore, a kinetic model of the key formaldehyde utilization systems was developed that successfully predicted key system dynamics. Our data resolve the specific roles for three interconnected metabolic modules that have two cellular outputs, assimilation and dissimilation. Furthermore, we have revealed a new paradigm for handling high-flux toxic intermediates: a dynamic metabolic loop that demonstrates graded response to changing metabolic needs.
Results
Detection of Serine-Derived Mass Fragments Using Gas Chromatography–Mass Spectrometry
A CD3OD label tracing strategy (Figure 2) was devised to directly determine what fraction of the methylene-H4F that entered the serine cycle was formed from the direct condensation of formaldehyde and H4F (direct route), versus the fraction formed through the alternative potential route involving oxidation of formaldehyde to formate by the H4MPT module, followed by assimilation through the H4F module (long route). The serine that is produced from methanol contains the carbon atom, and both hydrogens, from the methylene group of the methylene-H4F donor. Serine produced from CD3OD via the direct route contains two D, while that produced via the long route contains one D and in both cases these are relatively nonexchangeable C-D bonds. Therefore, at short labeling times (<1 min) the ratio of serine isotopomers with one or two D is an assay of the ratio of flux through the two routes.
Figure 2 GC–MS Method to Assay Ratio of Long Versus Direct Routes
(A) Simplified model of formaldehyde metabolism highlighting the deuterium (in red) label-tracing strategy. Oxidation of deuterated methanol (CD3OD) leads to the production of formaldehyde with two deuteriums (CD2O). Direct condensation with H4F (green arrows) and conversion to serine via the serine cycle (Figure 1) generates serine with two deuteriums. Alternatively, methylene-H4F may be produced through the long route (blue arrows; Figure 1), generating serine containing only one of the original deuteriums. Extraction and derivatizion of small molecules for analysis by GC–MS provides the ratio of (+1)/(+2) serine isotopomers, thereby assaying the proportion of methylene-H4F generated via the long route through formate or from the direct route from formaldehyde.
(B) Detection of serine by GC–MS. The small peak in total ion abundance detected by the MS denoted by the arrow represents serine.
(C) Analysis of the mass fragments present in this peak revealed the presence of ions with M/z values of 156 and 228, which are diagnostic for ECF–TFAA derivatized serine.
In order for this label tracing method to be successful, the ratio of serine isotopomers containing one or two deuteriums from CD3OD must be determined. Initially, cultures were labeled with standard methanol (CH3OH), added to boiling ethanol after labeling, and the derivatized H2O-soluble small molecules were prepared and analyzed via gas chromatography–mass spectrometry (GC–MS). Consistent with a derivatized serine standard and previous work [22,23], a peak was observed at approximately 8.6 min that contained two major ions with M/z of 156 and 228 (Figure 2B and 2C). The proportion of (+1) and (+2) M/z ions detected were within 1.1% ± 1.7% and −0.7% ± 0.5% of the predicted distribution (Isoform 1.02, National Institute of Standards and Technology) of naturally occurring heavy isotopomers for these fragments, indicating the feasibility of this GC–MS method for detecting serine isotopomers.
Deuterium Labeling Demonstrates Assimilation of C1 Units through Both Direct and Long Routes
Initially, the incorporation of deuteriums from CD3OD into serine was investigated with succinate-grown cell suspensions of wild-type M. extorquens AM1. Analysis of the derivatized H2O-soluble small molecule preparation from wild-type samples indicated a substantial increase in the proportion of fragments present as (+1) and (+2) isotopomers (>35% of total serine isotopomers). CD3OD labeling with a glyA mutant strain (CM239K.1), which lacks the initial serine-cycle enzyme, serine hydroxymethyltransferase, and was therefore completely unable to assimilate carbon from formaldehyde, produced no increase in (+1) or (+2) isotopomers (data not shown). Additionally, mutants defective for the proposed long route for methylene-H4F formation were tested for deuterium labeling. These included the ftfL (encodes formate-H4F ligase) mutant CM216K.1 [11], blocked for the H4F module, and the dmrA (encodes dihydromethanopterin reductase) mutant CM212K.1 [24], which has been shown to lack H4MPT [25,26]. Consistent with their proposed roles, the proportion of (+1) fragments dropped 8-fold for these mutants, compared to a modest 2-fold decrease in (+2) fragments. These data indicate that both the H4F and H4MPT modules affect labeling of serine and are required to generate the large increase in (+1) isotopomers seen with wild-type. These data also indicate that potential exchange reactions that could eliminate the deuteriums do not contribute measurably to the presence of (+1) ions. Collectively, these data indicate that the (+1) and (+2) serine mass fragments can serve as an accurate proxy for methylene-H4F generated through the long or direct routes. One caveat to this statement is that a portion of the NADPH involved in generating methylene H4MPT could be derived from the oxidation of methylene H4MPT to methenyl H4MPT and, therefore, could have become deuterium labeled. Based on the stoichiometry of the reactions and the known activity ratio of NADPH- versus NADH-producing enzymes for the methylene-H4MPT dehydrogenase reaction, we calculated that we at most overestimate the contribution of the direct pathway by 25% during growth on methanol, and by significantly smaller values at times with lower formaldehyde production. This prediction assumes an infinitely small intracellular concentration of NADPH, so depending on the actual pool of NADPH present, the error will be less. Therefore, our results are presented as maximum ratio changes.
When labeled with CD3OD, the succinate-grown wild-type cultures utilized to verify the GC–MS method produced a ratio of (+1) versus (+2) serine mass fragments of 8.0 ± 0.6. Thus, when succinate-grown cells are first exposed to methanol, the majority of methylene-H4F assimilated via the serine cycle is generated via the proposed long route. In contrast, CD3OD labeling of mid-exponential-phase methanol-grown cells indicated that the direct route dominated by up to 15-fold (measured ratio of [+1]/[+2] of 0.065 ± 0.006). Therefore, although both methylene-H4F production routes operated under both physiological conditions, a significant shift in the ratio of the two routes occurred, up to 100-fold.
Relative Contributions of the Long and Direct Routes of Methylene-H4F Formation during Transitions to and from Methylotrophic Growth
In order to understand the dynamics of the contribution of the long and direct routes for directing C1 units into assimilatory metabolism during transitions to and from methylotrophic growth, metabolic shift experiments were performed. One hour after samples were removed from succinate- and methanol-grown cultures for the labeling experiments described above, the remaining portions of the two cultures were harvested, washed, and resuspended into medium containing the other substrate (methanol or succinate, respectively). At four intervals during the transition to each of the new growth substrates (Figure 3) samples were harvested and analyzed via CD3OD labeling to determine the ratio of flux capacity through the two methylene-H4F formation routes. The ratio of the contribution of the long route for methylene-H4F formation to the direct route varied in a continuous fashion during the transition from succinate to methanol, or from methanol to succinate (Figure 3A). The cultures were followed for 7 or 10 h after the shift—sufficient time to observe the majority of the transition.
Figure 3 Change in Ratio of Flux through Long Versus Direct Methylene-H4F Formation Routes during Growth Transitions
(A) Experimental data as determined by GC–MS analysis of serine isotopomers. The bars for each transition represent a time series from cells harvested 1 h prior to the transition, and four time points following the transition (succinate to methanol: 1, 5, 7.5, and 10 h; methanol to succinate: 1, 3, 5, and 7 h).
(B) Predictions based on kinetic model simulations. The bars indicate the succinate to methanol transition (same time points as for the experimental data) and the methanol steady-state prediction.
Dynamics of C1 Fluxes during Transitions between Succinate and Methanol by 14C Labeling
The relative ratio of the routes provides only one of the parameters needed to understand the metabolic dynamics during this transition; the quantitative flux is also necessary. These values were obtained with 14C-labeling experiments. Concurrent with the CD3OD-labeling experiments described above, a portion of each sample was used to determine the rates of methanol oxidation, assimilation of C1 units, and CO2 production via 14C-CH3OH labeling [11]. Methanol oxidation was found to be 10-fold higher in methanol-grown cultures, and the percentage of carbon from methanol assimilated into biomass was 3-fold higher as compared to succinate-grown cultures (Table 1). The other values incorporated into the flux calculations are the stoichiometry of the serine cycle, in which two C1 units from methylene-H4F and one CO2 are incorporated for every C3 compound assimilated, and the proportion of external, unlabeled CO2 incorporated by the serine cycle [27]. The ten C1 fluxes (each branch arbitrarily labeled “A” through “J”) calculated using the concurrent CD3OD and 14C-methanol labeling methods are reported in Table 1 and shown in Figures 4 and 5.
Figure 4 C1 Fluxes during Transition from Succinate to Methanol
The fluxes determined are represented schematically (A). The other panels present flux for each branch, labeled A through J. The five bars for each flux represent a time series from cells harvested 1 h prior to the transition from succinate to methanol, and 1, 5, 7.5, and 10 h after the switch. Dissimilatory (B), methylene-H4F formation (C), and assimilatory (D) fluxes are presented separately with different scales for clarity. Flux F represents maximum fluxes.
Figure 5 C1 Fluxes during Transition from Methanol to Succinate
The fluxes determined are represented schematically (A). The other panels present flux for each branch, labeled A through J. The five bars for each flux represent a time series from cells harvested 1 h prior to the transition from methanol to succinate, and 1, 3, 5, and 7 h after the switch. Dissimilatory (B), methylene-H4F formation (C), and assimilatory (D) fluxes are presented separately with different scales for clarity. Flux F represents maximum fluxes.
Table 1 Calculated C1 Fluxes during Transitions between Succinate and Methanol at the Time (h) Relative to the Transition
All values are reported in nmol, min−1, mL−1, and OD600
−1
a First number represents flux for succinate to methanol; second number represents flux for methanol to succinate
A comparison of the values for succinate- versus methanol-grown cells shows that upon initial exposure of succinate-grown cells to methanol (Figure 4 and Table 1), the measurements suggest that most (at least 99%) of the formaldehyde was handled by the H4MPT module (flux B), and only a small amount flowed through the direct route (flux F). Of formate made from the H4MPT module (flux B), most (up to 88%) was converted to CO2 via formate oxidation (flux C), and a smaller amount (at least 12%) flowed through the H4F module and into assimilation (flux E), representing at least 90% of the assimilatory carbon. In contrast, for methanol-grown cells (Figure 5 and Table 1), less (only about 70%) of the formaldehyde generated from methanol flowed through the H4MPT module (flux B), with up to 30% handled by the direct route (flux F). Only a small portion of the assimilatory carbon (suggested to be about 6%) flowed through the H4F module (flux E), which represented about 3% of the formate generated via the H4MPT module. The remainder of the formate was oxidized to CO2 (flux C). These data indicate that, although the relative contribution of the long route to methylene-H4F formation decreased during the transition to growth on methanol (see Figure 3), the flux through the long route (flux E) increased significantly (see Figure 4). Flux through this route peaked 5 h after the transition to methanol, when it reached a value at least 8-fold higher than succinate-grown cells, and dropped somewhat afterward. The flux through the direct route (flux F) also increased to a maximum of up to 20% of the total formaldehyde flux at the final time point during the transition (see Figure 4). The fluxes for the transition from methanol to succinate represent the capacity for flux, as no methanol was present after the growth transitions. These changes, however, roughly mirrored the transition from succinate to methanol, but were not an exact reversal (see Figure 5). As noted for the deuterium-labeling experiments, the time periods followed in these experiments were sufficient to observe the majority of the transition.
Dynamic Mathematical Model of Formaldehyde Partitioning
In order to assess whether the known kinetic constraints of the three modules of formaldehyde metabolism were sufficient to account for the experimentally determined flux dynamics, a mathematical model was generated. The model simulated partitioning of C1 units through the three formaldehyde modules during growth of cells in methanol, and for the transition of succinate-grown cells to methanol. The model consisted of eight ordinary differential equations, based on known kinetic mechanisms, to describe the dynamics of the H4F and H4MPT modules and the direct condensation reaction. Most binding constants, rate constants, and cofactor concentrations were obtained from the literature (Table 2). For the six cases in which literature values are not known, these were estimated as described in Materials and Methods. Additionally, a dynamic simulation of the succinate to methanol transition was performed. The methanol uptake rate was set to the experimentally measured value at each time point (flux A, Table 1) and interpolated linearly between time points to create a smooth gradient. Starting with the values obtained for succinate or methanol growth, the parameters were increased throughout the shift at a rate corresponding to the increase in methanol uptake.
Table 2 Equilibrium Constants and Forward Rate Constants (Vmax) for Each Reaction in the Model Simulation
Equilibrium constants are all dimensionless, except for reaction 9, which has units of mM. Units for kinetic constants are mM/sec unless otherwise noted
a The literature value for this constant is 0.71 mM/s. A small adjustment was required to fit the data
b Constants of twice the literature values were assumed, due to the presence of multiple formate dehydrogenases
H4F, tetrahydrofolate; H4MPT, tetrahydromethanopterin; me-H4F, methylene-H4F; me-H4MPT, methylene-H4MPT; MFR, methanofuran; mn-H4F, methenyl-H4F; mn-H4MPT, methenyl-H4MPT; Irrev., irreversible; NA, not applicable
Two key results are apparent from the comparison of the model's predictions (see Figure 3B) to the measured flux ratio of the two methylene-H4F production routes (see Figure 3A). First, the model did not constrain the direction of flux through the H4F module. Therefore the prediction that the H4F module functions in assimilation both during steady-state methanol growth and upon the first exposure of succinate-grown cells to methanol indicates that the kinetic parameters of the module components are sufficient to account for this phenomenon. Second, the correspondence between the predicted and experimentally determined dynamics of the switch in methylene-H4F production routes confirms that the dynamics of the system are also largely attributable to the systems' kinetic constraints. That the kinetics did not exactly mimic the measured values is presumably partly due to differences between the actual induction of enzyme activities versus the model's simplifying assumption that all values change in a manner directly proportional to changes in methanol uptake. However, the model does not suggest a significant effect of methylene H4MPT-derived NADPD in the deuterium-labeling studies.
The H4F Module Could Not Be Eliminated during Growth on C1 Compounds
The combination of CD3OD and 14C-methanol label-tracing studies clearly demonstrate that the long route contributes methylene-H4F to the serine cycle and that the flux through the H4F module portion of the long route (flux E) increases significantly during the transition to growth on methanol. These results confirm the hypothesis of net reductive flux through this module [11,16,17]. However, this route contributes only 6% of the total methylene-H4F generated during growth on methanol. Therefore, it seemed possible that the H4F module might be required during transitions in and out of methylotrophy, but might not be required for continuous growth on methanol. Given the available genetic techniques, two strategies were employed in an attempt to obtain mutants in one of the key H4F module genes, formate-H4F ligase, during growth on C1 compounds. First, attempts were made to obtain null mutants via allelic exchange with cultures maintained on methanol or methylamine, but these efforts were unsuccessful. Second, cultures of the ΔftfL::kan mutant CM216K.1 [11] bearing the complementing plasmid pCM218 [11] were grown in medium containing methanol or methylamine without tetracycline for plasmid maintenance. No plasmid-free isolates were obtained for CM216K.1 with pCM218 during growth on methanol. However, they were obtained for wild-type with pCM218 on methanol, or CM216K.1 with pCM218 grown on succinate. Therefore, it appears that the H4F module plays an essential role in methylotrophy even after cells have already begun to grow on C1 compounds.
Discussion
In the formaldehyde metabolism of M. extorquens AM1, three interconnected metabolic modules are present, involved in two roles: converting formaldehyde to the key assimilatory intermediate methylene H4F and net oxidation of formaldehyde to CO2. Understanding paradigms for differential roles of redundant modules is central to enabling broadscale metabolic reconstruction from genome sequences. In addition, methylotrophy represents an intriguing example of a metabolic mode in which growth depends on high flux of a toxic metabolite, with subsequent partitioning of that metabolite. Other such modes are known that produce toxic aldehydes, for instance, growth on ethanolamine [28] and other alcohols [29]. Numerous other toxic intermediates are known in bacteria, such as the production of hydroxylamine by ammonia-oxidizing bacteria [30] and mono-oxygenase-dependent production of epoxyalkanes during growth on aliphatic alkanes [31]. In addition, the liver can be exposed to toxic metabolites, for instance, the production of formate from acute methanol poisoning [32]. However, the metabolic mechanisms that allow the balancing of flux and toxicity in such situations are not well understood. Understanding paradigms for such metabolic responses is important for assessing and possibly ameliorating toxicity problems in a variety of systems, including bioremediation of toxic compounds, chemical production in bioprocesses, and detoxification in tissues and organs.
Through a combination of 14C and deuterium label-tracing strategies, we have defined flux through each metabolic module in methylotrophic metabolism in M. extorquens AM1 during transitions into and out of methylotrophy, in which the flux of formaldehyde into the system changed by a factor of 10. These methods had the dual advantages of possessing sufficient sensitivity to detect flux under all conditions tested, and being free from the requirement of steady-state growth conditions, which allowed the dynamics of growth transitions to be examined. Furthermore, this approach complements a recently developed 13C-labeling method that measures flux through the multicarbon branches of central metabolism [27], but is inherently silent to the C1 fluxes measured here. The approach described here allowed us to test and confirm the hypothesis that the role of the H4F module during growth on C1 compounds is to supply methylene-H4F from formate [11,16,17], although the fraction of total flux passing through this route is always small.
Given the small percentage of total flux into assimilation via the H4F module during growth on methanol, why is this module required under this condition? The results presented here suggest that this requirement is not alleviated even when cells begin to actively grow on methanol. It is possible that this module generates an inducing signal for the serine cycle and, therefore, is necessary to maintain assimilatory flux during growth on methanol. This hypothesis is consistent with the genetic circuit, as two of the genes encoding key enzymes of the H4F module (mtdA and fch) are in an operon with serine-cycle genes and are under the control of a single regulatory protein, QscR [33].
Our results demonstrate a dramatic shift in flux through the primary methylotrophic modules during these transitions. It has long been known that all enzymes of methylotrophy increase 3–6 fold in activity after induction with methanol [14,16], predicting a sizable increase in total flux into the system. However, the flux measurements reported here show that a dynamic repartitioning occurs also. When M. extorquens AM1 encounters methanol, the methanol oxidation system is at low but significant activity [34]. Under these conditions, the flux of formaldehyde into the system is relatively low (Figure 6, left panel), and most of the formaldehyde is oxidized to CO2 via the H4MPT module and formate dehydrogenase, generating NAD(P)H. Only a trace amount is assimilated, almost all of that through the long route involving formate and H4F intermediates. As the flux of formaldehyde into the system increases, a greater percentage begins to flow through the direct route into assimilatory metabolism. A smooth transition occurs during the induction of the capacity in the system until approximately one-third of the total formaldehyde flows through this route, and assimilatory and dissimilatory metabolism are balanced for rapid growth on methanol (Figure 6, right panel). The metabolic elegance of this interconnected, dynamic metabolic loop creates an effective formaldehyde flux buffer for transitions, in which the cell has time to respond to the presence of a methylotrophic substrate, deriving benefit (energy) without risking buildup of a toxic intermediate. As the activity of the serine cycle begins to increase, more formaldehyde can be safely shunted to assimilatory metabolism via the direct, ATP-independent route, thereby ensuring the transition to growth on the C1 substrate without build up of formaldehyde.
Figure 6 An Interconnected Metabolic Loop for Handling the Toxic Intermediate Formaldehyde
A dynamic transition occurs from low to high formaldehyde flux, shifting the ratio of the direct versus long routes, and in the relative proportion of carbon oxidized to CO2 versus assimilated, creating a buffer system to accommodate large changes in formaldehyde flux.
What controls the rate of the nonenzymatic condensation of formaldehyde with H4F to form methylene-H4F, which was up to 150-fold greater during methanol growth than on succinate? The rate of this spontaneous reaction will be determined by the relative concentrations of reactants and products, with an equilibrium constant for this condensation of 3.2 × 10−4 [15]. Although this equilibrium constant favors the production of methylene-H4F, flux will only occur if either the concentrations of the reactants (formaldehyde and/or H4F) rise above the equilibrium concentration, or utilization of methylene-H4F is sufficient to keep the pool of this metabolite below the equilibrium concentration. At this time, it is not technically feasible to measure the intracellular concentrations of free formaldehyde or methylene-H4F. However, the most likely explanation for high flux through the nonenzymatic condensation of formaldehyde and H4F would be draw-off of the product (methylene-H4F) by the serine cycle. In order to test whether the known kinetic parameters explain the relative utilization of the two methylene-H4F production routes, a kinetic model was constructed and utilized to simulate formaldehyde partitioning during transitions to and from methylotrophic growth. The ability of the model to recapitulate the observed switch in route utilization (see Figure 3B) indicates that the architecture of the dynamic loop and the kinetic parameters of the responsible enzymes can predict operation of the H4F module in the assimilatory direction and are sufficient to account for partitioning of C1 units into assimilatory metabolism without accumulation of formaldehyde.
In summary, the dual-labeling approach described here for direct flux measurement during metabolic transitions has not only elucidated a key role for redundancy in the three metabolic modules responsible for formaldehyde assimilation and dissimilation, but has also revealed a new paradigm for accommodating high-flux toxic intermediates. It is likely that similar interconnected loop systems operate for other metabolites, toxic or not, and this example can now be used as a framework for predicting functions of other apparently redundant modules that may be involved in the handling of toxic metabolites.
Materials and Methods
Bacterial strains
Wild-type M. extorquens AM1 [35] and mutant strains were cultured at 30 °C in a minimal salts medium [36] containing 125 mM methanol or 15 mM succinate. A serine hydroxymethyltransferase mutant strain, CM239K.1 (ΔglyA::kan) was generated using the allelic exchange technique described previously [37].
CD3OD labeling and GC–MS
CD3OD (99.8%; Cambridge Isotope Laboratories, Andover, Massachusetts, United States) to a final concentration of 1 mM was added to washed cultures that had been resuspended to an OD600 = 1 in order to label cell metabolites with deuterium for analysis by GC–MS. After shaking for 20 s at room temperature the 2-ml suspension was added to three volumes of boiling 100% ethanol for instant lysis. Following centrifugation, the soluble fraction was dried, resuspended in distilled H2O, and centrifuged again to remove H2O-insoluble components. The resulting H2O-soluble small molecule fraction was then derivatized with ethyl chloroformate and trifluoroacetic acid as previously described [22,23]. All labeling experiments were performed three times.
GC–MS methods and data analysis
GC–MS experiments were performed using an Agilent 6890 gas chromatograph/Agilent 5973 quadrupole mass selective detector (electron impact ionization) operated at 70 eV equipped with an Agilent 7683 autosampler/injector (Hewlett-Packard, Palo Alto, California, United States). The MS was operated in selected ion monitoring mode to detect M/z = 156/157/158/228/229/230 from 7 min to the end of the method. The GC oven temperature started at an initial temperature of 60 °C, ramping at 20 °C min−1 to 130 °C, 4 °C min−1 to 155 °C, and then 120 °C min−1 to a final temperature of 300 °C that was held for 5 min. Flow through the column was held constant at 1 ml min−1. The injection volume was 1 μl and the machine was run in splitless mode. The temperature of the inlet was 230 °C, the interface temperature was 270 °C, and the quadrupole temperature was 150 °C. The column utilized was an HP-5MS (Hewlett-Packard).
GC–MS data were analyzed using Agilent Enhanced ChemStation G1701CA (Hewlett-Packard). The two mass clusters for serine, M/z = 156/157/158, and 228/229/230, represent fragments of ECF–TFAA derivatized serine (C10H14O6NF3) that have lost one or both of the carboxyl ethyl esters. The data were corrected for the natural abundance of heavy isotopes in the derivatized serine fragments, using proportions calculated with Isoform 1.02 (MS Search Program for Windows, National Institute of Standards and Technology, Gaithersburg, Maryland, United States). For each sample, the ratio of Δ + 1)/Δ + 2) was calculated for both mass clusters and averaged. The mean and standard error for these data were then calculated for the three replicates of each experiment.
Assimilation and CO2 production rates
The rate of 14C-CO2 production and assimilation of labeled carbon from 14C-methanol was determined concurrently with the CD3OD labeling described above using a modification of a previously described method [11]. A portion of the labeled cell suspensions was filtered (0.2 μM PVDF, Millipore, Billerica, Massachusetts, United States) to determine net assimilation. All measured and calculated fluxes were determined using the data from each of the three replicate experiments and then utilized to determine the mean and standard error for each flux.
Additional values incorporated into flux calculations
It has been determined previously that 63.3% of the total CO2 incorporated originates directly from CO2 produced from the oxidation of methanol [27]. This value cannot be determined under the nonsteady state conditions used in the experiments described here, so this value was incorporated directly into our calculations. The sensitivity of the calculated fluxes to a 2-fold increase or decrease in the determined ratio of 1.73:1.00 internal:external CO2 incorporated into the serine cycle was examined. Besides the direct effect on relative fluxes of internal and external CO2 into the serine cycle, the calculated incorporation of C1 units from methylene-H4F would vary no more than 7%, which would be balanced by a change in the dissimilatory flux through the H4MPT module and formate dehydrogenase of less than 6%. Therefore, deviations in the ratio of methanol-derived and external CO2 incorporation from the reported work [27] would not significantly alter the calculated fluxes.
Dynamic model
The dynamic model of the formaldehyde oxidation and assimilation modules consisted of eight ordinary differential equations, each describing the accumulation of a metabolite involved in the H4F and H4MPT modules. These equations were derived in a straightforward manner from the kinetic expressions given below. The production of formaldehyde from methanol was set to the measured rate of methanol uptake for each experiment. All enzymatic reactions were treated with either uni- or bimolecular reversible Michaelis–Menten kinetics, with the equilibrium constants taken from the literature [38]. In cases where Keq > 200, the reverse reaction was ignored for simplicity. Finally, since the dynamics of serine and glycine were not included in this model, serine hydroxymethyltransferase was modeled as an irreversible unimolecular Michaelis–Menten reaction, with the effects of all metabolites other than methylene-H4F accounted for in an effective Vmax. The total internal concentrations of H4F and H4MPT derivatives were set equal to 0.15 and 0.4 mM, respectively [38]. Concentrations of other energy and redox cofactors (ATP, NADH, etc.) were assumed equal to those present in Escherichia coli [39]. The parameters used in the simulation are listed in Table 2. All Kms could be obtained from the literature (see Table 2), except for that of reaction 5. This Km was set arbitrarily to 50 μM, which results in the reaction proceeding at half-maximal rate. Many of the values for Vmax could be directly calculated from specific activities found in the literature, for both growth on methanol and succinate. To allow for experimental error in the measured rate constants, and to account for the fact that kinetics measured in vitro do not necessarily correlate exactly with what occurs inside the cell, these values were allowed to vary within 50% during the fitting procedure described below. For the remaining parameters, a numerical error minimization technique was used to find the set of parameters yielding model predictions with the best fit to the experimental flux distributions, when integrated to steady state. This was first done for methanol growth, then repeated for succinate growth. The rate constant for spontaneous formaldehyde condensation (k6) was forced to be the same on succinate as on methanol, since this is a fundamental chemical property that is not affected by gene induction. All reverse rate constants were calculated directly from the forward constants, binding constants, and Keq. The spontaneous condensation of formaldehyde with H4MPT was assumed to be negligible under physiological conditions compared to the formaldehyde activating enzyme reaction [9]. All simulations were performed in MATLAB 6.5 (MathWorks, Natick, Massachusetts, United States) using the ODE solving function “ode15s.” The error minimization was also done in MATLAB, using an evolutionary algorithm written previously [27].
Abbreviations as in Table 2.
Supporting Information
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession numbers for genes discussed in this paper are dmrA (AY093431), ftfL (AY279316), and glyA (L33463).
We would like to thank L. Chistoserdova, M. Kalyuzhnaya, N. Korotkova, H. Rothfuss, S. Stolyar, R. Thauer, and J. Vorholt for their thoughtful discussion of our work, M. Sadilek for his invaluable assistance in developing the GC–MS method, and anonymous reviewers for helpful comments. This work was supported by a grant from the National Institutes of Health (GM 36296).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. CJM, SJVD, and MEL conceived and designed the experiments. CJM and SJVD performed the experiments. CJM, SJVD, and MEL analyzed the data. CJM and SJVD contributed reagents/materials/analysis tools. CJM, SJVD, and MEL wrote the paper.
¤Current address: Department of Microbiology and Molecular Genetics, Michigan State University, East Lansing, Michigan, United States of America
Citation: Marx CJ, Van Dien SJ, Lidstrom ME (2005) Flux analysis uncovers key role of functional redundancy in formaldehyde metabolism. PLoS Biol 3(1): e16.
Abbreviations
GC–MSgas chromatography–mass spectrometry
H4Ftetrahydrofolate
H4MPTtetrahydromethanopterin
==== Refs
References
Forster J Famili I Fu P Palsson BO Nielsen J Genome-scale reconstruction of the Saccharomyces cerevisiae metabolic network Genome Res 2003 13 244 253 12566402
Van Dien SJ Lidstrom ME Stoichiometric model for evaluating the metabolic capabilities of the facultative methylotroph Methylobacterium extorquens AM1, with application to reconstruction of C(3) and C(4) metabolism Biotechnol Bioeng 2002 78 296 312 11920446
Reed JL Vo TD Schilling CH Palsson BO An expanded genome-scale model of Escherichia coli K–12 (iJR904 GSM/GPR) Genome Biol 2003 4 R54 12952533
Hartwell LH Hopfield JJ Leibler S Murray AW From molecular to modular cell biology Nature 1999 402 C47 C52 10591225
Wolf DM Arkin AP Motifs, modules and games in bacteria Curr Opin Microbiol 2003 6 125 134 12732301
Aslund F Beckwith J The thioredoxin superfamily: Redundancy, specificity, and gray-area genomics J Bacteriol 1999 181 1375 1379 10049365
Chistoserdova L Chen SW Lapidus A Lidstrom ME Methylotrophy in Methylobacterium extorquens AM1 from a genomic point of view J Bacteriol 2003 185 2980 2987 12730156
Chistoserdova L Vorholt JA Thauer RK Lidstrom ME C1 transfer enzymes and coenzymes linking methylotrophic bacteria and methanogenic Archaea Science 1998 281 99 102 9651254
Vorholt JA Marx CJ Lidstrom ME Thauer RK Novel formaldehyde-activating enzyme in Methylobacterium extorquens AM1 required for growth on methanol J Bacteriol 2000 182 6645 6650 11073907
Hagemeier CH Chistoserdova L Lidstrom ME Thauer RK Vorholt JA Characterization of a second methylene tetrahydromethanopterin dehydrogenase from Methylobacterium extorquens AM1 Eur J Biochem 2000 267 3762 3769 10848995
Marx CJ Laukel M Vorholt JA Lidstrom ME Purification of the formate-tetrahydrofolate ligase from Methylobacterium extorquens AM1 and demonstration of its requirement for methylotrophic growth J Bacteriol 2003 185 7169 7175 14645277
Marx CJ Chistoserdova L Lidstrom ME Formaldehyde-detoxifying role of the tetrahydromethanopterin-linked pathway in Methylobacterium extorquens AM1 J Bacteriol 2003 185 7160 7168 14645276
Marx CJ Lidstrom ME Development of an insertional expression vector system for Methylobacterium extorquens AM1 and generation of null mutants lacking mtdA and/or fch
Microbiology 2004 150 9 19 14702393
Lidstrom ME Aerobic methylotrophic prokaryotes. In: Dworkin, M., editor, The prokaryotes, 3rd edition, release 3.7. New York: Springer-Verlag. Available: http://141.150.157.117:8080/prokPUB/index.htm
2001 Accessed 26 November 2004
Kallen RG Jencks WP The mechanism of the condensation of formaldehyde with tetrahydrofolic acid J Biol Chem 1966 241 5851 5863 5954363
Vorholt JA Cofactor-dependent pathways of formaldehyde oxidation in methylotrophic bacteria Arch Microbiol 2002 178 239 249 12209256
Pomper BK Saurel O Milon A Vorholt JA Generation of formate by the formyltransferase/hydrolase complex (Fhc) from Methylobacterium extorquens AM1 FEBS Lett 2002 523 133 137 12123819
Chistoserdova L Laukel M Portais JC Vorholt JA Lidstrom ME Multiple formate dehydrogenase enzymes in the facultative methylotroph Methylobacterium extorquens AM1 are dispensable for growth on methanol J Bacteriol 2004 186 22 28 14679220
Chistoserdova LV Lidstrom ME Genetics of the serine cycle in Methylobacterium extorquens AM1: Identification of sgaA and mtdA and sequences of sgaA, hprA and mtdA
J Bacteriol 1994 176 1957 1968 8144463
Pomper BK Vorholt JA Chistoserdova L Lidstrom ME Thauer RK A methenyl tetrahydromethanopterin cyclohydrolase and a methenyl tetrahydrofolate cyclohydrolase in Methylobacterium extorquens AM1 Eur J Biochem 1999 261 475 480 10215859
Attwood MM Quayle JR Crawford RL Hanson RS Formaldehyde as a central intermediary metabolite of methylotrophic metabolism Microbial growth on C1 compounds 1984 Washington, DC American Society for Microbiology 315 323
Husek P Amino acid derivatization and analysis in five minutes FEBS Lett 1991 280 354 356 2013337
Christensen B Nielsen J Isotopomer analysis using GC–MS Metab Eng 1999 1 282 290 10937821
Marx CJ O'Brien BN Breezee J Lidstrom ME Novel methylotrophy genes of Methylobacterium extorquens AM1 identified by using transposon mutagenesis including a putative dihydromethanopterin reductase J Bacteriol 2003 185 669 673 12511515
Caccamo MA Malone CS Rasche ME Biochemical characterization of a dihydromethanopterin reductase involved in tetrahydromethanopterin biosynthesis in Methylobacterium extorquens AM1 J Bacteriol 2004 186 2068 2073 15028691
Rasche ME Havemann SA Rosenzvaig M Characterization of two methanopterin biosynthesis mutants of Methylobacterium extorquens AM1 by use of a tetrahydromethanopterin bioassay J Bacteriol 2004 186 1565 1570 14973120
Van Dien SJ Strovas T Lidstom ME Quantification of central metabolic fluxes in the facultative methylotroph Methylobacterium extorquens AM1 using 13 C-label tracing and mass spectrometry Biotechnol Bioeng 2003 84 45 55 12910542
Roof DM Roth JR Ethanolamine utilization in Salmonella typhimurium
J Bacteriol 1988 170 3855 3863 3045078
Reid MF Fewson CA Molecular characterization of microbial alcohol dehydrogenases Crit Rev Microbiol 1994 20 13 56 8185833
Prosser JI Autotrophic nitrification in bacteria Adv Microb Physiol 1989 30 125 181 2700538
Ensign SA Microbial metabolism of aliphatic alkenes Biochemistry 2001 40 5845 5853 11352719
Barceloux DG Bond GR Krenzelok EP Cooper H Vale JA American Academy of Clinical Toxicology practice guidelines on the treatment of methanol poisoning J Toxicol Clin Toxicol 2002 40 415 446 12216995
Kalyuzhnaya MG Lidstrom ME QscR, a LysR-type transcriptional regulator and CbbR homolog, is involved in regulation of the serine cycle genes in Methylobacterium extorquens AM1 J Bacteriol 2003 185 1229 1235 12562792
Anthony C The oxidation of methanol in gram-negative bacteria FEMS Microbiol Rev 1990 7 209 214 1965566
Peel D Quayle JR Microbial growth on C1 compounds: 1. Isolation and characterization of Pseudomonas AM1 Biochem J 1961 81 465 469 14484877
Attwood MM Harder W A rapid and specific enrichment procedure for Hyphomicrobium spp Antonie Van Leeuwenhoek 1972 38 369 377 4561897
Marx CJ Lidstrom ME Broad-host-range cre-lox system for antibiotic marker recycling in gram-negative bacteria Biotechniques 2002 33 1062 1067 12449384
Vorholt JA Chistoserdova L Lidstrom ME Thauer RK The NADP-dependent methylene tetrahydromethanopterin dehydrogenase in Methylobacterium extorquens AM1 J Bacteriol 1998 180 5351 5356 9765566
Neidhardt FC Ingraham JL Schaechter M Physiology of the bacterial cell 1990 Sunderland (Massachusetts) Sinauer 506
| 15660163 | PMC539335 | CC BY | 2021-01-05 08:21:18 | no | PLoS Biol. 2005 Feb 4; 3(2):e16 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030016 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030020SynopsisEcologyEvolutionZoologyBirdsAncient DNA Tells Story of Giant Eagle Evolution Synopsis1 2005 4 1 2005 4 1 2005 3 1 e20Copyright: © 2005 Public Library of Science.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.
Ancient DNA Provides New Insights into the Evolutionary History of New Zealand's Extinct Giant Eagle
==== Body
The recent discovery of a Hobbit-like hominid on the Indonesian island of Flores was startling in some respects—its rather modern existence, for one—but it represents a classic case of Darwinian evolution. For reasons that are not entirely clear, when animals make their way to isolated islands, they tend to evolve relatively quickly toward an outsized or pint-sized version of their mainland counterpart. Following this evolutionary script, the Flores woman, presumably a downsized version of Homo erectus, appears to have shared her island home with dwarf elephants and giant rats.
Perhaps the most famous example of an island giant—and, sadly, of species extinction—is the dodo, once found on the Indian Ocean island of Mauritius. When the dodo's ancestor (thought to be a migratory pigeon) settled on this island with abundant food, no competition from terrestrial mammals, and no predators, it could survive without flying, and thus was freed from the energetic and size constraints of flight. New Zealand also had avian giants, now extinct, including the flightless moa, an ostrich-like bird, and Haast's eagle (Harpagornis moorei), which had a wingspan up to 3 meters. Though Haast's eagle could fly—and presumably used its wings to launch brutal attacks on the hapless moa—its body mass (10–14 kilograms) pushed the limits for self-propelled flight.
As extreme evolutionary examples, these island birds can offer insights into the forces and events shaping evolutionary change. In a new study, Michael Bunce et al. compared ancient mitochondrial DNA extracted from Haast's eagle bones with DNA sequences of 16 living eagle species to better characterize the evolutionary history of the extinct giant raptor. Their results suggest the extinct raptor underwent a rapid evolutionary transformation that belies its kinship to some of the world's smallest eagle species.
Giant Haast's eagle attacking New Zealand moa (Art: John Megahan)
The authors characterized the rates of sequence evolution within mitochondrial DNA to establish the evolutionary relationships between the different eagle species. Their analysis places Haast's eagle in the same evolutionary lineage as a group of small eagle species in the genus Hieraaetus. Surprisingly, the genetic distance separating the giant eagle and its more diminutive Hieraaetus cousins from their last common ancestor is relatively small.
Without the fossils to directly determine divergence times, Bunce et al. relied on molecular dating techniques that use the rate of sequence evolution in the genes studied to establish the relative evolutionary ages of the eagles. Proposing a divergence date of roughly 0.7–1.8 million years ago, the authors acknowledge that while this is the “best available approximation of the ‘true’ date,” additional molecular data could help refine the estimate. Whatever the date of divergence, the extinct giant eagle is clearly an anomaly among the eagles studied here. The increase in body size—by at least an order of magnitude in less than 2 million years—is particularly remarkable, Bunce et al. argue, since it occurred in a species still capable of flight.
The absence of mammalian competitors facilitated the evolution of much larger eagles and owls on Cuba and may have likewise precipitated the rapid morphological shift seen here. Haast's eagle, the authors write, “represents an extreme example of how freedom from competition on island ecosystems can rapidly influence morphological adaptation and speciation.” Given its similarity to the smaller Hieraaetus species, the authors recommend reclassifying the New Zealand giant as Hieraaetus moorei. This study shows how quickly morphological changes can occur in vertebrate lineages within island ecosystems. Could it be that anthropologists might some day uncover evidence of a giant version of the Flores woman?
| 0 | PMC539337 | CC BY | 2021-01-05 08:21:18 | no | PLoS Biol. 2005 Jan 4; 3(1):e20 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030020 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030025SynopsisBioinformatics/Computational BiologyEvolutionGenetics/Genomics/Gene TherapyNoneGenome Sequencing: Using Models to Predict Who's Next Synopsis1 2005 4 1 2005 4 1 2005 3 1 e25Copyright: © 2005 Public Library of Science.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.
A Model of the Statistical Power of Comparative Genome Sequence Analysis
==== Body
It's hard to believe it was just ten years ago that scientists reported the first complete genome sequence of an organism, the bacterial pathogen Haemophilus influenzae. The list has grown considerably since then: add over 160 bacterial species (and counting), most major model organisms, and an ever-growing list of mammals—including, of course, humans. With 99% of our genome now fully sequenced, the Human Genome Project's next major goal is to identify all the functional elements contained in our 2.85 billion nucleotides. Such an effort is hardly trivial: producing the sequence of a mammalian-size genome can run from $10 to $50 million, the estimated price tag of the Cow Genome Project.
In an ideal world, any organism would be fair game for sequencing, but in the real world, sequencing resources are scarce. Comparing genome sequences turns out to be a great way to identify regions that have important functions, but comparative genomics studies would be far more efficient if scientists could figure out in advance which genomes would reveal the most information about a particular question. Taking up that challenge, computational biologist Sean Eddy reports a statistical model that predicts how many genomes, and at what evolutionary distance, are needed for effective comparative genomic analyses. In addition to confirming some working principles of comparative genomics, the model also reveals a surprisingly simple guideline for future studies.
Comparative genomics works by aligning sequences of different organisms to identify patterns that operate over both large and small distances. Aligning mouse chromosomes with human chromosomes, for example, shows that 99% of our protein-coding genes align with homologous sequences in mice. Underlying such analyses is the principle that DNA sequences that are highly conserved are likely to be functionally important. A common assumption is that adding more comparative genomes to the alignment helps distinguish functionally significant from irrelevant conserved sequences.
How do you go about creating an abstract model that captures what Eddy calls the “essential flavor of comparative genomic analysis”? His model puts aside the specific characteristics of individual organisms, genomic features, and analysis programs in favor of identifying higher-level patterns and scaling relationships, specifically between the number of genomes, evolutionary distance, and feature size (features include genetic elements like exons and transcription factors).
The model shows that the number of genomes required to identify conserved regions—that is, regions evolving under selection—scales inversely with the size of the feature being sought. Thus, to look for conserved sequences half as long, you need twice as many genomes, assuming a constant evolutionary distance and statistical power. For example, to identify a conserved human feature the size of a coding exon (about 50 nucleotides), it is sufficient to compare just the human and mouse genomes. But to identify conserved single nucleotides, you would need 55 comparative genomes at “mouse-like” evolutionary distances (roughly 75 million years).
Things get a little trickier when varying evolutionary distance. We can see a substitution only at a given point in time: we can't tell how many times a site has changed, for example, or whether it changed at some point and then changed back. But at short evolutionary distances—where it's safer to assume no sites have changed more than once—the evolutionary distance is roughly the same as the fraction of sites identified as changed, and evolutionary distance and the number of genomes needed scale inversely. Therefore, the closer the evolutionary distance, the more genomes needed: one would need seven times as many comparative genomes using human/baboon distances, for example, compared to human/mouse distances. So when it comes to using primate sequences to study the human genome, our most distant relatives (such as lemurs) offer far more comparative analysis power than our next of kin (chimps and bonobos).
While this model confirms the intuitive assumption that identifying smaller features requires more genomes, it reveals an inverse scaling relationship far more direct, and precise, than previously imagined. With the next phase of the Human Genome Project under way, Eddy's model offers valuable guidelines for identifying which genomes and how many might best meet this ambitious goal.
| 0 | PMC539338 | CC BY | 2021-01-05 08:21:19 | no | PLoS Biol. 2005 Jan 4; 3(1):e25 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030025 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030026SynopsisBioinformatics/Computational BiologyBiophysicsNeuroscienceInsectsComputation Provides a Virtual Recording of Auditory Signaling Synopsis1 2005 4 1 2005 4 1 2005 3 1 e26Copyright: © 2005 Public Library of Science.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.
Disentangling Sub-Millisecond Processes within an Auditory Transduction Chain
==== Body
A small rodent rustles through a field in the still night, making just enough noise to betray its location to a circling barn owl. A female frog sits on the bank of a pond amid a cacophony of courting bullfrogs, immune to the mating calls of all but her own species. Thanks to a sophisticated sensory processing system, animals can cut through a vast array of ambient auditory stimuli to extract meaningful information that allows them to tell where a sound came from, for example, or whether they should respond to a particular mating call.
An acoustic stimulus arrives at the ear as sound energy in the form of air pressure fluctuations. The sound signal triggers oscillations in mechanical resonators such as the eardrum and hair sensilla. These oscillations convert sound energy into mechanical energy, opening ion channels in auditory receptor cells and producing electrical currents that change the neuron's membrane potential. This, in turn, produces the action potential that carries the sound signal to the brain. This multistep signal transduction process takes less than a millisecond, but exactly how it occurs at this time scale remains obscure. Direct measurements of the individual steps can't be made without destroying the mechanical structure; consequently, most measurements are taken downstream of the mechanical oscillations at locations like the auditory nerve. Likewise, the temporal resolution of most stimulus–response trials is far too imprecise to analyze processing at the sub-millisecond level.
Given these experimental limitations, Tim Gollisch and Andreas Herz turned to computational methods and showed that it's possible to reveal the individual steps of complex signal processing by analyzing the output activity alone. Using grasshopper auditory receptors as models, the authors identified the individual signal-processing steps from eardrum vibrations to electrical potential within a sub-millisecond time frame and propose a model for auditory signaling.
The crucial step in their study is the search for those sets of inputs (stimuli) that would yield a given fixed output (response). To get the parameters to describe the final output, the authors generated a sound stimulus (two short clicks) and recorded axon responses of receptor neurons in a grasshopper auditory nerve. From these recordings, they defined the fixed output as the probability of a receptor neuron firing a single action potential. They then asked how the various parameters, which were associated with different time scales, could produce the same predefined firing probability.
A schematic representation of auditory signaling
By varying the stimulus parameters and comparing the obtained values within their mathematical framework—and making certain assumptions, for example, that the steps signal through a “feedforward” process—they could then tease out the individual processing steps that contribute to the desired output within the required time frame. With this approach, Gollisch and Herz disentangled individual steps of two consecutive integration processes—which they conclude are the mechanical resonance of the eardrum and the electrical integration of the receptor neuron—down to the microsecond level. Surprisingly, this fine temporal resolution is achieved even though the neuron's action potentials jitter by about one millisecond.
Thus, using just the final output, this approach can extract the temporal details of the individual processes that contribute to the chain of auditory transduction events. While this method is best-suited for deconstructing unidirectional pathways, the authors suggest it could also help separate “feedforward” from feedback signaling components, especially when feedback is triggered by the final steps. But since many sensory systems share the same basic signal-processing steps, this method is likely applicable to a broad range of problems.
| 0 | PMC539339 | CC BY | 2021-01-05 08:21:18 | no | PLoS Biol. 2005 Jan 4; 3(1):e26 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030026 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030039SynopsisGenetics/Genomics/Gene TherapyPlant SciencePlantsSeparating Wheat from Chaff in Plant Genomes Synopsis1 2005 4 1 2005 4 1 2005 3 1 e39Copyright: © 2005 Public Library of Science.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.
Sorghum Genome Sequencing by Methylation Filtration
==== Body
Plant genome sizes span the modest—54 million base pairs (Mb) in the bitter cress Cardamine amara—to the enormous—124,000 Mb in the lily Fritillaria assyriaca. By comparison, fruitfly and human genomes have 180 Mb and 3,200 Mb, respectively. Genomes of important crops such as sorghum, soybean, maize, and wheat hover between 735 Mb and 16,900 Mb, and determining their complete sequences is daunting and costly.
Wide size variations do not necessarily reflect differences in gene content, but rather reflect the presence of repetitive sequence elements that do not generally code for genes. Repetitive elements account for at least 75% of the maize and sorghum genomes. In a new study, Joseph Bedell and his colleagues describe a way to filter away repetitive elements when sequencing the genome of sorghum (Sorghum bicolor), a staple crop in much of the developing world because of its resilience in arid climates.
The authors use an approach known as methylation filtration that has been employed before for pilot plant genome analyses. Here they present compelling evidence of the method's reliability when applied to large-scale genome sequencing. The approach is built on the observation that in plants, methylation—a chemical tagging of DNA with methyl groups—occurs at repetitive sequences to a much greater degree than at gene sequences. This provides an opportunity to concentrate sequencing efforts on the coding portion of the genome.
To eliminate repetitive sequences, the authors introduced small pieces of sorghum chromosomes into bacteria strains designed to specifically destroy DNA sequences that carry methyl groups. Using two independent assessments, they estimated that methylation filtration reduced the amount of sorghum DNA they would need to sequence by two thirds, from 735 Mb to approximately 250 Mb.
But were any genes lost in the filtration step? The authors compared their results to partial sequence information generated previously from bacterial artificial chromosomes (BACs). BACs offer the most comprehensive representation of the genome because they contain large pieces of unmodified sorghum chromosomal DNA. Of the 148 genes identified on 14 sorghum BACs, 133 appeared in the filtered set. This means that the methylation filtration method captured at least 90% of the genes in the sorghum genome and 96% (131/137) if a repeat cluster of 11 known methylated genes is removed from the analysis.
A field of hybrid sorghum
Methylation filtration also compared favorably to shotgun sequencing, a method that reads the whole genome in small fragments that are progressively assembled into larger pieces by computer analysis. The authors reported that after sequencing 285 Mb of filtered sorghum DNA—approximately 1.15 times the length of the sorghum coding regions—they obtained on average 65% of the length of 96% of the genes. Theoretical calculations and simulation based on the genome of Arabidopsis—a plant model organism—predicted that shotgun approach would yield similar results (67% of the length of 96% of the genes) after sequencing the equivalent of 1.15 times its total length (rather than 1.15 times the length of just the coding regions). Thus, methylation filtration can provide as much information on coding sequences as the shotgun approach, with less investment in sequencing.
Methylation filtration does not yield a complete genome map, but it offers quicker, more affordable access to genes than most commonly used sequencing approaches. Sorghum is closely related to maize and sugar cane, and more distantly to rice. The availability of its genome sequence offers the chance for more in-depth experiments into the evolution of the grass family, and promises important insights into the genetic control of drought resistance.
| 0 | PMC539340 | CC BY | 2021-01-05 08:21:18 | no | PLoS Biol. 2005 Jan 4; 3(1):e39 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030039 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030040SynopsisGenetics/Genomics/Gene TherapyInfectious DiseasesMicrobiologyEubacteriaMultiple Campylobacter Genomes Sequenced Synopsis1 2005 4 1 2005 4 1 2005 3 1 e40Copyright: © 2005 Public Library of Science.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.
Major Structural Differences and Novel Potential Virulence Mechanisms from the Genomes of Multiple Campylobacter Species
==== Body
In 1995, the first complete bacterial genome sequence was published. Now, nearly 200 bacterial genomes have been completed, and a new one hits the scientific press most weeks. This burgeoning industry is not just scientific “stamp collecting,” however. Having all these genome sequences may provide useful clues about why some bacteria cause human disease, how to control their spread, and how to treat the infections caused by them. By comparing genome sequences, scientists can learn much more about what makes a bacteria tick than they can learn from a single sequence.
Derrick Fouts and his colleagues have taken this comparative approach with Campylobacter. Infection with a Campylobacter species is one of the most common causes of human bacterial gastroenteritis. In the US, 15 out of every 100,000 people are diagnosed with campylobacteriosis every year, and with many cases going unreported, up to 0.5% of the general population may unknowingly harbor Campylobacter in their gut annually. Diarrhea, cramps, abdominal pain, and fever develop within 2–5 days of picking up a pathogenic Campylobacter species, and in most people, the illness lasts for 7–10 days. But the infection can sometimes be fatal, and some individuals develop Guillain-Barré syndrome, in which the nerves that join the spinal cord and brain to the rest of the body are damaged, sometimes permanently.
Campylobacteriosis is usually caused by C. jejuni, a spiral-shaped bacterium normally found in cattle, swine, and birds, where it causes no problems. But the illness can also be caused by C. coli (also found in cattle, swine, and birds), C. upsaliensis (found in cats and dogs), and C. lari (present in seabirds in particular). Disease-causing bacteria generally get into people via contaminated food, often undercooked or poorly handled poultry, although contact with contaminated water, livestock, or household pets can also cause disease.
Genome sequencing and comparison of four species of Campylobacter
In 2000, C. jejuni was the first food-borne pathogen to be completely sequenced, but we still know little about how Campylobacter species cause disease. In their search for clues, Derrick Fouts and coworkers have completely sequenced the genome of C. jejuni strain RM1221 (isolated from a chicken carcass) and compared it with the previously sequenced C. jejuni strain NCTC 11168 and with the unfinished sequences of C. coli strain RM2228 (a multi-drug-resistant chicken isolate), C. lari strain RM2100 (a clinical isolate), and C. upsaliensis strain RM3195 (taken from a patient with Guillain-Barré syndrome).
The researchers describe numerous differences and similarities between these different Campylobacter strains and species. For example, there are major structural differences between the genomes caused by the insertion of new stretches of DNA. Some of these pieces of DNA may carry genes that improve bacterial virulence or fitness, so their presence could help to explain the different biological behaviors of these strains. There are also major variations in the genes responsible for synthesis of molecules that are important for the interaction of Campylobacter with the environment. Such differences could underlie the host specificity of the different species.
Differences between the Campylobacter species in genes that are likely to be involved in aspects of bacterial virulence, such as adherence, motility, and toxin formation, are all detailed by Fouts et al., who also describe a new putative Campylobacter virulence locus. Further work is needed to relate these genomic differences to functional differences, but this detailed comparative genomic analysis provides the core blueprint for this important family of human pathogens. And in doing so, it lays the foundation for the development of new ways to monitor and control Campylobacter in the food chain and in human infection.
| 0 | PMC539341 | CC BY | 2021-01-05 08:21:19 | no | PLoS Biol. 2005 Jan 4; 3(1):e40 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030040 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030043SynopsisBotanyEcologyEvolutionPlant SciencePlantsDispersal or Drift? More to Plant Biodiversity Than Meets the Eye Synopsis1 2005 4 1 2005 4 1 2005 3 1 e43Copyright: © 2005 Public Library of Science.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.
Relaxed Molecular Clock Provides Evidence for Long-Distance Dispersal of Nothofagus (Southern Beech)
==== Body
Over 250 million years ago (mya), all the continents of Earth formed a single land mass called Pangaea. Some 50 million years later, this supercontinent began to split in two, forming Laurasia—now North America, Asia, and Europe—and Gondwana—present-day Antarctica, Australia, South America, Africa, and India. After another 50 million years, Gondwana, too, broke up. At the end of the Cretaceous period, New Zealand split off (about 80 mya), then South America and Australia separated from Antarctica (about 35 mya). Fairy-tale quality aside, the story of continental drift fits comfortably with the geological and fossil record and feeds our understanding of current distributions of plant biodiversity.
Although we know how and when Pangaea broke apart, the distribution of fossils of the same species on many different continents, separated by vast ocean waters, challenges us to explain how they got there. Plant life on New Zealand, for example, shares striking similarities to that on other Southern Hemisphere land masses, but scientists have yet to agree on how this came to pass. In particular, one genus, Nothofagus—the southern beech tree, a plant whose 80-million-year-old fossil history goes back to the days of Gondwana—has polarized views on the nature of Southern Hemisphere biogeography.
One theory suggests that geographic barriers (New Zealand and Australia are separated by the Tasman Sea) would have prevented species expansion after the break-up of the continents, so similar contemporary species must have already existed in both places before New Zealand broke away from Gondwana. In this scenario, called vicariance, ancestors of existing lineages drifted with the repositioned land masses. Another hypothesis, born of existing distributions and fossil data, suggests that long-range oceanic dispersal is more likely. But since Nothofagus seeds are not considered ocean-worthy vessels, many believe vicariance is the only possible explanation.
Peter Lockhart and colleagues argue that a clear picture of the divergence dates of various southern beech species could help clarify the relative contributions of vicariance versus dispersal. But they would need significant lengths of DNA sequences to reliably characterize the evolutionary history of each species.
Consequently, Lockhart and colleagues analyzed a 7.2-kilobase fragment of the chloroplast genome (which typically ranges from 110,000 bp to 160,000 bp) for 11 species of three Nothofagus subgenera—Lophozonia, Fuscospora, and Nothofagus—from South America, Australia, and New Zealand. Reconstructing the trees' evolutionary relationships (phylogeny) based on analyses of their chloroplast sequences, the authors discovered a nuanced evolutionary history that supports vicariance for some species and dispersal for others.
Nothofagus, the southern beech, on the slopes of Mt. Ruapehu in New Zealand (Photo: Peter Lockhart)
Assuming that beech was present throughout Gondwana (which fossil data support), the sequence of the Gondwana breakup should be reflected in the beech's phylogeny. New Zealand beeches should be more distantly related to both Australian and South American species, because of the greater period of separation—65 million years compared to 30 million years. Yet Australian and New Zealand beeches are more closely related to each other than to South American species, which reflects more recent relationships. Given that fossils of all beech subgenera extend back to the New Zealand Cretaceous period, the dating of splits and the nature of the relationships indicate extinction of beech lineages within current subgenera in New Zealand, and possibly in Australia and South America.
Lockhart and colleagues' analyses suggest that the relationships of the Australian and New Zealand Lophozonia and Fuscospora species are too recent to have roots in Gondwana, indicating a role for transoceanic dispersal. The evolutionary relationship between the Australasian and South American Fuscospora lineages, however, is consistent with vicariance. These divergence results, the authors conclude, indicate that current distributions of Nothofagus cannot be explained solely by continental drift (followed by extinction of some species) and that contemporary New Zealand Nothofagus species are not direct descendants of the beeches thought to have reached the island after the split from Antarctica.
Taken together, the results highlight the need for caution in evaluating fossil evidence. The fossil record doesn't necessarily capture when a species first appeared, and a continuous fossil presence can mask extinctions and reinvasions. The authors conclude that their molecular data make the case for investigating possible mechanisms of long-range dispersal—especially the dispersal properties of Nothofagus seeds—and stresses the need to consider more complex hypotheses to explain something as dynamic and complex as the evolutionary history of biodiversity.
| 0 | PMC539342 | CC BY | 2021-01-05 08:21:19 | no | PLoS Biol. 2005 Jan 4; 3(1):e43 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030043 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030055SynopsisCell BiologyMicrobiologyBiochemistryEubacteriaHow Bacteria Stopped Worrying and Learned to Love…Formaldehyde Synopsis2 2005 4 1 2005 4 1 2005 3 2 e55Copyright: © 2005 Public Library of Science.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.
Flux Analysis Uncovers Key Role of Functional Redundancy in Formaldehyde Metabolism
==== Body
Poring over the vast amount of sequence and genetic information now available for many organisms, scientists frequently encounter what appear to be redundant biochemical pathways. Redundant pathways take the same starting material and transform it into the same product, but through different routes. Why should cells maintain redundant pathways?
An interesting case is that of Methylobacterium extorquens, a bacterium that can grow on organic molecules with a single carbon atom such as the alcohol methanol. Bacteria in this species first oxidize methanol into formaldehyde, then use formaldehyde to make serine—the entry point for the synthesis of many of the cell's building blocks—via two apparently redundant pathways. The short pathway is a direct (non-enzymatic) reaction of formaldehyde with tetrahydrofolate to make methylene-tetrahydrofolate, which donates a single carbon atom for serine synthesis. A hypothesized long pathway could also lead to methylene-tetrahydrofolate through a long series of enzymatic reactions, one of which consumes energy.
Christopher Marx and colleagues now demonstrate that the bacteria modulate their use of each pathway during the course of acclimation to growth on methanol. In the process the authors offer new insights into the bacteria's rapid disposal of formaldehyde, a toxic chemical that would pickle cells in a minute if it was allowed to accumulate.
The short and long (pathways) of formaldehyde metabolism
The authors noticed that the short pathway transferred both hydrogen atoms of formaldehyde to the serine molecule while the long pathway transferred only one. If they fed bacteria methanol in which hydrogen had been replaced with its heavier form, deuterium: the resulting serine was slightly heavier than normal, as a result of acquiring one—or two—deuterium atoms.
By measuring the ratio of the two serine forms, Marx and colleagues inferred the relative contribution of the two pathways to serine synthesis. The long pathway dominated, accounting for eight times more serine than the short pathway, when cells first encountered methanol. But the situation was reversed after the cells acclimated to methanol: then the short pathway produced 15 times more serine than the long pathway.
The authors also measured absolute amounts of formaldehyde processed by each pathway, using methanol marked with a heavy form of carbon (14C). Although the relative contribution of the long pathway decreased during ramping-up to methanol growth, the absolute amount of formaldehyde that flowed through it increased 8-fold within the first half of the transition, and then decreased.
The authors generated a mathematical model based on known reaction rates from the short and long pathways. When they simulated methanol exposure, the model predicted a switch from long to short pathway very similar to what they had observed experimentally.
The authors conclude that the pathways are not in fact redundant, but fulfill different functions. The long pathway is not an efficient means of serine synthesis from formaldehyde; in fact, the small amount of serine it produces is at some energy cost. But it allows the cell to spend ATP—the molecular fuel—to jump-start formaldehyde assimilation while avoiding formaldehyde accumulation when the cells first experience methanol. The short pathway is a direct and efficient (energy-free) route to serine production (and hence growth), but one that is slower to reach its cruising speed. Thus, the cells use these two formaldehyde assimilation routes like a driver uses the transmission of a car: start with powerful low gears when first accelerating, and shift to more efficient gears once hurtling down the road.
The combination of both pathways represents an elegant solution to the problem of growth in toxic environments and provides a useful paradigm for detoxification in medical and environmental contexts.
| 0 | PMC539343 | CC BY | 2021-01-05 08:21:18 | no | PLoS Biol. 2005 Feb 4; 3(2):e55 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030055 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-5-541558142610.1186/1471-2202-5-54Research ArticleLeptin and insulin stimulation of signalling pathways in arcuate nucleus neurones: PI3K dependent actin reorganization and KATP channel activation Mirshamsi Shirin [email protected] Hilary A [email protected] Ke [email protected] Erin [email protected] Laura A [email protected] Alexander [email protected] Calum [email protected] Michael LJ [email protected] Division of Pathology & Neuroscience, Ninewells Hospital & Medical School, University of Dundee, Dundee DD1 9SY UK2 Division of Signal Transduction, School of Life Sciences, University of Dundee, Dundee DD1 5EH UK2004 6 12 2004 5 54 54 11 5 2004 6 12 2004 Copyright © 2004 Mirshamsi et al; licensee BioMed Central Ltd.2004Mirshamsi 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
Leptin and insulin are long-term regulators of body weight. They act in hypothalamic centres to modulate the function of specific neuronal subtypes, by altering transcriptional control of releasable peptides and by modifying neuronal electrical activity. A key cellular signalling intermediate, implicated in control of food intake by these hormones, is the enzyme phosphoinositide 3-kinase. In this study we have explored further the linkage between this enzyme and other cellular mediators of leptin and insulin action on rat arcuate nucleus neurones and the mouse hypothalamic cell line, GT1-7.
Results
Leptin and insulin increased the levels of various phosphorylated signalling intermediates, associated with the JAK2-STAT3, MAPK and PI3K cascades in the arcuate nucleus. Inhibitors of PI3K were shown to reduce the hormone driven phosphorylation through the PI3K and MAPK pathways. Using isolated arcuate neurones, leptin and insulin were demonstrated to increase the activity of KATP channels in a PI3K dependent manner, and to increase levels of PtdIns(3,4,5)P3. KATP activation by these hormones in arcuate neurones was also sensitive to the presence of the actin filament stabilising toxin, jasplakinolide. Using confocal imaging of fluorescently labelled actin and direct analysis of G- and F-actin concentration in GT1-7 cells, leptin was demonstrated directly to induce a re-organization of cellular actin, by increasing levels of globular actin at the expense of filamentous actin in a PI3-kinase dependent manner. Leptin stimulated PI3-kinase activity in GT1-7 cells and an increase in PtdIns(3,4,5)P3 could be detected, which was prevented by PI3K inhibitors.
Conclusions
Leptin and insulin mediated phosphorylation of cellular signalling intermediates and of KATP channel activation in arcuate neurones is sensitive to PI3K inhibition, thus strengthening further the likely importance of this enzyme in leptin and insulin mediated energy homeostasis control. The sensitivity of leptin and insulin stimulation of KATP channel opening in arcuate neurones to jasplakinolide indicates that cytoskeletal remodelling may be an important contributor to the cellular signalling mechanisms of these hormones in hypothalamic neurones. This hypothesis is reinforced by the finding that leptin induces actin filament depolymerization, in a PI3K dependent manner in a mouse hypothalamic cell line.
==== Body
Background
Leptin and insulin function as peripherally-derived hormone signals involved in the long-term regulation of energy balance [1-4]. Their circulating levels are directly proportional to adipose mass and CNS access occurs via saturable receptor-mediated processes. The primary CNS target for these adipostats is the ARC, where leptin and insulin receptors are highly expressed, and where direct administration of either hormone has a potent effect on food intake and body weight. Two specific ARC neurone populations have been strongly implicated in sensing changes in levels of circulating leptin and insulin and transducing these signals into neuronal outputs [1,3]. These "first-order" neurones encompass the melanocortin precursor, POMC containing neurones and NPY and AgRP co-containing neurones, the former associated with catabolic, the latter anabolic, outputs. Leptin and insulin increase POMC mRNA levels and decrease NPY & AgRP mRNA levels respectively.
However, transcriptional control is not the only effector mechanism elicited by these hormones on ARC neurones. Electrophysiological studies have shown that leptin depolarizes and increases the firing rate of ARC POMC neurones and inhibits the tone of NPY/AgRP neurones [5]. Although the electrophysiological actions of insulin have not been reported for identified POMC and NPY/AgRP neurones, both leptin and insulin have been demonstrated to inhibit, by hyperpolarization, the firing of a sub-population of ARC neurones, identified by their sensitivity to changes in extracellular glucose concentration [6,7]. For these latter neurones, termed glucose-responsive (GR), KATP channels have been identified as an effector mechanism through which leptin and insulin elicit neuronal inhibition. Consequently, leptin and insulin signal the status of body energy stores by activating their receptors on ARC neurones, eliciting changes in the electrical activity and amounts of releasable peptides in specific neuronal populations, leading to compensatory effector outputs, such as changes in food intake, energy balance and glucose homeostasis [8].
Obese humans have elevated leptin and insulin levels, indicative of central resistance to these hormones [9]. The mechanisms underlying this resistance are unclear, with defective hormone passage through the BBB and flawed receptor-signal transduction in ARC neurones being the prime candidates [10,11]. Consequently, it is important to understand the molecular mechanisms underlying leptin and insulin receptor modulation of ARC first-order neurones. Leptin and insulin, by stimulation of their respective receptors, have been demonstrated to activate various signalling pathways in peripheral tissues [10-13]. However, as these hormones induce seemingly identical actions on ARC neurones, both in terms of behavioural output and effects on ARC neurone excitability, some parallelism or convergence of signalling is likely [12,13]. Leptin, by binding to the long form of the leptin receptor (ObRb) has been demonstrated to activate three main signalling cascades, JAK2 – STAT3, MAPK and PI3K, the latter two of which are also intermediates in insulin receptor activation [14,15]. However, recent studies have strongly implicated PI3K as the key signalling intermediate in leptin and insulin actions on hypothalamic neurones influencing food intake and body weight [16,17].
Thus, to elucidate further the pathways that contribute to convergent actions of leptin and insulin on ARC neurones, we have examined the phosphorylation status of key leptin and insulin signalling intermediates in the ARC and have explored the linkage, with a focus on PI3K mediated signal transduction pathways, between these hormones and ARC neurone KATP channel activation.
Results
Leptin and insulin stimulate phosphorylation of signalling proteins in ARC
Rat hypothalamic tissue sections, predominantly made up of arcuate nucleus, were treated with aCSF alone or with leptin (10 nM) or insulin (0.1 or 1 nM, which produced identical results) for 1, 5, 15 and 30 minutes. Leptin and insulin stimulation induced comparable immunoblot profiles (Figure 1), with increased levels of phosphorylated STAT3 (p-STAT3), phosphorylated MAPK (p-MAPK), phosphorylated PKB/Akt (p-PKB) and GSK3 (p-GSK3). The phosphorylation status of the latter two proteins, PKB and its downstream effector GSK3, were utilised as a sensitive assay for hormone activation of PI3K. Leptin (10 nM) or insulin (0.1 nM) treatment was observed to cause an increase in phosphorylation of all four proteins. This increase in phosphorylation was generally transient with the highest levels of phosphorylation at the 1 and/or 5 minute time points. Subsequent to this peak level, in the majority of experiments, the phosphorylation was not sustained over the time period examined and returned to control values within 30 minutes (Figure 1A,1B). These data demonstrate that all 3 pathways potentially contribute to insulin and leptin signalling in ARC neurones and thus play a role in connecting leptin or insulin receptor activation to neuronal effector outputs. As both leptin and insulin signalling in the ARC require PI3K activity for reduction in food intake and body weight [16,17], we examined the sensitivity of the phosphorylation of PKB/GSK3 and MAPK to the presence of PI3K inhibitors. Isolated ARC sections were incubated either in control aCSF, 10 nM wortmannin or 10 μM LY294002, for 20 minutes prior to exposure to control aCSF, leptin (10 nM) or insulin (0.1 nM), in the continued presence of the appropriate inhibitor. The presence of LY294002 prevented leptin or insulin induced phosphorylation of PKB and GSK3 following 1 minute exposure to these hormones, illustrated in Figure 2A, as expected for proteins downstream of PI3K [18]. Furthermore, the presence of the PI3K inhibitor per se reduced p-PKB and p-GSK3 levels significantly, indicating that PI3K is active to a limited degree in these ARC neurones. Similar results were obtained for wortmannin (data not shown). However, surprisingly the PI3K inhibitors also reduced the leptin- and insulin-stimulated phosphorylation of MAPK (Figure 2B). These data further establish PI3K as a key component of neuronal leptin and insulin signalling in ARC neurones and suggest a potential role for PI3K in leptin and insulin driven transcriptional activity. Because the phosphorylation status of these signalling intermediates was examined in whole ARC extracts, this supplies little information as to the mechanisms by which adiposity hormones target specific ARC neurones. Thus, we have tried to delineate the molecular events connecting leptin and insulin receptor activation, PI3K activity and effector outputs. Here we focus on the activation of KATP channels, responsible for leptin and insulin inhibition of an electrophysiologically identified subset of ARC neurones [6,7].
Figure 1 Effects of leptin and insulin on phosphorylation of STAT3, MAPK, PKB and GSK3 Rat ARC wedges were incubated for 0, 1, 5, 15 or 30 minutes with 10 nM leptin (A) or 0.1 – 1 nM insulin (B) before cells were lysed and equal amounts of lysate were subjected to SDS-PAGE and transferred to nitrocellulose membrane. The phosphorylated levels of p42/p44 MAPK, PKB, STAT3 and GSK3α/β were detected by immunoblotting with appropriate specific antibodies. The total amount of PKB is also shown. Bands were quantified using densitometry. The values are expressed as relative to the corresponding aCSF control group, and normalized for protein loading. Values represent the mean ± SEM for between 4–6 animals for each time point. * P < 0.05 and ** P < 0.01.
Figure 2 Changes in phosphorylation of PKB, GSK3 and MAPK by inhibition of PI3K Rat ARC wedges were pretreated with 10 μM LY294002 or aCSF for 20 minutes and incubated for 1 minute with 10 nM leptin or 1 nM insulin or aCSF. Equal amounts of protein lysate were subjected to SDS-PAGE and transferred to nitrocellulose membrane. The phosphorylated levels of PKB, and GSK3αβ (A) and p42/p44 MAPK (B) were detected by immunoblotting with appropriate specific antibodies. The total amount of PKB is also shown. Bands were quantified using a densitometer. The values are expressed as relative to the corresponding aCSF control group, and normalized for protein loading. Values represent the mean ± SEM for between 4–5 and 3–5 animals for each time point with leptin and insulin respectively. * P < 0.05 and ** P < 0.01.
Leptin and insulin activate KATP channels in acutely isolated ARC neurones
Cell-attached recordings from rat isolated ARC neurones were used to confirm that leptin and insulin activate the large conductance KATP channel as previously described [6,7]. Leptin, present in the recording electrode during cell-attached recordings, increased mean K+ channel activity in 45% of unidentified neurones (n = 25/55). Mean channel activity (Nf.Po), 1–2 minutes following cell attached formation was 0.08 ± 0.02 and increased to 0.38 ± 0.03 (n = 10, P < 0.01) after peak activation had occurred (10.6 ± 1.0 minutes after patch formation). In control cell-attached recordings of between 10–25 minutes, with 10 nM leptin in the pipette solution, there was no effect on K+ channel currents observed in other ARC neurones (n = 30). During peak KATP channel activation by leptin, bath application of the KATP channel inhibitor tolbutamide (200 μM), reduced mean channel activity by 56 ± 12% (n = 4; P < 0.05), an effect reversible on washout of drug (Figure 3A). Following leptin-induced increase in channel activity, patch excision into the inside-out configuration allowed channel sensitivity to ATP to be assessed. At a patch potential of 0 mV and in asymmetric cation gradients, mean channel activity was 0.40 ± 0.09 (n = 3) and bath application of 3 mM MgATP (Figure 3B) reversibly reduced channel activity by 67.1 ± 9.7% (P < 0.05). Current-voltage relations under symmetrical K+ conditions were linear, with a mean single channel conductance of 156 ± 15 pS (n = 3). The sensitivity to tolbutamide, ATP and single channel characteristics are consistent with leptin activation of the large conductance KATP channel of GR neurones [6,19]. In a separate series of cell-attached recordings from isolated ARC neurones, bath application of insulin (0.1 – 10 nM) also increased KATP channel activity in 45% of unidentified neurones (n = 14/31). Nf.Po increased from 0.14 ± 0.03 under control conditions to 0.49 ± 0.08 (n = 7; P < 0.01) after approximately 10 – 20 minutes exposure to insulin (Figure 3C). Insulin had no effect on other K+ channel currents in cell-attached recordings from the remaining neurones (n = 17). Consequently, these data are in agreement with previous studies on rat ARC GR neurones [6,7].
Figure 3 Leptin and insulin activate large conductance KATP in ARC neurones A, representative cell-attached recording from an acutely dissociated ARC neurone. Leptin (10 nM), present in the electrode solution, increased the activity of a K+ channel, which was inhibited reversibly by bath application of 200 μM tolbutamide. Upward deflections in this and subsequent cell-attached recordings are extracellularly recorded action current activity. B, representative recording from an inside-out patch under asymmetrical K+ conditions and held at 0 mV obtained from an acutely dispersed ARC neurone, following cell-attached leptin-induced increase in channel activity. Note that bath application of 3 mM MgATP reversibly inhibited K+ channel activity. Inserts are expanded regions of traces showing channel activity in more detail. C, representative cell-attached recording from an acutely dissociated ARC neurone. Under control conditions, few channel openings are observed. Subsequent to bath application of 0.1 nM insulin, there is a marked increase in KATP channel activity. Corresponding diary plot of channel activity (Nf.Po) with time displays the insulin induced increase in activity.
Leptin and insulin activation of ARC neurone KATP is PI3K dependent
As leptin and insulin activation of KATP channels in ARC neurones is rapid (<5–10 minutes) and the leptin increase in KATP activity demonstrated in isolated patches [6], this action is unlikely to be mediated by changes in transcription. Furthermore, in cell-attached recordings, following leptin (10 nM) stimulated KATP channel activity, application of the MAPK pathway inhibitor, PD98059 (10 μM; n = 4) had no effect on Nf.Po (Figure 4A). A previous study has demonstrated that insulin activated ARC neurone KATP channels are similarly insensitive to this MAPK pathway inhibitor [7]. However, inhibition of PI3K does reverse both leptin (Figure 4B) and insulin-induced activation of ARC neurone KATP channels. Leptin (10 nM) increased KATP mean Nf.Po from 0.21 ± 0.10 to 0.68 ± 0.28 (n = 3, P < 0.02), and subsequent bath application of 10 nM wortmannin reduced KATP activity to a mean value of 0.33 ± 0.13 (n = 3, P < 0.02) over a period of 15–20 minutes, an Nf.Po indistinguishable from control (P > 0.4). Similarly, in a separate series, leptin increased Nf.Po from 0.15 ± 0.04 to 0.35 ± 0.05 (n = 4, P < 0.05) and subsequent application of 10 μM LY294002 reduced Nf.Po to 0.19 ± 0.04 (n = 4, P < 0.01) within 15–20 minutes. Essentially identical data have been reported previously for the effects of these PI3K inhibitors on insulin-activated ARC KATP channel activity [7]. Thus these results demonstrate that leptin and insulin signalling pathways converge on PI3K to elicit GR neurone hyperpolarization, and confirm that PI3K is a key enzyme in individual ARC neurone responsiveness to both leptin and insulin.
Figure 4 PI3K activity mediates leptin activation of KATP Representative cell-attached recordings with leptin (10 nM), present in the recording electrode. A, traces illustrate that leptin-induced increase in KATP activity is not reversed on bath application of the MEK inhibitor, PD 98059 (50 μM). The corresponding diary plot for part of the recording, initiated 20 mins after recording began, and following attainment of maximal leptin-induced KATP channel activity is shown. B, traces show that leptin-induced KATP channel activity is inhibited by subsequent application of the PI3K inhibitors, wortmannin (10 nM) or LY294002 (10 μM). Corresponding diary plots for Nf.Po from a single experiment for each PI3K inhibitor are shown below the relevant traces.
Such a central role for PI3K suggests that its main lipid product, PtdIns(3,4,5)P3 may serve as an important second messenger for downstream effectors such as the KATP channel. The mechanism by which PtdIns(3,4,5)P3 recognises downstream target proteins is by binding to specialised phosphatidylinositol recognition sites, such as the pleckstrin homology (PH) domain [20]. Thus, to demonstrate that PtdIns(3,4,5)P3 production is elevated in ARC neurones following exposure to leptin and insulin, we used the PH domain of GRP-1, which selectively binds PtdIns(3,4,5)P3 [21] coupled to GFP (PH-GRP1-GFP) in an overlay assay on fixed freshly isolated ARC neurones. In non-stimulated ARC neurones there is significant labelling of all neurones with PH-GRP1-GFP (Figure 5). This is likely due to inherent PI3K activity of the neurones, rather than non-specific binding, as a PH-GRP1-GFP fusion protein with a single point mutation (K273A), which does not bind PtdIns(3,4,5)P3 [21,22], displays very little reactivity with non-stimulated ARC neurones (n = 7). Stimulation of isolated ARC neurones with leptin (10 nM) for 10 minutes resulted in a proportion (38 ± 8%; n = 5) of dispersed neurones displaying increased fluorescence after treatment with wild type fusion protein (Figure 5A). In addition, exposure of dispersed neurones to insulin (1 nM) for 5–10 minutes induced increased binding of PH-GRP1-GFP fusion protein in a similar proportion of neurones (43 ± 15%; n = 6), although insulin appeared to induce greater levels of binding/fluorescence than leptin (Figure 5B). However, leptin and insulin driven phosphorylation of PKB and GSK3 along with induction of elevated PtdIns(3,4,5)P3 levels in ARC neurones, are only indicative of increased PI3K activity. Thus we examined IRS-2 associated PI3K activity [16] in isolated ARC wedges exposed for one or two minutes to leptin (up to 50 nM) or insulin (up to 100 nM). Although we observed an increase in activity in 4/8 and 4/7 experiments for leptin and insulin respectively, there was no overall significant increase observed (data not shown). This may be due to the relative paucity of leptin and /or insulin sensitive neurones in the overall cellular population.
Figure 5 Leptin and insulin increase PtdIns(3,4,5)P3 in isolated neurones Acutely isolated ARC neurones were incubated in the absence and presence of 10 nM leptin (A) or 1 nM insulin (B) for 10 minutes. Cells were fixed and permeabilized, as described in Methods, prior to incubation with wild type (wt) or K273A mutant (mt) PH-GRP1-GFP fusion protein for 1 hour. Cells were subsequently processed for visualising GFP by confocal microscopy. Note that leptin and insulin increased the binding of wild type PH-GRP1-GFP in ARC neurones, and this is shown as both the fluorescence image (upper panels in A, B) and as a false colour image (lower panels in A, B), where blue represents low or non-detectable fluorescence and red the highest fluorescence intensity.
Leptin and insulin activation of GR neurone KATP requires actin filament re-organisation
Previous studies have demonstrated that the phosphatidylinositol lipid second messenger, PtdIns(3,4,5)P3 activates KATP channels in an insulin-secreting cell line when applied directly to the internal aspect of isolated patches [23]. However, activation of KATP is probably not due to direct binding of the lipid to channel subunits, as the effect of PtdIns(3,4,5)P3 on KATP was prevented by the presence of the actin stabilizing agent, phalloidin. Additionally, leptin-induced opening of this insulin-secreting cell KATP channel was occluded when phalloidin was present in the cell interior [23]. Thus we examined whether adiposity hormone signalling in ARC GR neurones also requires actin remodelling in order to manifest a specific effector output, the activation of hypothalamic neurone KATP channels. As our assessment of hormone activation of GR neurone KATP channels uses cell-attached recordings, we used the membrane permeable actin stabilizing toxin, jasplakinolide to induce actin polymerization [24]. In preliminary experiments, jasplakinolide (100 nM) was demonstrated to have no effect when applied directly to isolated inside-out patches obtained from ARC neurones, under asymmetrical recording conditions, containing spontaneously active KATP channels (n = 4; P > 0.5; data not shown). Cell-attached recordings with leptin (10 nM) present in the pipette solution, increased mean KATP channel activity from 0.07 ± 0.02 to 0.44 ± 0.07 (n = 4; P < 0.01). Subsequent bath application of jasplakinolide (50 – 100 nM) reversed the leptin-induced KATP activation (Figure 6A), with channel activity returning to 0.10 ± 0.02 (n = 4; P < 0.01) within 5–10 minutes, a level indistinguishable from pre-leptin controls (P > 0.5). In a second series of cell-attached recordings, bath application of insulin (0.1 – 10 nM) increased mean KATP activity from 0.06 ± 0.01 to 0.64 ± 0.16 (n = 4; P < 0.01) and subsequent bath application of jasplakinolide (100 nM), in the continuous presence of insulin, reduced channel activity to 0.17 ± 0.06 (n = 4; P < 0.01) within 5–10 minutes (Figure 6B). The inhibition of leptin and insulin stimulated KATP channel activity by jasplakinolide was reversible on washout of the toxin in 2/4 and 3/4 patches for leptin and insulin respectively.
Figure 6 Actin dynamics mediate leptin and insulin activation of KATP A, representative cell-attached recording from an acutely isolated ARC neurone with leptin (10 nM) in the electrode solution. Following attainment of increased KATP activity, bath application of jasplakinolide (100 nM) reversibly reduced channel activity. The corresponding diary plot for this experiment is shown. B, representative cell-attached recording from an ARC neurone. Bath application of insulin (10 nM) increased KATP channel activity and subsequent bath addition of 100 nM jasplakinolide, concomitant with 10 nM insulin, reversibly inhibited the insulin stimulated KATP activity. The corresponding diary plot is shown.
PI3K mediates leptin-induced actin filament reorganisation in GT1-7 cells
Preliminary experiments labelling rat ARC slices with rhodamine-conjugated phalloidin to stain for F-actin were unsatisfactory due to the overall high levels of staining in the slices and inability to distinguish clearly individual neurones and their responses to hormone stimulation. Similarly, use of freshly isolated ARC neurones was precluded as the phalloidin staining was inconsistent among individual neurones within a single preparation and between neuronal preparations. Thus we have used the mouse hypothalamic cell line, GT1-7, to demonstrate that leptin utilises a PI3K-dependent signalling cascade to modify cytoskeletal dynamics. A previous study indicated that GT1-7 cells express ObRb [25], although others do not concur with this conclusion [26]. Using RT-PCR, we detected the presence of leptin receptor mRNA in GT1-7 cells by amplification of a common extracellular domain of the mouse receptor. Further analysis using primers specific to ObRb, which contains a long cytosolic domain with the intracellular protein motifs required for signalling [10], demonstrates the presence of this receptor isoform in GT1-7 cells (Figure 7A). We have also used this hypothalamic cell line to examine whether leptin is capable of increasing PI3K activity directly. In response to 50 nM leptin, IRS-2-associated PI3K activity was modestly, but significantly, increased (Figure 7B). Thus, using native GT1-7 cells leptin (1–10 nM) induced a decrease in cortical F-actin as visualised by alexa 488 conjugated phalloidin staining, which was prevented or reversed by the presence of 100 nM jasplakinolide (n = 8; Figure 8 upper panels). As cellular cortical actin structure is determined by the dynamic equilibrium between F- and G-actin, a reduction in F-actin at the plasma membrane should be accompanied by a corresponding increase in the concentration of free G-actin in the cells [27]. Figure 8 (middle panel) demonstrates, using alexa 594 conjugated DNase I staining of the same cells, that leptin does indeed increase the concentration of G-actin and that this effect is also sensitive to the presence of jasplakinolide (n = 8). Indeed, dual staining of the GT1-7 cells demonstrates (Figure 8, lower panel) that leptin alteration of the cortical cytoskeleton is due to a concomitant increase in the content of G-actin at the expense of F-actin, and that this action is completely inhibited in the presence of jasplakinolide. The alteration in cytoskeletal dynamics by leptin is also PI3K dependent as shown in Figure 9A, where the presence of either 10 nM wortmannin (n = 13) or 10 μM LY294002 (n = 13) substantially reduced the ability of leptin to decrease the levels of F-actin and increase G-actin as assessed by phalloidin and DNase I staining respectively. This cell staining method of assessing leptin stimulated changes in actin dynamics was compared to direct quantitative analysis of actin. Live cells were treated with leptin (10 nM) ± jasplakinolide (100 nM) or LY294002 (10 μM) or wortmannin (10 nM) for 20 minutes, Triton-X-100 soluble (G) and insoluble (F) actin fractions separated and run on a gel [28]. Exposure of cells to leptin did not alter total cellular actin, whereas G-actin levels increased by 2 fold, at the expense of F-actin, the levels of which declined by 65% (n = 4; Figure 9C,9D). Thus, leptin induced a change in the G/F actin ratio from a control value of 0.54 to 3.17. These data correlate well with the change in fluorescence intensity observed in leptin-treated fixed cells (n = 8), where G-actin levels were also increased by 2 fold and F-actin decreased by 70% (Figure 9B). Exposure of cells to the F-actin stabilizing agent, jasplakinolide or the PI3K inhibitors, LY294002 or wortmannin prevented leptin from inducing F-actin disassembly as observed by either assay (Figure 9A,9B,9C,9D). In addition, protein overlay experiments using wild type PH-GRP1-GFP fusion protein binding to assess PtdIns(3,4,5)P3 levels in GT1-7 cells demonstrate that leptin increases PI3K activity concurrently with the re-organization of cortical actin in this cell line, with leptin stimulation inducing little change in the K273A mutant PH-GRP1-GFP binding to these cells (Figure 10).
Figure 7 Leptin stimulates PI3K activity in GT1-7 cells A, expression of the leptin receptor mRNA in mouse GT1-7 cells. Lanes 1–3, RT-PCR detection of the common ObR isoform in hypothalamus (lane 2) and GT1-7 cells (lane 3), together with a negative control (lane 1). Lanes 4–6, RT-PCR detection of the ObRb isoform in the hypothalamus (lane 5) and GT1-7 cells (lane 6) together with a negative control (lane 4). Note the presence of PCR products of the appropriate sizes in GT1-7 and hypothalamus (465 bp ObR and 647 bp ObRb). B, PI3K activity associated with IRS-2 in GT1-7 cells stimulated with 50 nM leptin. PI3K activity was measured in immunoprecipitates and was quantitated using a Phosphoimager. Data are mean ± SEM for 4 experiments. ** P < 0.01.
Figure 8 Leptin disrupts cortical actin filaments in GT1-7 cells Cultured GT1-7 cells were incubated in the absence and presence of leptin (10 nM) ± jasplakinolide (100 nM) for 30 minutes (jasplakinolide added 10 minutes prior to leptin). Following treatment cells were fixed and permeabilized, as described in the Methods, incubated with Alexa 488 conjugated phalloidin and Alexa 594 conjugated DNase I and subsequently processed for visualising F- and G-actin respectively by confocal microscopy.
Figure 9 PI3K mediates F-actin disruption in GT1-7 cells A, Cultured GT1-7 cells were incubated in the absence and presence of leptin (10 nM) ± wortmannin (10 nM) or LY 294002 (10 μM) for 30 minutes. Following treatment cells were fixed and permeabilized, as described in the Methods, incubated with Alexa 488 conjugated phalloidin and Alexa 594 conjugated DNase I and subsequently processed for visualising F- and G-actin respectively by confocal microscopy. B, Plot of average alexa 488-phalloidin fluorescence intensity (green) and alexa 594-DNase 1 (red) in fixed cells treated with 10 nM leptin (L) alone or cells treated with leptin and jasplakinolide (100 nM; J + L), LY294002 (10 μM; L + L) or wortmannin (10 nM; W + L) relative to cells untreated (C) with drug (n = 8 separate experiments, with 8 cells measured under each condition for each experiment). C, GT1-7 cells were incubated with PBS only (C), 10 nM leptin (L) or leptin and jasplakinolide (100 nM; J + L), LY294002 (10 μM; L + L) or wortmannin (10 nM; W + L). Cells were treated to extract actin pools as described in Methods and equal amounts of pool lysate were subjected to SDS-PAGE and transferred to nitrocellulose membrane. The levels of actin were detected by immunoblotting with an actin monoclonal antibody. D, Plot of average Triton-X-100 soluble (G, red), Triton-X-100 insoluble (F, green) and total actin (gray) concentration from live cells, relative to control untreated cells (n = 4 separate experiments), for data as shown in C. Error bars indicate s.e.m. and * significance of P < 0.01.
Figure 10 Re-organisation of F-actin is associated with raised PtdIns(3,4,5)P3 levels in GT1-7 cells Cultured GT1-7 cells were incubated in the absence and presence of 10 nM leptin for 20 minutes. Cells were fixed and permeabilized, as described in Methods, prior to incubation with rhodamine-conjugated phalloidin and wild type (wt) or K273A mutant (mt) PH-GRP1-GFP fusion protein for 60 minutes. Cells were subsequently processed for visualising GFP and rhodamine by confocal microscopy.
Discussion
PI3K – a pivotal enzyme in ARC signalling
Previous studies have demonstrated that leptin applied in vivo stimulates hypothalamic ObRb to increase phosphorylation of the signalling protein intermediates STAT3 and MAPK and that both leptin and insulin increase hypothalamic PI3K activity [12,29]. Here we have applied hormones directly to ARC wedges isolated from hypothalamic slices to enable improved signal detection (with respect to amplitude and temporal resolution), localisation of signalling to the arcuate nucleus and to fix external conditions so that potential compensatory changes associated with in vivo studies are obviated. Exposure of ARC wedges to leptin or insulin induced rapid (≤1 minute) phosphorylation of MAPK (ERK1 & 2 subfamilies), STAT3 and the PI3K activity indicators, PKB and its downstream target GSK3. These hormone-induced increases in phosphorylation were transient in the majority of experiments, usually lasting 1–5 minutes at ~34°C with return to control levels of phosphorylation within 30 minutes. Such rapid recovery has also been noted in other studies [13,30] and may be due to activation of endogenous phosphatases such as PTP1B curtailing this acute signalling process [14,31]. The phosphorylation of MAPK is quite modest and at present there are few data which link this pathway directly with the actions of either insulin [32] or leptin [16] on energy homeostasis, although recently it has been shown that centrally driven insulin-mediated sympathoactivation of brown adipose tissue is MAPK-dependent [33].
As expected, exposure of ARC wedges to leptin induced an increase in tyrosine phosphorylated STAT3 [11,13,29,34]. However, unexpectedly insulin also induced an increase in tyrosine phosphorylation of STAT3 in ARC neurones. In a previous study [30] in vivo application of insulin (icv) demonstrated no such change, unless leptin was co-applied. The data reported here indicate that insulin per se is capable of increasing STAT3 phosphorylation, as no exogenous leptin was present or endogenous leptin likely to remain in the ARC sections following the extensive washes and incubations prior to stimulation. This difference may be due to an increased signal to noise delivered using ARC tissue over whole hypothalamus and that rapid transient signals are more readily detectable by this method. Both leptin and insulin rapidly increased the phosphorylation of PKB and its downstream effector GSK3 in a wortmannin and LY294002 sensitive manner, indicative of increased PI3K activity in ARC neurones, in agreement with previous in vivo studies [17,35].
However, our results did not demonstrate that either leptin or insulin induced a significant increase in IRS-2-associated PI3K activity measured directly in ARC tissue. This may be due to a low signal to noise ratio, as only a (unknown) proportion of cells would be expected to respond to the hormones in the ARC tissue block, and/or that hormone mediated increases in PI3K activity are limited to plasma membrane microdomains. This is supported by the very modest increase in PI3K activity detected in GT1-7 cells when stimulated by leptin. Although hypothalamic activation of PKB by insulin has been reported previously [17], these are the first reports that leptin increases PKB activity and that both hormones increase the phosphorylation of GSK3 in the ARC. The presence of the PI3K inhibitors, wortmannin or LY294002, also reduced the leptin and insulin driven increase in MAPK phosphorylation. The mechanism by which leptin and insulin cause phosphorylation of this protein is most likely through the Ras pathway, as this protein has been demonstrated to interact directly with the catalytic subunit of PI3K [36] and inhibitors of PI3K have been reported to inhibit insulin induced increased MAPK activity, for example in rat adipocytes [37]. The insulin mediated enhanced STAT3 tyrosine phosphorylation in an interesting observation that requires further examination. Although phosphorylation of tyrosine-705 on STAT3 is a prerequisite for dimerisation and translocation of STAT3 to the nucleus [38], phosphorylation of serine-727 may also be required for maximal activation of STAT3 DNA binding [39].
Interestingly one pathway candidate for phosphorylating serine-727 is the Ras/Raf/MEK signalling cascade, and indeed a recent study has demonstrated that leptin can induce S727 phosphorylation of STAT3 in a PD98059 dependent manner in macrophages, and this is required to produce full stimulation of STAT3 [40]. Insulin mediated serine phosphorylation of STAT3 has also been reported, using transfected Chinese hamster ovary cells, to be mediated by a MEK-dependent pathway [41]. A similar mechanism in hypothalamic neurones would indicate an inter-connection between the three identified signalling pathways activated by these hormones and an important effector molecule, STAT3. Studies are underway to examine this proposal.
The importance of STAT3 signalling to the central mechanisms that control energy homeostasis has recently been directly demonstrated by transgenic mouse studies. Using a 'knock-in' strategy to induce defective STAT3 binding to ObRb [42] or a 'knock-out' strategy to ablate STAT3 from some hypothalamic neurones [43], loss or reduction in hypothalamic STAT3 signalling initiates hyperphagia, increased body weight and adiposity with alterations in glucose homeostasis. Indeed, the JAK2-STAT3 and IRS2-PI3K signalling pathways are purported to underpin the genomic and acute or membrane functions of these signalling pathways respectively [12]. Clearly, further work is required to determine the exact signalling mechanisms controlling insulin stimulated STAT3 phosphorylation in hypothalamic neurones.
Leptin and insulin signalling to KATP channels
Leptin and insulin cause inhibition, by hyperpolarization through activation of a sulphonylurea-sensitive K+ conductance, of a subset of hypothalamic neurones, defined by their acute sensitivity to changes in external glucose concentration, termed GR neurones [6,7,19]. Single channel recordings from acutely isolated ARC neurones demonstrate that both hormones activate the same K+ channel, the sulphonylurea-sensitive large conductance KATP channel. This action is rapid and independent of transcriptional events, so most likely is mediated by MAPK or PI3K signalling. Pharmacological inhibition of the MAPK pathway with PD98059 did not reverse leptin (as shown above) or insulin [7] stimulated KATP activity, abrogating this pathway from causing the hyperpolarising response. In contrast, inhibition of PI3K with either wortmannin or LY294002, reversed both leptin (as shown above) and insulin [7] raised KATP activity. Furthermore, use of the fusion protein PH-GRP1-GFP as a specific detector of PtdIns(3,4,5)P3 in isolated neurones also demonstrated that both hormones rapidly increase the cellular content of this PI3K lipid product in a sub-population of neurones. These results are consistent with class 1 PI3K [44] acting as a point of convergence for leptin and insulin signal transduction pathways to KATP channels in GR neurones. The functional significance of PI3K in the control of energy balance has been demonstrated by in vivo studies, which show that leptin [16] and insulin [17] stimulate IRS2-associated PI3K activity in the hypothalamus and pharmacological inhibition, using wortmannin and LY294002, of hypothalamic PI3K activity prevents the anorectic actions of icv leptin or insulin, whereas the MAPK inhibitor PD98059 had no effect on leptin driven attenuation of food intake [16].
Remodelling of cortical actin filaments as a leptin and insulin signalling event
Leptin and insulin stimulated KATP activity in isolated ARC neurones was also reversed, within 5–10 minutes, by the marine sponge toxin, jasplakinolide. This toxin binds to F-actin with high affinity, resulting in its stabilization and prevention of depolymerization to its monomer G-actin [24]. These data indicate that the adiposity hormones require actin filament depolymerization for KATP activation to occur. Such a mechanism is supported by reports that agents, which promote actin depolymerization, activate KATP channels in cardiac myocytes [45,46] and the insulin-secreting cell line, CRI-G1 [47]. Furthermore, in this latter study leptin stimulated KATP channel activity was also shown to depend on actin filament depolymerization. Insulin is also well documented to cause actin filament re-organization in peripheral cells associated with various functional outputs, which depend on PI3K activity, including metabolic and mitogenic effects [48]. The reversal of hormone-stimulated KATP activity by jasplakinolide was faster (5–10 minutes) than for the PI3K inhibitors (15–20 minutes). This temporal difference suggests that the site of jasplakinolide action is downstream from the PI3K signal transduction pathway to KATP channels.
However, alteration of the cellular cortical actin structure is inferred through the use of natural agents like jasplakinolide. In order to verify directly that hormone-driven structural re-arrangements did occur we decided to use the hypothalamic cell line, GT1-7, as preliminary experiments using freshly isolated neurones did not produce reliable and reproducible data due to the presence of dead and dying cells showing as false positives for hormone induced actin depolymerization. Use of this cell line also obviated any problems with identification of ObRb containing neurones and neuronal subtypes in slices. RT-PCR analysis indicates that this cell line does express the main signalling form of the leptin receptor and analysis of PI3K activity shows functional coupling of this receptor to this signalling pathway. We have shown, by cell staining of fixed cells and, independently by analysis of cellular G- and F-actin concentration from live cells, that leptin disrupts cortical actin structure by disturbing the processes that maintain the equilibrium between F-actin and G-actin, in the direction of depolymerization to G-actin. This effect of leptin was completely inhibited by the presence of either jasplakinolide or the PI3K inhibitors. In addition, there is a good temporal and spatial association between PtdIns(3,4,5)P3 production, as determined by PH-GRP1-GFP binding, and actin filament depolymerization. Thus, leptin and insulin signalling in, at least some sub-groups of hypothalamic neurones maintains a close parallel with leptin signalling in insulin-secreting cells, where it has been reported that leptin increases KATP activity by a PI3K-dependent cortical actin re-arrangement [47].
Conclusions
The effect of leptin and insulin on the phosphorylation status of various cellular signalling intermediates and on KATP channel activation in arcuate neurones indicates that both hormones activate the same signalling cascades, and can produce common outputs. The sensitivity of both KATP opening and the phosphorylation of certain intermediates to PI3K inhibition is significant as this enzyme has been previously demonstrated to play an important role in leptin and insulin mediated energy homeostasis control. Furthermore it is interesting that leptin and insulin induce rapid phosphorylation of MAPK and STAT3 as these data support the view that these hormones may influence genomic and membrane neuronal outputs by common mechanisms. The inhibition of leptin and insulin stimulation of KATP channel opening of arcuate neurones by jasplakinolide suggests a role for cytoskeletal dynamics in modulation of membrane events such as neuronal hyperpolarization. This hypothesis is further strengthened by the finding that leptin induces actin filament depolymerization in a mouse hypothalamic cell line, which is PI3K dependent, demonstrating that this cell line may be a useful model for further analysis of leptin signalling mechanisms in hypothalamic neurones.
Methods
Preparation of hypothalamic lysates and immunoblots
Male Sprague-Dawley rats (50–100 g) were killed by cervical dislocation in accordance with Schedule 1 of the UK Government Animals (Scientific Procedures) Act (1986). The brain was rapidly transferred to ice-cold aCSF solution, containing (in mM): 128 NaCl, 5 KCl, 1.2 NaH2PO4, 26 NaHCO3, 1.2 CaCl2, 2.4 MgSO4, and 10 glucose, equilibrated with 95% O2, 5% CO2 to give a pH of 7.4. The tissue was maintained in ice-cold aCSF whilst horizontal 400 μm coronal brain slices were prepared using a Vibratome (Intracel, Royston, Herts. UK). Slices containing the ARC were incubated in aCSF at room temperature for 20 minutes, and then at 33–35°C for 1 hour. Hypothalamic wedges, predominantly containing the ARC were cut, and these were incubated in aCSF ± hormones and/or kinase inhibitors (10 mls) for the required time. The reaction was stopped by the addition of 2 ml of cold lysis buffer containing (in mM) 100 NaCl, 10 NaF, 25 Tris HCl, 10 NaPPi, 5 EGTA, 1 EDTA, 1 Na3VO4, 1 Benzamidine, 0.1 PMSF, 0.1% (v/v) mercaptoethanol, 1% Tritron X-100 (v/v) and 92 mg ml-1 sucrose. The tissue was homogenised on ice, the lysate sonicated for two 10 s periods and then centrifuged for 10 minutes at 12000 rpm at 4°C. The supernatant was retained and the pellet discarded. The protein content of the clarified lysate was determined by the method of Bradford [49]. Proteins (10 μg) were separated by SDS-PAGE, and subsequently transferred to nitrocellulose membranes. Membranes were incubated in blocking buffer (10% non-fat dried milk in TBST (20 mM Tris HCl, 150 mM NaCl, 0.5% Tween, pH 7.4)) for 1 hour at room temperature following which phospho-specific p44/p42 MAPK (Thr202/Tyr204), phospho-specific STAT3 (Tyr705), phospho-specific GSK-3α/β(Ser21/9), phospho-specific PKB (Thr308) and PKB (all polyclonal and used at 1:1000) antibodies were applied overnight at 4°C with gentle shaking. All antibodies were obtained from Cell Signalling Technology Inc. The membranes were washed four times with TBST and incubated for 1 hour at room temperature with horseradish peroxidase conjugated Goat anti-Rabbit IgG (1:5000). After further washing with TBST, total amount of specific protein was visualised by enhanced chemiluminescence detection as described by the manufacturer (NEN Life Science Products). Immunoreactive bands were scanned and quantified using AIDA software. As an internal control, the membranes were immunoblotted with a monoclonal anti β-actin antibody (Sigma: used at 1:5000) or with the PKB antibody. The values for proteins were normalized with respect to the internal control to account for variations in gel loading.
Determination of PI 3-kinase activity
Cell and tissue lysates were made as described. The immunoprecipitation and PI3K activity assay were carried out as previously described [50]. Briefly, frozen samples were thawed before centrifugation to remove precipitated material. 10 μl Protein-G-Sepharose beads pre-coupled to 5 μg anti-IRS2 antibody (Upstate Biotechnology) was used to immunoprecipitate PI3K activity from ~0.5 mg cell lysate. The immunoprecipitated material was washed once with ice cold lysis buffer and three times with ice cold assay buffer, both of which were freshly supplemented with protease inhibitors, reducing agent and sodium vanadate as described [51]. Washed beads were re-suspended in 40 μl assay buffer supplemented with 1 μM unlabelled ATP, 25 μCi/assay radiolabelled ATP and phosphatidylinositol/phosphatidylethanolamine vesicles (final concentration of each lipid 100 μM). Samples were incubated at 37°C for 30 mins and the reaction was stopped by addition of 0.6 ml methanol/chloroform/12 M HCl (80:40:1, v/v), 0.2 ml chloroform and 0.32 ml 0.1 M HCl. Samples were processed and PtdIns(3)P separated from contaminating materials by thin layer chromatography (TLC) as previously described [51]. Bands corresponding to [32P]PtdIns(3)P were located using a phosphorimager (Fuji FLA 5000) and analyzed with AIDA software.
Preparation of acutely isolated ARC neurones and electrophysiology
Coronal slices containing the medial hypothalamus were obtained (as described above) and sections containing the ARC were removed. The sections were transferred to 5 ml aCSF containing 1 mg ml-1 protease XIV (Sigma-Aldrich, Dorset, U.K) and incubated for 1 hour at room temperature. The aCSF was continuously gassed with 95% O2: 5% CO2 for the entire incubation period. Sections were removed and washed in 50 ml aCSF five times prior to re-suspension in 5 ml normal saline containing (in mM): 135 NaCl, 5 KCl, 1 MgCl2, 1 CaCl2, 10 HEPES, 3 glucose, pH 7.4. Sections were sequentially triturated with fire polished Pasteur pipettes with decreasing tip size. The cell suspension was evenly distributed onto concanavalin A (Sigma-Aldrich) pre-treated 35 mm diameter culture dishes. The culture dishes were left for 15–20 minutes allowing cell adhesion prior to use.
Cell-attached single channel currents were recorded from single neurones at room temperature, using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA USA). Patch pipettes were prepared from thick walled borosilicate glass and had open tip resistances of 8 – 15 MΩ when filled with high K+ solution containing (in mM) 140 KCl, 1 MgCl2, 1 CaCl2, 10 HEPES, pH 7.2. This solution was used in order to allow easy identification of K+ currents in the cell-attached configuration [19]. All cell-attached recordings were made in the presence of normal saline, with no applied pipette potential, thus utilizing the cell membrane potential to drive current flow (with inward current shown as downward deflections). Single channel recordings from inside-out patches isolated from ARC neurones were made either under asymmetrical conditions, in the presence of normal saline, or under symmetrical K+ conditions with the intracellular aspect of the membrane exposed to a bathing solution containing (in mM): 140 KCl, 1 MgCl2, 2.7 CaCl2, EGTA 10 (free Ca2+ of 100 nM), HEPES 10, pH 7.2.
Data were recorded onto digital audio-tape using a Biologic DTR 1200 recorder and analysed off-line. Pre-recorded data were transferred via a Digidata 1200 interface into a PC, digitised at 10 kHz and measured using the PCLAMP6 software, Fetchan 6. The mean current (I) and single channel amplitude (i) were determined for recordings ranging in duration from 30 s to 120 s and channel activity (Nf.Po) determined as described previously [52], where Nf is the number of functional channels and Po is the open probability. Drug effects were measured by comparison of Nf.Po from individual patches in the presence and absence of the drug. Data for a given set of experiments were normalised and statistical significance determined by employing the Students t-test for unpaired data. Results are presented as mean ± SEM and the number of experiments denoted by 'n.'
Leptin receptor mRNA expression
Reverse transcription was performed in a 20 μl reaction containing 1 × First Strand Buffer, 1 mM DTT, 0.5 mM of dNTP, 0.5 μg anchored oligo(dT)18, 4 μg RNA and 1 μl (200 U) M-MLV Reverse Transcriptase (Gibco), at 25°C for 5 minutes, 42°C for 60 minutes, 70°C for 15 minutes and stored at -20°C. After RT, a 2 μl aliquot of the reaction was added to 48 μl of PCR mix. The mix containing 1 × PCR buffer, 2.5 mM MgCl2, 0.5 mM PCR nucleotide mix, 1 μM each of the gene specific primers (mObRcom F:ggaatgagcaaggtcaaaa; mObRcom R:gtgacttccatatgcaaacc; mObRb F:tcttctggagcctgaacccatttc; mObRb R:ttctcaccagaggtccctaaact; ref [53]) and 5 units of Taq DNA polymerase (Promega). PCR was performed using the following profile: 94°C for 5 minutes, 25 cycles at 94°C for 30 seconds, 55°C for 30 seconds, 72°C for 30 seconds, with a final extension at 72°C for 7 minutes.
GT1-7 cell culture, staining and actin analysis
The mouse hypothalamic cell line GT1-7 [54] was grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Sigma), 1 mM L-glutamine and 1% penicillin-streptomycin at 37°C in a humidified atmosphere of 95% air and 5% CO2. Cells were passaged every 3–4 days, plated on poly-L-lysine (Sigma) coated glass coverslips in 3.5 cm Petri dishes and used 1–2 days after plating. Cells were treated with 10 μM LY294002 or 10 nM wortmannin or 100 nM jasplakinolide (all Sigma) in normal saline for 10 minutes, prior to a challenge with 10 nM leptin (or saline), in the continuous presence of inhibitor, for 20 or 60 minutes before fixing. GT1-7 cells were fixed in 4% methanol-free formaldehyde in cytoskeletal buffer (10 mM MES, 3 mM MgCl2, 138 mM KCl, 2 mM EGTA, pH 6.1) with 0.32 M sucrose for 30 minutes [28]. They were then washed in phosphate-buffered saline (PBS), permeabilised in PBS/0.5% Triton X-100 for 10 minutes, rinsed in PBS, blocked with 20% goat serum (Sigma) for 30 minutes, rinsed in PBS and incubated with rhodamine conjugated phalloidin, or 2 μg ml-1 Alexa 594-DNase I and 2 U ml-1 Alexa 488-phalloidin (all Molecular Probes) for 90–120 minutes, rinsed in PBS, and mounted on coverslips. Cells were observed with a 63X oil objective and images acquired using a laser-scanning confocal microscope (Zeiss LSM 510), under identical conditions with randomly selected regions of each coverslip. For quantitative analysis of G- and F-actin cellular pools, we used a direct method to partition the actin pools from live cells [28]. In brief, equal cell numbers were added to 3.5 cm culture dishes and cells grown to 80% confluence. The Triton-X-100 soluble (G-actin) pool was isolated first, by incubating cells for 5 minutes at room temperature with 1 ml PBS containing 1% Triton-X-100, protease inhibitors and 1 μg ml-1 phalloidin (to prevent filament dissociation). Cells were then washed with PBS, and the Triton-X-100 insoluble pool (F-actin) prepared by addition of 1 ml of PBS lysis buffer, containing 1% Triton-X-100, protease inhibitors, 2% SDS and 1 μg ml-1 phalloidin for 5 minutes prior to harvesting cells from dishes. For determination of total actin, cells were exposed to the second step only. Each cellular pool was passed through a 25 gauge needle and total protein concentration determined, before equal amounts of protein were loaded onto SDS-PAGE gels, and actin detected using an actin monoclonal antibody (Chemicon). Quantitative measurements of G- and F-actin in fixed cells were made using Velocity software (Improvision), where individual cell total fluorescence, normalized to cell area, was determined and background fluorescence subtracted. Average fluorescence intensity was calculated for 8 cells in each experiment, and expressed relative to control (non-drug exposed cells). Actin bands on gels were quantified by densitometry, where total density was determined with respect to constant area, background subtracted and average relative band density calculated.
PH-GRP1-GFP fusion protein overlays
Following stimulation with hormone for 10–20 minutes, acutely isolated neurones (room temperature) or GT1-7 cells (37°C) were fixed at room temperature with 2–4% paraformaldehyde for 15 and 30 mins, respectively. Cells were permeabilized by washing with 0.05% PBS-Tween 20 (PBS-T; x2 for 10 mins). Non-specific binding was minimised by blocking with 3% BSA for 1 hour at room temperature. Cells were subsequently washed with 0.05% PBS-T prior to incubation with wild type PH-GRP1-GFP or K273A mutant PH-GRP1-GFP (50 μg ml-1) fusion protein for 1 hour at room temperature, and images acquired by confocal microscopy.
List of abbreviations
ACSF, artificial cerebrospinal fluid; AgRP, agouti-related protein; ARC, arcuate nucleus; NPY, neuropeptide Y; POMC, proopiomelanocortin; KATP, ATP-sensitive potassium channel; BBB, blood-brain-barrier; DNase I, deoxyribonuclease I; F-actin, filamentous actin; G-actin, globular actin; GFP, green fluorescent protein; GR neurone, glucose-responsive neurone; GRP1, general receptor for phosphoinositides-1; GSK3, glycogen synthase kinase 3; IRS2, insulin receptor substrate 2; JAK2, janus kinase 2; MAPK, mitogen-activated protein kinase; MEK, MAPK kinase; ObR, leptin receptor; PH domain, pleckstrin homology domain; PI3K, phosphoinositide 3-kinase; PKB, protein kinase B; PtdIns(3,4,5)P3, phosphatidylinositol 3,4,5-trisphosphate; STAT3, signal transducer and activator of transcription 3;
Authors' contributions
SM carried out the electrophysiology studies, participated in the western blot and protein overlay experiments. HL carried out the arcuate western blot experiments and PI3K activity measurements. KN carried out the actin imaging and actin quantitative analysis experiments. EA contributed to the actin imaging experiments, western blots and participated in the protein overlay experiments. LB performed all tissue culture and participated in the western blot experiments. AG made the fusion proteins and participated in the design of the overlay experiments. CS participated in the design and implementation of the western blot experiments. MA conceived of the study, participated in its design and co-ordination and drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
Supported by the Wellcome Trust (068692), Tenovus Scotland and Biovitrum. CS is a Diabetes (UK) Senior Fellow.
==== Refs
Schwartz MW Woods SC Porte D Seeley RJ Baskin DG Central nervous system control of food intake Nature 2000 404 661 671 10766253
Woods SC Seeley RJ Adiposity signals and the control of energy homeostasis Nutrition 2000 16 894 902 11054594 10.1016/S0899-9007(00)00454-8
Ahima RS Saper CB Flier JS Elmquist JK Leptin regulation of neuroendocrine systems Front Neuroendocrinol 2000 21 263 307 10882542 10.1006/frne.2000.0197
Spiegelman BM Flier JS Obesity and the regulation of energy balance Cell 2001 104 531 543 11239410 10.1016/S0092-8674(01)00240-9
Cowley MA Hypothalamic melanocortin neurons integrate signals of energy state Eur J Pharmacol 2003 480 3 11 14623345 10.1016/j.ejphar.2003.08.087
Spanswick D Smith MA Groppi V Logan SD Ashford MLJ Leptin inhibits hypothalamic neurons by activation of ATP-sensitive potassium channels Nature 1997 390 521 525 9394003 10.1038/37379
Spanswick D Smith MA Mirshamsi S Routh VH Ashford MLJ Insulin activates ATP-sensitive K+ channels in hypothalamic neurones of lean, but not obese rats Nat Neurosci 2000 3 757 758 10903566 10.1038/77660
Flier JS Obesity wars:molecular progress confronts an expanding epidemic Cell 2004 116 337 350 14744442 10.1016/S0092-8674(03)01081-X
Cummings DE Schwartz MW Genetics and pathophysiology of human obesity Annu Rev Med 2003 54 453 471 12414915 10.1146/annurev.med.54.101601.152403
Sweeney G Leptin signalling Cell Signal 2002 14 655 663 12020765 10.1016/S0898-6568(02)00006-2
Sahu A Leptin signalling in the hypothalamus: emphasis on energy homeostasis and leptin resistance Front Neuroendocrinol 2003 24 225 253 14726256 10.1016/j.yfrne.2003.10.001
Niswender KD Schwartz MW Insulin and leptin revisited: adiposity signals with overlapping physiological and intracellular capabilities Front Neuroendocrinol 2003 24 1 10 12609497 10.1016/S0091-3022(02)00105-X
Hegyi K Fülöp K Kovács K Tóth S Falus A Leptin-induced signal transduction pathways Cell Biol Int 2004 28 159 169 14984741 10.1016/j.cellbi.2003.12.003
Saltiel AR Pessin JE Insulin signalling pathways in time and space Trends Cell Biol 2002 12 65 71 11849969 10.1016/S0962-8924(01)02207-3
Khan AH Pessin JE Insulin regulation of glucose uptake: a complex interplay of intracellular signalling pathways Diabetologia 2002 45 1475 1483 12436329 10.1007/s00125-002-0974-7
Niswender KD Morton GJ Stearns WH Rhodes CJ Myers MG Schwartz MW Intracellular Signaling: Key enzyme in leptin-induced anorexia Nature 2001 413 794 795 11677594 10.1038/35101657
Niswender KD Morrison CD Clegg DJ Olson R Baskin DG Myers MG Seeley RJ Schwartz MW Insulin activation of phosphatidylinositol 3-kinase in the hypothalamic arcuate nucleus. A key mediator of insulin-induced anorexia Diabetes 2003 52 227 231 12540590
Lizcano JM Alessi DR The insulin signalling pathway Curr Biol 2002 12 R236 R238 11937037 10.1016/S0960-9822(02)00777-7
Ashford MLJ Boden PR Treherne JM Glucose-induced excitation of hypothalamic neurons is mediated by ATP-sensitive K+ channels Pflugers Arch 1990 415 479 483 2315006 10.1007/BF00373626
Bottomley MJ Salim K Panayotou G Phospholipid-binding protein domains Biochim Biophys Acta 1998 1436 165 183 9838094
Klarlund JK Tsiaras W Holik JJ Chawla A Czech MP Distinct polyphosphoinositide binding selectivities for pleckstrin homology domains of GRP-1-like proteins based on diglycine versus triglycine motifs J Biol Chem 2000 275 32816 32821 10913124 10.1074/jbc.M002435200
Evans CA Tonge R Blinco D Pierce A Shaw J Lu Y Hamzah HG Gray A Downes CP Gaskell SJ Spooncer E Whetton AD Comparative proteomics of primitive hematopoietic cell populations reveals differences in expression of proteins regulating motility Blood 2004
Harvey J McKay NG Van Der Kaay J Downes CP Ashford MLJ Essential role of phosphoinositide 3-kinase in leptin-induced KATP channel activation in the rat CRI-G1 insulinoma cell line J Biol Chem 2000 275 4660 4669 10671495 10.1074/jbc.275.7.4660
Bubb MR Spector I Beyer BB Fosen KM Effects of jasplakinolide on the kinetics of actin polymerization. An explanation for certain in vivo observations J Biol Chem 2000 275 5163 5170 10671562 10.1074/jbc.275.7.5163
Magni P Vettor R Pagano C Calcagno A Beretta E Messi E Zanisi M Martini L Motta M Expression of a leptin receptor in immortalized gonadotropin-releasing hormone-secreting neurons Endocrinology 1999 140 1581 1585 10098491 10.1210/en.140.4.1581
Kaszubska W Falls HD Schaefer VG Haasch D Frost L Hessler P Kroeger PE White DW Jirousek MR Trevillyan JM Protein tyrosine phosphatase 1B negatively regulates leptin signalling in a hypothalamic cell line Mol Cell Endocrinol 2002 195 109 118 12354677 10.1016/S0303-7207(02)00178-8
Pollard TD Borisy GG Cellular motility driven by assembly and disassembly of actin filaments Cell 2003 112 453 465 12600310 10.1016/S0092-8674(03)00120-X
Cramer LP Briggs LJ Dawe HR Use of fluorescently labelled deoxyribonuclease I to spatially measure G-actin levels in migrating and non-migrating cells Cell Motil Cytoskeleton 2002 51 27 38 11810694 10.1002/cm.10013
Bates SH Myers MG The role of leptin receptor signalling in feeding and neuroendocrine function Trends Endocrinol Metab 2003 14 447 452 14643059 10.1016/j.tem.2003.10.003
Carvalheira JBC Siloto RMP Ignacchitti I Brenelli SL Carvalho CRO Leite A Velloso LA Gontijo JAR Saad MJA Insulin modulates leptin-induced STAT3 activation in rat hypothalamus FEBS Lett 2001 500 119 124 11445068 10.1016/S0014-5793(01)02591-1
Zabolotny JM Bence-Hanulec KK Stricker-Krongrad A Haj F Wang Y Minokoshi Y Kim Y-B Elmquist JK Tartaglia LA Khan BB Neel BG PTP1B regulates leptin signal transduction in vivo Dev Cell 2002 2 489 495 11970898 10.1016/S1534-5807(02)00148-X
Lazar DF Wiese RJ Brady MJ Mastick CC Waters SB Yamauchi K Pessin JE Cuatrecasas P Saltiel AR Mitogen-activated protein kinase inhibition does not block the stimulation of glucose utilization by insulin J Biol Chem 1995 270 20801 20807 7657664 10.1074/jbc.270.35.20801
Rahmouni K Morgan DA Morgan GM Liu X Sigmund CD Mark AL Haynes WG Hypothalamic PI3K and MAPK differentially mediate regional sympathetic activation to insulin J Clin Invest 2004 114 652 658 15343383 10.1172/JCI200421737
Zabeau L Lavens D Peelman F Eyckerman S Vandekerckhove J Tavernier J The ins and outs of leptin receptor activation FEBS Lett 2003 546 45 50 12829235 10.1016/S0014-5793(03)00440-X
Zhao AZ Huan J-N Gupta S Pal R Sahu A A phosphatidylinositol 3-kinase-phosphodiesterase 3B-cyclic AMP pathway in hypothalamic action of leptin on feeding Nat Neurosci 2002 5 727 728 12101402
Rodriguez-Viciana P Warne PH Dhand R Vanhaesebroeck B Gout I Fry MJ Waterfield MD Downward J Phosphatidylinositol-3-OH kinase as a direct target of ras Nature 1994 370 527 532 8052307 10.1038/370527a0
Sajan MP Standaert ML Bandyopadhyay G Quon MJ Burke TR Farese RV Protein kinase C-ζ and phosphoinositide-dependent protein kinase-1 are required for insulin-induced activation of ERK in rat adipocytes J Biol Chem 1999 274 30495 30500 10521430 10.1074/jbc.274.43.30495
Ihle JN The Stat family in cytokine signaling Curr Opin Cell Biol 2001 13 211 217 11248555 10.1016/S0955-0674(00)00199-X
Decker T Kovarik P Serine phosphorylation of STATs Oncogene 2000 19 2628 2637 10851062 10.1038/sj.onc.1203481
O'Rourke L Shepherd PR Biphasic regulation of extracellular-signal-regulated protein kinase by leptin in macrophages: role in regulating STAT3 Ser727 phosphorylation and DNA binding Biochem J 2002 364 875 879 12049654 10.1042/BJ20020295
Ceresa BP Horvath CM Pessin JE Signal transducer and activator of transcription-3 serine phosphorylation by insulin is mediated by a Ras/Raf/Mek-dependent pathway Endocrinology 1997 138 4131 4137 9322921 10.1210/en.138.10.4131
Bates SH Stearns WH Dundon TA Schubert M Tso AWK Wang Y Banks AS Lavery HJ Haq AK Maratos-Flier E Neel BG Schwartz MW Myers MG STAT3 signalling is required for leptin regulation of energy balance but not reproduction Nature 2003 421 856 859 12594516 10.1038/nature01388
Cui Y Huang L Elefteriou F Yang G Shelton JM Giles JE Oz OK Pourbahrami T Lu CYH Richardson JA Karsenty G Li C Essential role of STAT3 on body weight and glucose homeostasis Mol Cell Biol 2004 24 258 269 14673160 10.1128/MCB.24.1.258-269.2004
Shepherd PR Withers DJ Siddle K Phosphoinositide 3-kinase: the key switch mechanism in insulin signalling Biochem J 1998 333 471 490 9677303
Terzic A Kurachi Y Actin microfilament disrupters enhance K (ATP) channel opening in patches from guinea-pig cardiomyocytes J Physiol 1996 492 395 404 9019537
Yokoshiki Y Katsube Y Suinagawa M Seki T Sperelakis N Disruption of actin cytoskeleton attenuates sulphonylurea inhibition of cardiac ATP-sensitive K+ channels Pflugers Archiv 1997 434 203 205 9136675 10.1007/s004240050384
Harvey J Hardy SC Irving AJ Ashford MLJ Leptin activation of ATP-sensitive K+ (KATP) channels in rat CRI-G1 insulinoma cells involves disruption of the actin cytoskeleton J Physiol 2000 527 95 107 10944173 10.1111/j.1469-7793.2000.00095.x
Tsakiridis T Tong P Matthews B Tsiani E Bilan PJ Klip A Downey GP Role of actin cytoskeleton in insulin action Microsc Res Tech 1999 47 79 92 10523787
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 10.1006/abio.1976.9999
Batty IH Fleming IN Downes CP Muscarinic-receptor-mediated inhibition of insulin-like growth factor-1 receptor-stimulated phosphoinositide 3-kinase signalling in 1321N1 astrocytoma cells Biochem J 2004 379 641 51 14769130 10.1042/BJ20031700
Herbert TP Kilhams GR Batty IH Proud CG Distinct signalling pathways mediate insulin and phorbol ester-stimulated eukaryotic initiation factor 4F assembly and protein synthesis in HEK 293 cells J Biol Chem 2000 275 11249 56 10753934 10.1074/jbc.275.15.11249
Lee K Rowe ICM Ashford MLJ Characterisation of an ATP-modulated large conductance Ca2+-activated K+ channel present in rat cortical neurones J Physiol 1995 488 319 337 8568673
Morton NM Emilsson V Liu YL Cawthorne MA Leptin action in intestinal cells J Biol Chem 1998 273 26194 26201 9748302 10.1074/jbc.273.40.26194
Mellon PL Windle JJ Goldsmith PC Padula CA Roberts JL Weiner RI Immortalization of hypothalamic GnRH neurons by genetically targeted tumorigenesis Neuron 1990 5 1 10 2196069 10.1016/0896-6273(90)90028-E
| 15581426 | PMC539348 | CC BY | 2021-01-04 16:39:08 | no | BMC Neurosci. 2004 Dec 6; 5:54 | utf-8 | BMC Neurosci | 2,004 | 10.1186/1471-2202-5-54 | oa_comm |
==== Front
BMC Nucl MedBMC Nuclear Medicine1471-2385BioMed Central London 1471-2385-4-31557920810.1186/1471-2385-4-3Research ArticleEvaluation of the clinical value of bone metabolic parameters for the screening of osseous metastases compared to bone scintigraphy Schoenberger Johann [email protected] Silke [email protected] Eva [email protected] Christoph [email protected] Department of Nuclear Medicine, University of Regensburg, Germany2004 4 12 2004 4 3 3 5 8 2004 4 12 2004 Copyright © 2004 Schoenberger et al; licensee BioMed Central Ltd.2004Schoenberger 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
Bone metastases are common in many types of cancer. As screening methods different imaging modalities are available. A new approach for the screening of osseous metastases represents the measurement of bone metabolic markers. Therefore aim of this study was to evaluate the usefulness of the determination of bone metabolic markers aminoterminal propeptide of type I procollagen (PINP, osteoblastic activity) and the carboxyterminal pyridinoline cross-linked telopeptide of type I collagen (ICTP, osteoclastic activity) for the detection of bone metastases associated with other malignancies.
Methods
88 patients aged 21 – 82 years with malignant tumors were prospectively studied. The serum concentrations of PINP and ICTP were measured and compared to the results of bone scintigraphy, radiological bone series, CT, MRI and clinical follow-up.
Results
Osseous metastases were found in 21 patients. 19 of them were correctly identified by bone scintigraphy (sensitivity: 90%). For bone metabolic markers results were as follows: ICTP sensitivity: 71%, specificity: 42%; PINP sensitivity: 24%, specificity: 96%.
Conclusions
As markers of bone metabolism PINP and ICTP showed low sensitivity and/or specificity for the detection of osseous metastases. The presented markers did not seem to be sufficient enough to identify patients with bone metastases or to replace established screening methods.
==== Body
Background
Bone metastases are common in advanced cancers of the lung, breast, kidney, prostate and others. In autopsy studies the prevalence ranges from 47–85% [1]. In patients with osseous metastases early detection is needed, since without effective treatment these bone metastases can cause severe complications leading to considerable morbidity and reduced quality of life.
The screening for bone metastases is usually based on bone scintigraphy, confirmed by supplemental radiographic bone surveys, computer tomography, or magnetic resonance imaging. Bone scans are simple to perform and examine the whole skeleton, but although the sensitivity of this method is high its specificity is poor. For this reason, in many cases a positive scan requires confirmation by other imaging modalities, which leads to higher costs and time-consuming investigations. Therefore, an alternative cost-effective screening method with a similar sensitivity and a higher specificity would be very welcome.
For the formation of osseous metastases the extracellular matrix consisting of collagens combined with noncollagenous glycoproteins and proteoglycans, including the basement membrane and the interstitial stroma play an important role. Normally, the extracellular matrix serves as a barrier for the attachment and invasion of malignant cells, but the proteolytic activity of tumor cells leads to the destruction of its collagenic components, thus facilitating the local invasion of malignant cells and finally the development of bone metastases [2]. The major collagen in bone is type I collagen, which is synthesized by osteoblasts and accounts for about 90% of the organic matrix [3].
Recently, bone metabolic markers have been reported to be useful in diagnosing bone metastases [4,5]. Newly developed methods are able to quantitatively determine concentrations of collagen metabolites. The synthesis of type I collagen can be analyzed by measuring the serum concentration of the aminoterminal propeptide of type I procollagen (PINP) using a specific radioimmunoassay. In addition, bone resorption can also be analyzed by a radioimmunoassay, which measures the serum concentration of the carboxyterminal pyridinoline cross-linked telopeptide of type I collagen (ICTP).
Both parameters have been identified as potential candidates for the early detection of bone metastases. Therefore, the purpose of this study was to evaluate the usefulness of PINP and ICTP in patients with newly diagnosed cancer for the screening of osseous metastases compared to current standard protocols.
Methods
Patients
For this study 88 consecutive patients (35 female; 53 male) with malignant tumors (20 lung cancer, 20 breast cancer, 19 head/neck cancer, 5 prostate cancer, 4 thyroid cancer, 3 sarcoma, 2 esophagus cancer, 2 pancreatic cancer, 2 urothel cancer, 1 gastric cancer, 2 plasmocytoma, 2 histiocytosis X, 1 melanoma, 1 rectal cancer, 1 hypernephroma, 2 carcinoma of unknown primary, 1 breast cancer and hypernephroma) between the ages of 21 and 82 were included. None of these patients were under a tumor specific therapy or presented primary bone disease such as osteoporosis or Paget's disease, which could interfere with the results of this study. All patients gave written consent to participate in this prospective study, which was approved by the local ethics committee.
Marker assays
Blood samples for measuring PINP and ICTP were collected on the same day as bone scintigraphy. Due to higher PINP values at night, all samples were taken early in the morning and stored at -20°C until assayed. Apparent hemolytic serum was excluded.
Serum concentrations of ICTP and PINP were measured by using commercially available RIA kits (both: Orion Diagnostica, Espoo, Finland). According to the kit description, the normal range of ICTP was 1.6–5.3 μg/l for females and 1.4–5.2 μg/l for males. For PINP the normal range was 19–102 μg/l for females and 21–78 μg/l for males.
Bone scanning
Two double-head gamma cameras (ECAM duet and Bodyscan; Siemens, Erlangen, Germany) and a triple-head gamma camera (Multispect III, Siemens, Erlangen, Germany) were used for planar and tomographic (SPECT) bone scans, respectively. Low-energy, high-resolution collimators (1024 × 256 matrix) were used and data acquisition was started 2–4 hrs after intravenous injection of 550–700MBq Tc99m-DPD (3,3-Diphosphono-1,2-propandicarbonacid). SPECT acquisitions were performed from suspected regions (128 × 128 matrix; 64 steps; 150000–200000 counts/step; Butterworth filter; cut-off level 0.4). The total acquisition time ranged from 100 to 130 min for planar and SPECT scans. The bone-scanning procedure was performed in accordance with procedural guidelines published by the Society of Nuclear Medicine [6].
Interpretation of bone scintigraphy
Two experienced nuclear medicine physicians interpreted planar bone and SPECT images. Initially, neither of the readers knew of the findings of the other or the results of other imaging modalities. Increased tracer-uptake located at joints or on the edge of vertebral bodies adjacent to disk spaces was interpreted as arthritis or osteophytes, respectively. Lesions were classified as fractures when they showed a typical linear and curved pattern e.g. adjacent lesions in the ribs. Multiple lesions of varying size, shape and intensity, elongated rib lesions or photopaenic areas (cold spots) were classified as osseous metastases or suspicious lesions, where further analysis or imaging methods were necessary. In general, interpretation was performed following the criteria described by Krasnow et al. [7]. Finally, any discrepant interpretations between the two readers were resolved by consensus. Patients were classified as having osseous metastases when other imaging modalities (radiographic, MRI, CT) or histological findings confirmed the diagnosis. Patients were classified as not having bone metastatic disease when no imaging technique or histological finding indicated osseous involvement. To reduce the possibility that bone metastases were not yet visible by the cited imaging methods, a clinical follow-up period of between 9–14 months was used as the gold standard in all patients. Follow-up consisted of a clinical examination, control of tumor markers, imaging techniques (e.g. computed tomography, MRI or plain radiographic), etc. Overall, follow-up was done in compliance with the guidelines for tumor patients published by the cancer society.
Data analysis
Sensitivity and specificity were calculated. Values are expressed as mean ± standard deviation. Statistical analysis was performed using SPSS-Software version 10.0 (SPSS, Inc.). The Mann-Whitney U test was applied to compare concentrations of PINP and ICTP, respectively. A value of p < 0.05 was considered to be statistically significant. Furthermore, we performed a receiver operating characteristic (ROC) analysis to assess the impact of bone metabolic markers in detecting bone metastases.
Results
Osseous metastases were found in 21 patients with the following tumors: 8 patients with breast cancer, 3 patients with head/neck cancer and local osseous tumor invasion, 3 prostatic cancer, 2 plasmocytoma, 1 each lung cancer, histiocytosis X, thyroid cancer and sarcoma and 1 patient had breast cancer and hypernephroma.
Bone scintigraphy
Using planar and tomographic bone scintigraphy, 19 patients were correctly diagnosed as patients with osseous metastases by both investigators independently and confirmed by other imaging modalities or follow-up (sensitivity: 90%). Two patients were false negative, 1 patient with histiocytosis X (diffuse infiltration of the spine and pelvis, identified by computed tomography) and 1 had breast cancer and hypernephroma (multiple osteolytic lesions with diameters up to 1.5 cm in the pelvic bone, which was correctly diagnosed by computed tomography). In 43 patients there was no evidence for osseous metastases by bone scintigraphy or other imaging modalities. In 24 patients, changes in bone scintigraphy were described not typical for osseous metastases, but additional imaging methods were recommended for verification. Most of these lesions were located in the ribs (singular focus) and the spine, typically for fractures, traumatic injuries or osteoporotic changes (compression fracture). Additional verification was recommended especially in those patients where no history of traumatic injuries, degenerative processes or osteoporosis was known. Verification was done in most cases by plain radiography, computed tomography or within the staging by FDG-PET.
ICTP
Serum ICTP was elevated above the upper reference limit (>5.2 μg/ in males and >5.3 μg/l in females) in 56 patients (31 male, 25 female). In females the mean value was 7.64 ± 4.25 μg/l. In the group with metastases (n = 10) the mean value was 10.71 ± 5.90 μg/l compared to 6.41 ± 2.66 μg/l in the other group (n = 25) without metastases (p = 0.11). In males the mean value was 8.74 ± 9.49. The patients with metastases (n = 11) showed a value of 9.23 ± 7.62 compared to 8.61 ± 9.99 in patients (n = 42) without bone metastases (p = 0.24). Figures 1 and 2 present the values of all patients, separated by male and female, showing the different range of normal values. In females sensitivity was 70% and specificity 32% and in males sensitivity was 73% and specificity 48%. For both groups combined sensitivity was 71% and specificity 42%.
Figure 1 Values of ICTP in males (Reference interval: 1.4–5.2 μg/). Red columns indicate patients with osseous metastases.
Figure 2 Values of ICTP in females (Reference interval: 1.6–5.3 μg/). Red columns indicate patients with osseous metastases.
PINP
Serum PINP was elevated above the upper reference limit (>78 μg/ in males and >102 μg/l in females) in 8 patients (5 male, 3 female). In females the mean value was 57.42 ± 38.50 μg/l. In the group of patients with metastases (n = 10) the mean value was 73.29 ± 62.38 μg/l compared to 51.07 ± 22.23 μg/l in the group (n = 25) without metastases (p = 0.95). In males the mean value was 45.59 ± 43.20. The patients with metastases showed a value of 60.70 ± 83.86 compared to 41.63 ± 23.99 in patients without bone metastases (p = 0.70). Figures 3 and 4 show the values of all patients. Sensitivity in females was 30 %, specificity 100 %, in males sensitivity was 18 %, specificity 93 %. Combined sensitivity was 24 % and specificity 96 %. As examples, figures 6 and 7 show patients with non-small-cell-lung cancer as the primary tumor.
Figure 3 Values of PINP in males (Reference interval: 21–78 μg/). Red columns indicate patients with osseous metastases.
Figure 4 Values of ICTP in females (Reference interval: 19–102 μg/). Red columns indicate patients with osseous metastases.
ROC-analysis
Neither parameter achieved statistical significance by ROC analysis. According to the ROC analysis, the optimal cut-off level of ICTP that maximizes sensitivity and specificity was 7.5 μg/l in females (60% sensitivity and 88% specificity) and 5.9 μg/l in males (73% sensitivity and 57% specificity), for PINP 85.5 μg/l in females (40% sensitivity and 96% specificity) and 25.8 μg/l in males (46% sensitivity and 74% specificity).
Figure 5 shows ROC curves and summarizes the results.
Figure 5 left side ROC curves for ICTP and PINP in males Marker AUC ± SE (95%CI) ICTP 0.62 ± 0.1 (0.47–0.75) PINP 0.54 ± 0.1 (0.40–0.68) AUC: area under curve; SE: standard error; 95%CI: 95% confidence interval right side ROC curves for ICTP and PINP in females Marker AUC ± SE (95%CI) ICTP 0.67 ± 0.1 (0.49–0.82) PINP 0.51 ± 0.1 (0.33–0.68) AUC:area under curve; SE: standard error; 95%CI: 95% confidence interval
The types of tumors in our study were very heterogeneous, which reflects the normal day-to-day situation seen in a department of nuclear medicine. To obtain additional information concerning special tumor types, we performed a separate examination of two different tumor types with a higher number of patients: breast cancer and head/neck cancer.
Breast cancer
The group with breast cancer consisted of 20 patients. Subsequently, bone metastases could be verified in eight of these patients. Six cases could be identified clearly by bone scan, in one patient additional plain radiographic analysis was recommended to confirm the diagnosis and one patient was false positive (focal area of increased tracer uptake in the shaft of the femur, identified as local necrosis of the bone by plain radiography and confirmed by follow-up). For bone markers the following results were observed:
ICTP: sensitivity: 63%, specificity: 25%. PINP: sensitivity: 25%, specificity: 100%. Figure 8 shows an example of a patient with osseous metastatic disease and normal values for ICTP and PINP.
Head and neck cancer
19 patients with head/neck cancer were examined (4 female, 15 male). In these patients determination of whether local osseous structures are involved is essential for the preoperative planning of further treatment. In three patients osseous structures were affected by the tumor and all of these were correctly diagnosed by bone scintigraphy. ICTP and PINP were right positive in two separate patients. Sensitivity for PINP and ICTP was 33%, specificity for ICTP was 56% and for PINP 94%.
Discussion
Metastatic bone disease is a serious clinical problem. Complications associated with osseous metastases are pain, fractures, spinal cord compression, paralysis, etc., which lead to a significant reduction in the quality of life of tumor patients. Several diagnostic tools are available to detect bone metastases, including plain radiography, computer tomography, magnetic resonance imaging and bone scintigraphy. Of these, radionuclide bone scanning using Tc-99m labeled diphosphonates is the most widely used and accepted method for the detection of bone metastases. Bone scintigraphy can detect osseous metastases several months before changes in plain radiographs can be seen, thus making bone scintigraphy an excellent diagnostic tool. However, this method is expensive, is not always available in every hospital and has the disadvantage of showing a positive reaction even to bone inflammation, degenerative changes and fractures or flare reaction, which leads to a reduced specificity. This is the reason why many authors have described outcomes regarding diagnosis of bone metastases and observation of the clinical course using markers of bone turnover.
In recent years, there have been important advances in the field of biochemical markers of bone turnover and new methods have emerged [8]. Measurement of the metabolites of type I collagen, the predominant collagen in bone, has been reported to be useful for monitoring bone turnover in many different disorders, including diseases with bone metastases [9].
Yoshida et al. reported on the serum concentration of type I collagen metabolites as a quantitative marker of bone metastases in patients with prostate cancer [10]. They concluded that, especially in patients with high grade carcinoma cells, the determination of bone metabolic markers should be more useful in evaluating metastatic spread to bone than prostate specific antigen. In our collective, patients with prostate cancer (n = 5) showed the highest correlation between the presence of osseous metastases and elevated markers for PINP and ICTP. Only in one case we found a false positive ICTP. However, due to the small size of our patient groups it prevents any real conclusive statements from being made.
For other tumor types we did not observe a similar high correlation. Horiguchi et al. reported on the usefulness of ICTP as a marker for bone metastases in patients with lung cancer [11]. He suggested that measurement of ICTP is an excellent serological diagnostic method for identifying bone metastases in patients with lung cancer and can also help predict when it might be useful to undertake other examinations like bone scintigraphy. In our study however, we saw a high rate (11 out of 19) of false positive results of increased ICTP levels in this group of patients. One reason for this might be the presence of non-detectable micro-metastases at the time of ICTP measurement. To avoid the possibility of false results, all patients had a follow-up examination between 9 and 14 months, during which micro-metastases would have become apparent. None of the 11 patients developed osseous metastases during this time period so that the likelihood of the presence of such metastases was very low.
Another major group of patients in our study were females with breast cancer. The literature on these tumors and the value of bone metabolic markers for detection of osseous metastases is very controversial. Blomqvist et al. reported a positive and significant correlation between ICTP and PICP and the number of bone metastases, plus Wada et al. suggested that ICPT might be a useful marker for screening and monitoring bone metastases in breast cancer [12,13]. In contrast Ulrich et al. showed in a study with 106 patients that the sensitivity for diagnosing bone metastases was 65% [14]. These results are more similar to our study. However, Ulrich reported a high specificity of 91%, whereas we observed only 25% specificity for ICTP.
The types of tumors in our study were very heterogeneous, which seems to reflect the circumstances seen on a daily basis. Due to the fact that tumors can metastasize to bone in different ways (osteoblastic and/or osteolytic) we established parameters for both possibilities. For PINP, the marker for osteoblastic activity, the specificity was high but with a poor sensitivity. For ICTP, the marker for osteolytic activity, both characteristics of sensitivity and specificity were low, such that a general recommendation for the use of this marker as a screening parameter cannot be made.
Potential indications for bone metabolic markers might be the therapy control in patients with bone metastases, in which increased parameters had been proven before therapy. Particularly in these patients, the so-called "bone flare reaction", the repair of destroyed bone structures by tumor cells, complicates the assessment of bone scintigraphy. Increased activity in known osseous metastases after or during therapy can be caused by either repair or further tumor growth. Blomqvist and colleagues and Koizumi et al. reported on patients with breast cancer and osseous metastases, where only patients with progressive disease showed an increase in ICTP values during therapy compared to other patients with response to therapy [9,15]. Another indication might be in benign bone disorders like rheumatoid arthritis or Paget's disease. Aman et al. reported a correlation between ICTP values and the disease progression of patients with rheumatoid arthritis [16].
Conclusions
In summary, when a new method is recommended for the diagnostic work up, it is necessary to demonstrate that this new modality is as sensitive and specific as existing routine imaging procedures. The determination of bone metabolic parameters like ICTP or PINP is less expensive than bone scanning, but this prospective study has shown that the results from bone metabolic markers are not yet sufficient enough to demonstrate a bone involvement in different type of malignancies with high sensitivity and specificity.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JS and CE designed the study, performed the statistical analysis and drafted the manuscript.
SR and EW recruited patients and carried out the nuclear medicine investigations.
All authors read and approved the final manuscript.
Figure 6 Bone scan of a 59-year-old female with non-small-cell lung cancer and multiple osseous metastases. Both parameters are increased ICTP: 13 μg/L (1.6–5.3) PINP: 113.8 μg/L (19–102)
Figure 7 Bone scan of a 47-year-old male with non small cell lung cancer. Bone scintigraphy and follow-up showed no evidence of osseous metastatic disease. ICTP: 7.3 μg/L (1.4–5.2) and PINP: 102 μg/L (21–78) are above the upper reference limit.
Figure 8 55-year-old female with metastatic breast cancer. On the bone scan multiple osseous metastases can be seen especially in the spine, pelvic region and calotte. ICTP: 4.8 μg/L (1.6–5.3) and PINP: 31.5 μg/L (19–102) are within the reference limit.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank Mrs. Henriette Dam, who carried out the radio-immunoassays, for her excellent technical assistance.
==== Refs
Galasko C Weiss L, Gilbert AH The anatomy and pathways of skeletal metastases In Bone metastases 1981 Boston, MA: GK Hall 49 63
Terranova PV Hujanen ES Martin G Basement membrane and the invasive activity of metastatic tumor cells J Natl Cancer Inst 1986 77 311 316 3525956
Prockop DJ Kivirikko KI Tuderman L Guzman NA The biosynthesis of collagen and its disorders NEJM 1979 301 13 23 449904
Koizumi M Yamada Y Takiguchi T Nomura E Furukawa M Kitahara T Yamashita T Maede H Takahashi S Aiba K Bone metabolic markers in bone metastases J Cancer Res Clin Oncol 1995 121 542 548 7559734
Lorente JA Morote J Raventos C Encabo G Valenzuela H Clinical efficacy of bone alkaline phosphatase and prostatic specific antigen in the diagnosis of bone metastasis in prostate cancer J Urol 1996 155 1348 1351 8632571 10.1097/00005392-199604000-00060
Donohoe KJ Henkin RE Royal HD Brown ML Collier BD O'Mara RE Carretta RF Procedure guideline for bone scintigraphy: 1.0. Society of Nuclear Medicine JNM 1996 37 1903 1906 8917200
Krasnow AZ Hellman RS Timins ME Collier D Anderson T Isitman AT Diagnostic bone scanning in oncology Semin Nucl Med 1997 27 107 141 9144855
Delmas PD Papapoulos SE, Lips P, Pols HAP, Johnston CC, Delmas PD Biochemical markers of bone remodeling in osteoporosis Osteoporosis 96. International Congress, series 1118 1996 Amsterdam: Elsevier 191 204
Blomqvist C Risteli I Risteli J Virkkunen P Sarna S Elomaa I Markers of type I collagen degradation and synthesis in the monitoring of treatment response in bone metastases from breast carcinoma Br J Cancer 1996 73 1074 1079 8624266
Yoshida K Sumi S Arai K Koga F Umeda H Hosoya Y Honda M Yano M Moriguchi H Kitahara S Serum concentration of type I collagen metabolites as a quantitative marker of bone metastases in patients with prostate cancer Cancer 1997 80 1760 1767 9351545
Horiguchi T Tachikawa S Kondo R Hirose M Teruya S Ishibashi A Banno K Usefulness of serum Carboxy-terminal telopeptide of type I collagen (ICTP= as a marker of bone metastases from lung cancer Jpn J Clin Oncol 2000 30 174 179 10830985 10.1093/jjco/hyd043
Blomqvist C Risteli L Risteli J Virkkunen P Sarna S Elomaa I Markers of type I collagen degradation and synthesis in the monitoring of treatment response in bone metastases from breast carcinoma Br J Cancer 1996 73 1074 1079 8624266
Wada N Fujisaki M Ishii S Ikeda T Kitajima M Evaluation of bone metabolic markers in breast cancer with bone metastases Breast Cancer 2001 8 131 137 11342986
Ulrich U Rhiem K Schmolling J Flaskamp C Paffenholz I Salzer H Bauknecht T Schlebusch H Cross-linked type I collagen C- and N-telopeptides in women with bone metastases from breast cancer Arch Gynecol Obstet 2001 264 186 190 11205705 10.1007/s004040000105
Koizumi M Matsumoto S Takahashi S Yamashita T Etsuro O Bone metabolic markers in the evaluation of bone scan flare phenomenon in bone metastases of breast cancer Clin Nucl Med 1999 24 15 20 9890487 10.1097/00003072-199901000-00004
Aman S Paimela L Leirisalo-Repo M Risteli J Kautiainen H Helve T Hakala M Prediction of disease progression in early rheumatoid arthritis by ICTP, RF and CRP. A comparative 3-year follow-up study Rheumatology 2000 39 1009 1013 10986307 10.1093/rheumatology/39.9.1009
| 15579208 | PMC539349 | CC BY | 2021-01-04 16:30:50 | no | BMC Nucl Med. 2004 Dec 4; 4:3 | utf-8 | BMC Nucl Med | 2,004 | 10.1186/1471-2385-4-3 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-4-611559834510.1186/1471-2458-4-61Research ArticleThe effect of cigarette price increase on the cigarette consumption in Taiwan: evidence from the National Health Interview Surveys on cigarette consumption Lee Jie-Min [email protected] Tsorng-Chyi [email protected] Chun-Yuan [email protected] Sheng-Hong [email protected] Department of Logistics Management, National Kaohsiung Marine University, Kaohsiung, Taiwan2 Department of Applied Economics, National Chung Hsing University, Taichung, Taiwan3 Department of International Trade, Overseas Chinese Institute of Technology, Taichung, Taiwan2004 14 12 2004 4 61 61 20 6 2004 14 12 2004 Copyright © 2004 Lee et al; licensee BioMed Central Ltd.2004Lee 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 study uses cigarette price elasticity to evaluate the effect of a new excise tax increase on cigarette consumption and to investigate responses from various types of smokers.
Methods
Our sample consisted of current smokers between 17 and 69 years old interviewed during an annual face-to-face survey conducted by Taiwan National Health Research Institutes between 2000 to 2003. We used Ordinary Least Squares (OLS) procedure to estimate double logarithmic function of cigarette demand and cigarette price elasticity.
Results
In 2002, after Taiwan had enacted the new tax scheme, cigarette price elasticity in Taiwan was found to be -0.5274. The new tax scheme brought about an average annual 13.27 packs/person (10.5%) reduction in cigarette consumption. Using the cigarette price elasticity estimate from -0.309 in 2003, we calculated that if the Health and Welfare Tax were increased by another NT$ 3 per pack and cigarette producers shifted this increase to the consumers, cigarette consumption would be reduced by 2.47 packs/person (2.2%). The value of the estimated cigarette price elasticity is smaller than one, meaning that the tax will not only reduce cigarette consumption but it will also generate additional tax revenues. Male smokers who had no income or who smoked light cigarettes were found to be more responsive to changes in cigarette price.
Conclusions
An additional tax added to the cost of cigarettes would bring about a reduction in cigarette consumption and increased tax revenues. It would also help reduce incidents smoking-related illnesses. The additional tax revenues generated by the tax increase could be used to offset the current financial deficiency of Taiwan's National Health Insurance program and provide better public services.
==== Body
Background
One of the problems in controlling tobacco in Taiwan is that cigarette prices are lower in Taiwan than in most countries [1,2]. In India, smokers have to work 77 minutes to afford a pack of cigarettes, in Indonesia 62 mins, in China 56 mins, but in Taiwan they need only work 7 to 10 mins to afford a pack [1]. As long as the domestic cigarette price remains rather low, we probably will not see much of a decrease in the number of smokers.
The smoking population increased to 4.5 million persons (total population 22,520,776) in 2002 [3]. One out of three adults smoked. Currently, due to illnesses and death associated with cigarette smoking, smokers currently account for approximately 20 billion NT dollar of extra medical expense annually and account for a 160 billion NT dollar loss in GDP (Gross Domestic Product)[4]. This economic burden in putting pressure on the government to further increase the existing Health and Welfare Tax on tobacco.
Because high excise taxes on cigarettes have been found able to reduce cigarette consumption [5-8], such measures are becoming one of the most important means of controlling tobacco [9-11]. The new tax scheme enacted on 1 January 2002 in Taiwan resulted in NT $16.8 tax excise. This tax included the existing taxes for NT $11.8 and a NT $5 Health & Welfare Tax for a 20-pack of cigarettes. The government also levies 5% sales taxes. Under that tax scheme, the cigarette tax revenues account for 40% of the retail price, which is about NT $42.2. While 40% sounds high, it is actually lower than the taxes imposed on cigarettes in developed countries that have seen some success at lowering cigarette consumption.
The government should take elasticity of demand into consideration when deciding whether to increase or add an excise tax levy. If there is a price elasticity below 1, a tax increase brings about a decline in consumption and an increase in total tax revenues. Since Hsieh, Hu and Lin (1999) found that figure to be -0.6 in Taiwan, it can be reasonable to assume that a tax increase would be more likely to reduce cigarette consumption more significantly in Taiwan than in other countries with lower price elasticities at least in the short-term and medium term [12]. Higher taxes would also generate higher total tax revenues.
Taiwan's Tobacco and Wine Tax Law is currently under review in the Legislative Yuan of Taiwan. Some legislators are seeking to increase the Tobacco Health and Welfare Tax by NT $3 per pack, raising it from NT $5 to NT $8 per pack. Just as they have done in the past, cigarette sellers will probably shift the tax increase to the consumers letting them be responsible for the increase. The effect of the price increase on demand depends on cigarette price elasticity – the larger the elasticity, the larger the reduction in consumption. Therefore, an estimation of price elasticity for domestic cigarettes could be a very important indicator of the possible effect a "Tobacco Health and Welfare Tax" would have on cigarette consumption and could be used to adjust Tobacco Health and Welfare Tax accordingly.
Price elasticity of cigarette in Taiwan has been mostly estimated using time-series data [12,13], though this method might overlook the impact of smuggled cigarettes on the price elasticity and overestimate the price elasticity. It might be more appropriate and useful to use cross-section data from the Health Interview Survey to estimate the effect of cigarette tax on cigarette consumption and to compare differences in cigarette price elasticity with various smoker characteristics.
Methods
This study uses data on current smokers from 17 to 69 years of age during years 2000 to 2003. First, the demand function of the current smokers was established. Respondents who answered "everyday" or "some of the days" to the question how often you smoke were classified as "current smokers". Then, we used a random sampling of how they consumed and how much they paid for it between 2000 to 2003 to calculate cigarette price elasticity. We then analyzed differences in cigarette price elasticity in smokers categorized according to gender, age, education, income standard, and how much they smoked.
Demand function
Cigarette demand function was estimated by OLS and expressed by double logarithm function. The estimated demand function we used was:
ln Qig = α0 + β1 ln Pig + γ2 ln Iig (1)
where
lnQig is i'th current smokers' logarithm representing monthly amount smokers consumed per person in group g. Current smokers were categorized according to age, gender, education, monthly income, and amount smoked. lnPig is i'th current smokers' logarithm representing cigarette price per pack (NT$) for smoking characteristics group g. lnIig is i'th current smokers' logarithm representing income per capita (NT$) for various categories of smokers in group g.
The α0, β1, and γ2 are parameters to be estimated. In order to measure how price change might affect cigarette consumption, the determination of price elasticity was particularly important. Price elasticity of demand for cigarettes is defined as the percent change in consumption resulting from a price increase. Cigarette price elasticity of demand β1 and income elasticity of demand γ2 can be derived from logarithmically differentiation (1) according to price and income.
Data collection
Using an annual face-to-face survey on cigarette consumption from 2000 to 2003 by Taiwan National Health Research Institutes, we collected data on how many packs current smokers consumed, how much they paid for a pack of cigarettes, how much they earned per month and how much they spent on cigarettes per month. Current smokers were categorized into gender, age, education, income, and amount smoked. Calculations of cigarette price were based on the average retail price of the top 3 most consumed cigarettes, calculations of number of packs smoked per month were done by dividing the monthly cigarette consumption by the average retail price. Calculations of income were based on personal monthly income.
Certain background characteristics are listed in Table 1. More than 90% of the smokers were men; less than 10% women. The numbers of young smokers were rising at the time of the study. Young people between the ages of 17 and 24 years old made up 5.4% of the sample in 2000, while they made up 12.8% in 2003, a 1.4 percent increase. The number of elderly smokers above the age of 55 gradually declined from 15.5% in 2000 to 10.8% in 2003. People with higher educational backgrounds tended to smoke more. More than 60% of the current smokers had senior or junior high school educations between 2000 and 2002. Second only to smokers with senior high school degrees, the percentage of the smokers with college degree increased by 27.2% in 2003, accounting for almost 40% of all the smokers. Thirty-five to forty percent earned between NT $20,000 to NT $30,000 per month. Those who smoked less than one pack were defined as light smokers; 1~2 packs (2 packs excluded), medium smokers; and 2 packs and above, heavy smokers. The proportion of light smokers had gradually increased from 50% in 2000 to 60% in 2003, and the proportion of heavy smokers gradually decreased from 11.8% in 2000 to 6.5% in 2003, showing an overall tendency toward reducing consumption.
Table 1 Background characteristics of the current smokers in Taiwan, 2000–2003
2000 2001 2002 2003
Characteristics No. % No. % No. % No. %
Total 856 632 521 493
Gender
Male 789 92.2 599 94.8 496 95.0 460 93.3
Female 67 7.8 33 5.2 25 5.0 33 6.7
Age
17–24 46 5.4 35 5.5 41 7.9 63 12.8
25–34 179 20.9 133 21.0 94 18.0 101 20.5
35–44 299 34.9 222 35.1 190 36.5 177 35.9
45–54 199 23.2 149 23.6 122 23.4 99 20.1
55- 133 15.5 93 14.7 74 14.2 53 10.8
Education
College and above 168 19.6 128 20.3 107 20.5 134 27.2
Senior high school 317 37.0 245 38.8 195 37.4 196 39.8
Junior high school 216 25.2 153 24.2 136 26.1 98 19.9
Primary school or lower 155 18.1 106 16.8 83 15.9 65 13.2
Month income
No income 98 11.4 93 14.7 63 12.1 75 15.2
<NT $20,000 146 17.1 94 14.9 94 18.0 84 17.0
NT $ 20,000–39,999 335 39.1 237 37.5 192 36.9 174 35.3
NT $ 40,000–59,999 181 21.1 144 22.8 117 22.5 105 21.3
≥ NT $ 60,000 96 11.2 64 10.1 55 10.6 55 11.2
Smoking degree
Light smokers 445 50.0 319 50.5 281 53.9 296 60.0
Medium smokers 310 36.2 250 39.6 200 38.4 165 33.5
Heavy smokers 101 11.8 63 10.0 40 7.7 32 6.5
Results
Cigarette consumption, retail price and personal monthly income were used in equation (1), the OLS method, to calculate cigarette price and income elasticities. The overall cigarette price elasticity was negative, less than one, indicating that cigarette consumption or demand in Taiwan was inelastic during the study period (Table 2). Taken into consideration that cigarette price elasticity is inelastic and that reduction of the cigarette consumption is done with a strategy of raising domestic cigarette price, we speculate that cigarette prices need to be even higher, to lower consumption enough to have a clear, strong impact in improved public health outcomes. While income elasticities were not statistically different from 2000 to 2002, they did reach a statistically significant level (5%) in 2003. At that time, estimated income elasticity was positive, indicating that cigarettes were normal goods. A positive value normally means that demand for normal goods would increase as incomes rise. But the value could vary greatly among normal goods. Interestingly, in our study, we found that as incomes were rose, income elasticity became low, indicating that income increases in 2003 only had a slight effect on cigarette consumption.
Table 2 The estimated overall cigarette price and income elasticities of the current smokers, 2000–2003a
2000 2001 2002 2003
Price elasticity -0.3134 (-2.894)* -0.3684 (-3.482)* -0.5274 (-3.143)* -0.3090 (-1.531)
Income elasticity 0.0069 (0.708) 0.0042 (0.504) -0.0012 (-0.125) 0.0320 (2.916)*
a. t ratios are shown in parentheses.
* p < 0.05.
After Taiwan's accession to WTO in 2002, Taiwan's first Tobacco Health and Welfare Tax added NT $5 the price of cigarettes, resulting in an increase in cigarette retail price from NT $35.2 in 2001 to NT $42.2 in 2002, about an NT $7 increase. In 2001, smokers 15 years old or older consumed an average 126.52 packs/person [14]. In 2002, the cigarette price elasticity became -0.5274, meaning that this price increase caused a reduction of cigarette consumption by 13.27 packs/person (10.5% per person) and a reduction of about 0.235 billion packs in the total consumption. Meanwhile, new cigarette consumption in 2002 was reduced to 2 billion packs for the population at and above the age of 15 in 2001. Provided there was a NT $16.8 tax on every pack of cigarettes, the tax revenue for the government would be 33.6 billion NT dollar.
However, the tax has been in force for two years, and a significant reduction in consumption has been shown to date. Cigarette taxes accounted for 40% of the retail price in 2002. Although forty percent sounds high, this proportion is still rather low when, as mentioned earlier, comparing it with the 66% or more of the retail price going for cigarettes prices in high-income countries (with the notable exception of the United States)[9]. Consequently, cigarette prices in Taiwan are well below those of many high-income countries, who have seen significant reductions in cigarette consumption. Those successes certainly impose a pressure on the government to implement another tax increase in the existing Health and Welfare Tax.
Inter-party negotiations at the Legislative Yuan resulted in changes to Tobacco and Liquor Tax Law which led to a NT $3 rise in the health tax levied on cigarettes, from NT$5 to NT$8 per pack. The cigarette industry is likely to pass the tax increase in the form of higher prices on to the consumer. A price increase of 3 NT per pack in cigarette would reflect a consumption reduction of 2.47 packs per capita, totaling 44.7 million packs, 2.2% per capita, in cigarette consumption. Cigarette consumption would be reduced to 2 billion packs (based on a population count of 2003), and the government tax revenue would be 33.46 billion, including the Health and Welfare Tax of 16 billion. We estimate that with a Health and Welfare Tax increase from 5 NT to 8 NT per pack, there would be a 6 billion NT increase in revenue from the Health and Welfare Tax.
In 2000 and 2001, the cigarette price elasticities of the current smokers were -0.3134 and -0.3684, respectively. In 2002, it had reached its maximum of -0.5274, indicating that as cigarette prices increased, so did the price elasticity. Consumers responded to the higher prices by cutting consumption. Then in 2003, after Taiwan's accession to WTO, cigarette price elasticity then lowered to -0.309, indicating that the effect of a cigarette price hike can diminish over time.
Two sets of previous data concerning the cigarette consumption, retail price and personal monthly income of the current smokers obtained during 2000–2001 and during 2002–2003, were made available to us. Using the OLS method and the rise of cigarette price in 2002 as the baseline, we computed the cigarette price and income elasticities for these two sets of data. Table 3 shows the differences in the overall cigarette price elasticity of the current smokers between before and after Taiwan's accession to WTO in 2002: before -0.4062 and after -0.3352. Our result clearly indicated that smokers were more responsive to the price after the cigarette price rose in 2002–2003. Yet, -0.4062 and -0.3352 also revealed that there was no big change in cigarette price elasticity before and after Taiwan's accession to WTO. The reason for this might be that the rise of average cigarette price was only NT $7 per pack. We speculate that a possible continuous and substantial rise in cigarette prices in the future might increase the overall price elasticity, which in turn could allow for a more effective use of the tax increases or price increases to control tobacco.
Table 3 The estimated cigarette price and income elasticities of different types of current smokers during 2000–2001 and during 2002–2003a
2000–2001 2002–2003
Characteristics Price elasticity Income elasticity Price elasticity Income elasticity
Overall -0.3352 (-4.355)* 0.0055 (0.849) -0.4062 (-3.102)* 0.0174 (2.331)*
Gender
Male -0.3107 (-4.036)* 0.0020 (0.291) -0.3930 (-2.935)* 0.0194 (2.450)*
Female -0.1223 (-0.312) -0.0206 (-0.820) -0.1406 (-0.251) -0.0476 (-1.903)
Age
17–24 -0.4022 (-0.645) 0.0368 (1.151) -0.1057 (-0.144) 0.0420 (1.532)
25–34 -0.0528 (-0.274) -0.0038 (-0.227) 0.2300 (0.734) 0.0315 (1.822)
35–44 -0.1835 (-1.425) -0.0028 (-0.206) -0.2154 (-1.023) -0.0060 (-0.422)
45–54 -0.1540 (-1.087) 0.0048 (0.388) -0.4100 (-1.624) -0.0107 (-0.724)
55- -0.2453 (-1.073) -0.0080 (-0.608) -0.0475 (-0.118) 0.0008 (0.044)
Education
College and above -0.1070 (-0.514) -0.0133 (-0.584) -0.7007 (-2.047)* 0.0919 (4.148)*
Senior high school -0.2772 (-2.273)* 0.0211 (1.825) -0.5369 (-2.703)* 0.0233 (2.069)*
Junior high school -0.5766 (-4.093)* 0.0275 (2.552)* 0.1789 (0.833) 0.0279 (2.058)*
Preliminary or lower 0.0560 (0.302) -0.0056 (-0.440) -0.0392 (-0.118) -0.0191 (-1.244)
Month income
No income -0.5103 (-2.193)* -0.8363 (-2.014)*
<NT $20,000 -0.4570 (-2.373)* -0.7478 (-2.316)*
NT $ 20,000–39,999 -0.4805 (-3.953)* -0.2861 (-1.345)
NT $ 40,000–59,999 -0.1562 (-0.951) -0.2624 (-1.022)
≥ NT $ 60,000 0.2341 (1.050) -0.1152 (-0.301)
Smoking degree
Light smokers -0.1984 (-2.046)* 0.0087 (1.120) -0.5320 (-3.293)* 0.0172 (1.956)
Medium smokers -0.0228 (-0.914) -0.0005 (-0.198) -0.2600 (-5.068)* -0.0046 (-1.514)
Heavy smokers 0.2573 (1.973)* -0.0083 (-0.820) -0.0006 (-0.005) 0.0010 (0.137)
a. t ratios are shown in parentheses.
* p < 0.05.
Cigarette price elasticity for the male smokers reached statistically significance (5%), which was higher than the price elasticity of the female smokers, indicating that men were more responsive to price elasticity than the women. The cigarette price elasticity of the male smokers in 2002–2003 was -0.393, which was higher than -0.3107 male smokers in 2000–2001. The income elasticity of male smokers in 2002–2003 was 0.0194, reaching statistical significance (5%). There was no statistical difference found in estimated cigarette price and income elasticity of smokers among the different age sub-groups, though, according to some reports in foreign countries, teenagers are more sensitive changes to cigarette prices than adults [15-17].
Education level seemed to make a difference. The price elasticity smokers with junior high school educations was -0.5766 in 2000–2001, which was higher than that of senior high school level smokers, -0.2772. In 2002-2003, the price elasticity of college level smokers was -0.7007, which was higher than that of the smokers of senior high school level, -0.5369. Therefore, between 2000 and 2001, the more educated group had a larger price elasticity, though the less educated group had the higher coefficient in 2002–2003. Our findings cover two time spans and are different from those reported by foreign researchers who have found that consumers at lower education levels respond stronger to the change of cigarette price [18]. In our study, those with no income had the greatest cigarette price elasticity, -0.5103 in 2000–2001 and -0.8363 in 2002–2003. Those with no income were more sensitive to the price rise than those with income. Light smokers had the highest cigarette price elasticity, -0.1984 in 2000–2001 and -0.532 in 2002–2003. Heavy smokers had a price elasticity of 0.2573 in 2000–2001, indicating that with the rise of cigarette price, these smokers increased their consumption of tobacco.
Conclusions
In this study, we evaluate the effect of a new tax on the consumption of tobacco by calculating the cigarette price elasticities of various kinds of smokers and comparing those values with their reactions to an increase cigarette, and then using knowledge gained from that study, we assess the possible effect of another increase on consumption. We found that the new tax scheme implemented after Taiwan joined the WTO reduced cigarette consumption by 13.27 packs/person (10.5%). We estimated that an additional NT $3 increase in Taiwan's Health and Welfare Tax would reduce cigarette consumption here by 2.47 packs/person (2.2%).
In this study, cigarette price elasticity was less than one, meaning that in addition to reducing cigarette consumption, an additional tax would also generate additional tax revenues. Continuing price increases should reduce cigarette consumption significantly. Provided that the tax increases are proportionately larger than the resulting reduction in cigarette consumption, cigarette tax revenues will increase and can be used to reduce current National Health Insurance deficits and possibly reduce the damage and death caused by smoking related diseases.
Based on estimated cigarette price elasticities for various kinds of current smokers, we found that male smokers without income and light smokers were more sensitive to changes in cigarette prices. Teenagers (17 – 24 years old), however, were not found to be significantly influenced by the change in cigarette price, which means it will take more than just tax increases to decrease consumption among our youth. Schools will need to commit to preventive education by inculcating the students with the knowledge of tobacco hazards. Only through early preventive education starting from their childhood can we expect to see significant reduction in cigarette consumption.
Finally, the R-squared statistics of the each empirical estimation was below 0.1 for probably sake of to calculate elasticity and exclude variables like advertisement. Future research should also attempt to include these factors with cigarette demand function.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JML performed physical measurements, collected data, and drafted the manuscript. TCH reviewed the manuscript. CYY and SHC carried out the statistical analysis and participated in data collection. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
A partly-article of this research was support by extramural grants from Bureau of Health Promotion, Department of Health in Taiwan (Grant Number: BHP-92- Anti-Tobacco-2H01). This study is based (in part) on data from the Taiwan Interview Survey on Cigarette Consumption (TISCC) Original Database provided by National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of Department of Health and National Health Research Institutes.
==== Refs
Guindon GE Tobin S Yach D Trends and affordability of cigarette price: ample room for tax increases and related health gains Tobacco Control 2002 11 35 43 11891366 10.1136/tc.11.1.35
Tsai YW Chen CS Liu TC Yang CL Chuang CC Chen JI Wen Chi-Pang Analysis of cigarette taxation in Taiwan In Collection of Research Papers on Tobacco or Health in Taiwan 2002 2002 Taiwan: Division of Health Policy Research, National Health Research Institutes 188 211
Wen CP Levy DT Wen Chi-Pang Smoking policy at the crossroads: opportunities and challenges In Collection of Research Papers on Tobacco or Health in Taiwan 2002 2002 Taiwan: Division of Health Policy Research, National Health Research Institutes 4 49
Lin HS Wen CP Hou CC Hou CC Chu ST Chung WS Wen Chi-Pang Research on tobacco litigation in Taiwan In Collection of Research Papers on Tobacco or Health in Taiwan 2002 2002 Taiwan: Division of Health Policy Research, National Health Research Institutes 68 90
Sung HY Hu TW Keeler TE Cigarette taxation and demand: an empirical model Contemporary Economic Policy 1994 12 91 100
Hu TW Sung HY Keeler TE Reducing cigarette consumption in California: tobacco taxes vs. anti-smoking media campaign American Journal of Public Health 1998 88 1218 1222
Lanoie P Leclair P Taxation or regulation: looking for a good anti-smoking policy Economics letters 1998 58 85 89 10.1016/S0165-1765(97)00258-9
Hu TW Mao Z Effects of cigarette tax on cigarette consumption and the Chinese economy Tobacco Control 2002 11 105 108 12035000 10.1136/tc.11.2.105
Chaloupka FJ Hu TW Warner KE Jacobs Rowena Yurekli Ayda Jha Prabhat, Chaloupka Frank The taxation of tobacco products Tobacco control in developing countries 2000 London: Oxford University Press 237 272
Hopkins DP Briss PA Richard CJ Reviews of evidence regarding interventions to reduce use and exposure to environmental tobacco smoke American Journal of Preventive Medicine 2001 20 16 66 11173215 10.1016/S0749-3797(00)00297-X
United States Department of Health and Human Services Reducing tobacco use: a report of the Surgeon General. 2000 Atlanta, Ga, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Heath Promotion, Office on Smoking and Heath
Hsieh CR Hu TW Lin C-FJ The demand for cigarettes in Taiwan: domestic versus imported cigarettes Contemporary Economic Policy 1999 17 223 234
Shi MS Hsieh CR Cigarette demand, cigarette taxation and anti-smoking: review of empirical literature and the case of Taiwan Journal of Social Sciences and Philosophy 1999 11 301 334
Lee JM Yen CJ A Study of the Impacts of Trade Liberalization and Tobacco Hazard Control on Cigarettes Demand and Industry Structure in Taiwan 2004 Taiwan: Bureau of Health Promotion, Department of Health
Chaloupka FJ Wechsler H Price, tobacco control policies and smoking among young adults Journal of Health Economics 1997 16 359 373 10169306 10.1016/S0167-6296(96)00530-9
Lewit EM Coate D The potential for using excise taxes to reduce smoking Journal of Health Economics 1982 1 121 45 10263952 10.1016/0167-6296(82)90011-X
Lewit E Coate MD Grossman M The effects of government regulations on teenage smoking Health Economics 1981 24 545 569
Chaloupka FJ Rational addictive behavior and cigarette smoking Journal of Political Economy 1991 99 722 742 10.1086/261776
| 15598345 | PMC539350 | CC BY | 2021-01-04 16:28:47 | no | BMC Public Health. 2004 Dec 14; 4:61 | utf-8 | BMC Public Health | 2,004 | 10.1186/1471-2458-4-61 | oa_comm |
==== Front
BMC Struct BiolBMC Structural Biology1472-6807BioMed Central London 1472-6807-4-101557921010.1186/1472-6807-4-10Research ArticleCrystal structure of subunit VPS25 of the endosomal trafficking complex ESCRT-II Wernimont Amy K [email protected] Winfried [email protected] European Molecular Biology Laboratory (EMBL), 6 rue Jules Horowitz, 38042 Grenoble, France2004 4 12 2004 4 10 10 15 9 2004 4 12 2004 Copyright © 2004 Wernimont and Weissenhorn; licensee BioMed Central Ltd.2004Wernimont and Weissenhorn; 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
Down-regulation of plasma membrane receptors via the endocytic pathway involves their monoubiquitylation, transport to endosomal membranes and eventual sorting into multi vesicular bodies (MVB) destined for lysosomal degradation. Successive assemblies of Endosomal Sorting Complexes Required for Transport (ESCRT-I, -II and III) largely mediate sorting of plasma membrane receptors at endosomal membranes, the formation of multivesicular bodies and their release into the endosomal lumen. In addition, the human ESCRT-II has been shown to form a complex with RNA polymerase II elongation factor ELL in order to exert transcriptional control activity.
Results
Here we report the crystal structure of Vps25 at 3.1 Å resolution. Vps25 crystallizes in a dimeric form and each monomer is composed of two winged helix domains arranged in tandem. Structural comparisons detect no conformational changes between unliganded Vps25 and Vps25 within the ESCRT-II complex composed of two Vps25 copies and one copy each of Vps22 and Vps36 [1,2].
Conclusions
Our structural analyses present a framework for studying Vps25 interactions with ESCRT-I and ESCRT-III partners. Winged helix domain containing proteins have been implicated in nucleic acid binding and it remains to be determined whether Vps25 has a similar activity which might play a role in the proposed transcriptional control exerted by Vps25 and/or the whole ESCRT-II complex.
==== Body
Background
Endosomal compartments receive membrane bound cargo from both the biosynthetic and the endocytic pathways. Receptor downregulation by endocytosis includes transport to early endosomes and either recycling or sorting into late endosomes. The latter have the morphological characteristics of multivesicular bodies (MVB) [3] that can undergo homotypic fusion or heterotypic fusion with lysosomes, which deliver MVB cargo for proteolytic degradation [4]. In addition to receptor downregulation, MVB formation has been implicated in antigen presentation [5] and in the release of enveloped viruses [6,7].
Gene deletion and inactivation studies in yeast have identified 17 proteins that directly affect MVB formation (yeast class E compartment) by resulting in aberrant endosomal/vacuolar morphology [4]. All proteins are required for vacuolar protein sorting (VPS) into the class E compartment and are recruited to endosomal membranes from the cytosol in order to assemble into three ESCRT (Endosomal Sorting Complexes Requited for Transport) complexes that function in MVB formation [8-11]. Receptor mono-ubiquitinylation has been shown to serve as a signal to enter the MVB pathway [12]. Initial recognition of ubiquitinated cargo by Vps27 recruits the ubiquitin binding protein Vps23 [11,13], which in turn leads to the assembly of the multi-protein complex ESCRT-I (VPS23, VPS28, and VPS37) [10]. ESCRT-I subsequently recruits ESCRT-II, composed of Vps22, Vps25, and Vps36, which in turn activates ESCRT-III subcomplexes [8,9]. Assembly of ESCRT-III at the endosome initiates the sorting and concentration of ubiquitinated cargo; ubiquitin is removed and Vps4, an AAA-type ATPase, dissociates the ESCRT complexes concomitantly with membrane invagination and budding of vesicles into the lumen of the endosome [4].
Two recent crystal structures of a core of the ESCRT-II complex reveal a trilobal complex, containing two copies of Vps25, one copy of Vps22 and the C-terminal region of Vps36. Each subunit is composed of two winged helix domains and an N-terminal region of Vps25 interacts with Vps22 and Vps36 [1,2].
Although ESCRT-II is essential for the MVB pathway, since cells missing ESCRT-II components fail to localize ESCRT-III to late endosomes [8,9] the complex has also been found "moonlighting" in the nucleus. The human and rat homologues of ESCRT-II were originally identified as the EAP complex (ELL Associating Protein; Vps22/EAP30; Vps25/EAP20; Vps36/EAP45), associated with the RNA polymerase II elongation factor ELL in the nucleus [14,15]. Consistent with a role in transcriptional control, yeast Vps22 (or SNF8) as well as Vps25 and Vps36 have been implicated in glucose-dependent gene expression control [15,16]. To date, it is not clear whether the role of ESCRT-II in MVB formation is independent of its function as a transcriptional activator or whether both processes are linked. Here, we report the crystal structure of full-length yeast Vps25, composed of two homologous winged-helix domains.
Results and discussion
Structure of Vps25
The structure of Vps25 was solved by single wavelength anomalous diffraction (SAD) using selenomethionine-derivatized crystals. Vps25 consists of two homologous winged helix domains as detected by the program GRATH that are arranged in tandem (Figure 1A). Winged helix folds are compact alpha/beta structures with secondary structure elements arranged in a typical order (H1-S1-H2-H3-S2-W1-S3-W2optional) [17], which fold into a mostly helical part followed by a twisted anti-parallel beta-sheet and two large loops (wings, W). The fold of Vps25 deviates slightly from the canonical fold. The N-terminal domain 1 (residues 1 to 126) contains two additional N-terminal 3/10 helices, implicated in the interaction with either Vps22 or Vps36 [1,2], followed by the canonical helix 1 and strand 1. It lacks canonical helix 2, which instead folds into a large disordered loop followed by strands 3 and 4 that connects to helix 2 (at the corresponding position of canonical helix 3). Strands 5 and 6 then form, together with strand 1, a twisted anti-parallel beta-sheet with wing W1 protruding from the structure (Figure 1A and Figure 2). Domain 1 also lacks wing W2, as in the cases of winged helix domain containing transcription factors E2F4 and DP2 [18]. Strand 6 flows directly into domain 2, which also has a canonical winged helix fold except for the absence of wing W2 (Figure 1A and Figure 2). Domains 1 and 2 are tightly packed against each other and their C alpha atoms can be superimposed with an r.m.s. deviation of 3.4 Å (Figure 1B), confirming their structural relatedness. The domain interface is dominated by van der Waals contacts including conserved and non conserved residues Trp44, Phe122, Leu104, Leu124, Trp125 in domain 1 and Leu128, Trp131, Met168, Pro169 and Leu172 in domain 2 (Figure 2).
Figure 1 Vps25 contains two winged helix domains arranged in tandem. (A) Ribbon diagram of Vps25; the two domains are shown in orange and yellow. Secondary structure elements are labeled. The major missing loop region connecting strands 1 and 3 is indicated by a dashed line. (B) Superposition of the Calpha positions of the N- and C-terminal domains (residues 23 to 48 and 85 to 101 with corresponding C-terminal domain residues; r.m.s.d. 3.4 Å). Note that the positions of helices 1/3 and helices 2/5 as well as wing positions W1 match up well.
Figure 2 Structure based sequence alignment of Vps25. Sequences aligned using S. cerevisiae Vps25 (gene bank #CAA89632) and Vps25 orthologues from H. sapiens (#BE386260), D. melanogaster (#AAF59066) and from C. elegans (#T26073). Identical residues are shown on red background, similar residues are drawn in red and sequence similarity is underlined by blue boxes. Secondary structure elements are shown. Disordered regions in the Vps25 structure are indicated by dashed lines.
Structural comparision of unliganded Vps25 and Vps25 in complex with Vps22 and Vps36 (ESCRT-II)
Two recent crystal structures of the ESCRT-II core reveal trilobal structures with head to tail interactions of one copy of Vps25 with Vps22 and the other copy of Vps25 with Vps36 at the center. In both cases a conserved proline rich N-terminal region of Vps25 (Figure 2) together with conserved Arg83 mediate key interactions [1,2]. Therefore it was of interest to analyse whether Vps25 undergoes any conformational changes upon participation in ESCRT-II complex formation. Superposition of the C alpha atoms with one copy of Vps25 from either ESCRT-II complex structure ([1,2]; pdb entries 1U5T and 1W7P) revealed r. m. s. displacements of 1.2/1.2 Å (residues 3 to 51), 1.5/1.7 Å (residues 74 to 155) and 2.3/2.9 Å (residues 159–199) respectively. The major changes are confined to both wings W1 and W2 indicating their conformational flexibility (Figure 3). In contrast, the conserved N-terminal segment, which is implicated in Vps22 and Vps36 interactions shows no substantial changes (Figure 3).
Figure 3 Comparison of unliganded and liganded Vps25. Superposition of unliganded Vps25 (red) with Vps25 from both ESCRT-II structures [1, 2] (blue, pdb code 1U5T chain C; green, pdb code 1W7P chain B). The peptide backbones are shown as coils. Vps25 is shown in the same orientation as in figure 1A. The position of Arg83 is indicated by an arrow.
In the unliganded Vps25 structure, this helical segment constitutes the 1192 Å2 dimerization interface of two identical Vps25 dimers present in the asymmetric crystal unit. The dimer contact is mediated by hydrophobic residues Pro5, Pro6, Val7, Phe10, Pro11, and Pro12, which is similar to the contact region described for Vps25 interactions with Vps22 and Vps36 [1,2]. In the Vps25 structure Arg83 does not participate in dimerization but hydrogen bonds to Thr15 instead of forming salt bridges with either Vps36 Asp548 or Vps22 Asp214 as observed in the ESCRT-II complex [1,2]. Arg83 locates to a beta hairpin (strand 4; Figure 2) in the unliganded form of Vps25. Although the position of Arg83 is unchanged in all Vps25 structures (Figure 3) the position of the preceding loop region varies which might be due to differences in secondary structure assignment [1,2]. Therefore Vps25 seems to dock as a rigid body onto either Vps22 or Vps36 upon ESCRT-II complex formation. Although we do not detect Vps25 dimer formation in vitro, a dimeric form of Vps25 might be stabilized through other unknown interactions.
Structural homology of Vps25 with nucleic acid binding winged helix domains
Analysis of the full-length structure with DALI [19] revealed seven structural homologues displaying nucleic acid binding winged helix domains with a Z score above 5 for Vps25 domain 1. The top two hits were the selenocysteine-specific elongation factor fragment (PDB 1lva, Z score 6) and double-stranded RNA specific adenosine deaminase (ADAR) Z-alpha domain (PDB 1qbj, Z score 5.5). Winged helix family members interact with nucleic acids mostly via the "specificity helix" that binds to the major groove of the DNA with two flanking loops contributing to DNA interactions [17]. Superposition of Vps25 domain 1 onto the winged helix domain of E2F-4 bound to DNA [18] matching the "specificity helices" (Vps25 helix H2) revealed a potential fit with only minor clashes at the helix H1 loop region (data not shown). A potential nucleic acid interaction of Vps25 might be interesting in light of the described role of Vps25 and the other ESCRT-II subunits in glucose-dependent gene regulation [15,16] and complex formation with RNA polymerase II elongation factor ELL [14,15], although no biochemical data exist so far to support such a proposed function.
Vps25 participates in protein complex formation
The ESCRT-II complex assembles at the endosomal membrane downstream of ESCRT-I and recruits ESCRT-III subcomplexes [8-10]. Consistent with such a sequential assembly, further ESCRT-II interactions of Vps25 have been described, namely with Vps28 (ESCRT-I) and with Vps20 (CHMP6; ESCRT-III) [7,20]. Surface electrostatic potential maps of Vps25 reveal a negatively charged surface within domain 2 that is characterized by a patch of conserved residues such as Glu153, Glu170 and Tyr152 (Figure 4A and Figure 2). Tyr152 is also part of the highly conserved domain 2, helix 4 (Figure 2). Domain 2 is the outer domain of Vps25 in the ESCRT-II complex and this region would thus be freely accessible for potential interaction(s) with Vps28 or Vps20. Similarly, basic residues (Lys99 and Arg23) potentially implicated in nucleic acid recognition are part of a conserved patch on domain 1 (Figures 4B and 2).
Figure 4 Surface charge distribution of Vps25. (A) Surface potential representation of Vps25 with regions where electrostatic potential <-10 kBT are red, while those >+10 kBT are blue (kB, Boltzmann constant; T, absolute temperature). (B) Horizontal rotation (180°). Exposed residues are labeled for orientation. Note that one face of the molecule carries a mainly negative charge (A) while the other one carries a mainly positive charge (B).
Vps25 contains additional features, which are unique to S. cerevisiae, as evidenced from multiple sequence analysis [15,16]. Vps25 orthologues have a shorter strand 2 to strand 3 connection (19 residues), whose sequence is composed of mostly charged residues and is disordered in our structure as well as in the ESCRT-II structures [1,2]. Furthermore, domain 1 wing W1 is shorter (7 residues) (Figure 2), which might indicate S. cerevisiae unique protein-protein interaction sites.
Conclusions
Clear evidence suggests that ESCRT-II recruitment is involved in MVB formation leading to plasma membrane receptor downregulation [4]. On the other hand ESCRT-II seems to play a role in transcription regulation [15]. Similarly, other ESCRT components such as Tsg101 (Tumor susceptibility gene; Vps23; ESCRT-I) and members of the CHMP protein family (ESCRT-III; Chromatin Modifying Protein; Charged Multivesicular body Protein) are also found to act in the nucleus as well as in the cytosol and at endosomal membranes [21-23]. Interestingly, both Vps25 and Vps36 have been implicated in regulating stress and pheromone response pathways [24] and pheromone receptor Ste2 is downregulated via the endosomal pathway [12]. Similarly, SNF8 (Vps 22; EAP30), Vps36 and Vps25 are all directly involved in derepression of glucose-repressed genes, which might be linked to sorting of sucrose receptors via the endosomal pathway [15,25]. Protein sorting into MVB involves monoubiquitylation of cargo, which is recognized by ESCRT members. ESCRT-II Vps36 contains an ubiquitin binding NZF zinc finger motif that is necessary for protein sorting [26]. Therefore, ESCRT-II complexes may sense the turnover of specific ubiquitylated receptors at the endosomal membrane together with other unknown signals. As ESCRT-II only transiently associates with endosomal membranes [9] a signal within the MVB process might induce nuclear localization of ESCRT-II, where it could stimulate gene expression leading to up or down regulation of specific membrane receptors.
Methods
Protein expression, purification and crystallization
Full length yeast Vps25 DNA (gene bank #CAA89632) was cloned into expression vector pETM30 (EMBL, Protein Expression Facility) and the Vps25 GST fusion protein was expressed in E. coli BL21 codon+ cells. For purification, cell pellets from 6 liter cultures were lysed in 150 mls of buffer A (50 mM Tris-HCl, pH 8.5, 200 mM NaCl, 0.2 mM DNaseI, 2 mM β-ME, 2 complete EDTA-free protease inhibitor tablets (Pierce)) and 0.1 mg/ml lysozyme for one hour on ice. The cell lysate was cleared by centrifugation and loaded onto a GST-sepharose (Pharmacia) column. The column was extensively washed with buffer B (50 mM Tris pH 8.5, 200 mM NaCl) and Vps25 fusion protein was eluted with buffer B containing 5 mM reduced glutathione. GST was then removed by TEV cleavage (w/w; 1:200) at 4°C overnight. His-tagged GST and TEV were subsequently both removed on a Ni2+ chelating sepharose column. Vps25 was further purified on a superdex75 column (Pharmacia) in buffer C (50 mM Tris 8.5, 200 mM NaCl, 2 mM βME). Selenomethione-labeled Vps25 was produced using standard procedures and purified as described above.
Crystallization conditions for Vps25 (7 mg/ml) were first determined by screening 600 conditions using a Cartesian crystallization robot. Initial conditions were refined using the hanging drop method, and the final crystallization condition (100 mM Na cacodylate pH 6.5, 200 mM Mg or Ca acetate, 5–7% glycerol, and 15–18% polyethylene glycol 8000) produced rectangular- and wedge-shaped selenomethionine-labeled Vps25 crystals in the same drop. Native Vps25 crystallized initially only with rectangular morphology and wedge-shaped crystals were produced by microseeding with the original SeMet crystals. For cryogenic data collection, the crystals were equilibrated in 25% glycerol and flash cooled in a gaseous nitrogen stream at 100 K.
Crystallization produced rectangular crystals that belong to space group P422 with unit cell dimensions a = b = 78 Å, c = 54 Å and diffract to 3.2 Å resolution. However, all data sets collected from these crystals proved to be almost perfectly merohedrally twinned. The second crystal form, wedge-shaped, belonged to space group P212121 with unit cell dimensions as indicated (table 1), contained 4 molecules per asymmetric unit, diffracted X-rays to 3.1 Å resolution and was used for structure solution.
Table 1 Data Collection and Refinement.
Crystal VPS25-SEMET VPS25-NATIVE
Space Group P212121 P212121
Wavelength 0.97914 0.931
Unit Cell (Å)
a 53.44 53.36
b 124.11 123.66
c 139.48 140.30
Resolution (outer shell) (Å) 100-3.20 (3.31-3.20) 100 - 3.10 (3.21-3.10)
Total Reflections (outer shell) 111407 (10016) 65477 (6064)
Unique Reflections 29230 16330
Completeness (%) (outer shell) 98.9 (93.1) 93.1 (91.3)
Rmerge (outer shell) 0.090 (0.335) 0.053 (0.307)
Average I/sigma (outer shell) 12.3 (4.9) 20.9 (4.9)
Phasing
Number of Se Sites 14
SOLVE FOM 0.351
RESOLVE FOM (ncs) 0.694
Refinement
Resolution (outer shell) (Å) 25.0-3.10 (3.29-3.10)
Number of reflections (test set) 16404 (790)
R factor 0.275
Free R factor 0.327
Number of protein/solvent atoms 5760/16
Average B factor (Å2) 51.3
Rms deviation bond lengths (Å) 0.009
Ramachandron Mol A Mol B MolC MolD
Most favored
Additionally favored 83.8
15.7 79.5
20.5 82.1
17.9 67.2
32.8
Data Collection
Native data for Vps25 were collected at the European Synchrotron Radiation Facility (ESRF) beamline ID14-EH3 and data from SeMet-labeled crystals were collected at the ESRF beam line ID29 at three wavelengths (table 1). Data were processed and scaled with XDS [27].
Phasing and refinement
Significant radiation damage had occurred for data collected at the inflection and remote wavelengths, therefore only data collected at the peak wavelength (table 1) were used for SAD phasing. ShelXD [28] was used to find 14 out of 16 selenium sites, which were further refined with SOLVE [29]. Four-fold non-crystallographic symmetry was imposed on the sites in addition to solvent flattening with RESOLVE [30]. Phasing statistics are listed in table 1. The initial model was built with O [31] guided by the SeMet positions and clear tryptophan (7 per mol) and tyrosine (8 per mol) densities followed by refinement with CNS [32]. Strict four-fold NCS and phases were initially kept throughout the initial chain-tracing and refinement. During model building it was observed that molecules A and B and molecules C and D are arranged in the same dimer configuration and strict NCS was changed to restrained NCS during refinement. The packing also indicated tight interactions between molecules A, B, and C while molecule D showed only very few crystal contacts yet formed the "bridge" between two-dimensional layers formed by molecules A, B and C. The electron density maps for molecules A, B, and C were clear and well defined, while electron density for molecule D was poorly defined for side chains and loops. The model was improved by alternating cycles of model building and conjugate gradient minimization and restrained individual B-factor refinement using CNS [32]. The final coordinates were refined against the native dataset (30 to 3.1 Å) using the MLHL maximum likelihood target with the RESOLVE phases as constraint and retaining the original test set reflections. In the final stage of refinement, a maximum likelihood target and model phases alone were used.
The final model lacks two to five flexible loops (molecule mol A, residues 56–72, 114–115, 156–157; mol B, residues 53–73, 157–158; mol C, residues 57–72, 155–158; mol D, residues 19–21, 55–73, 107–120, 156–160, 185–186). Accordingly, mol D is poorly defined (43 residues missing out of 204). The final R factor and R free (0.275/0.327) reflect missing residues and the poor model for molecule D. The model exhibits otherwise overall good stereochemistry with no outliers in the Ramachandran plot as defined in PROCHECK (table 1) [33]. The coordinates have been deposited in the RCSB Protein Data Bank accession code 1XB4 [PDB:1XB4].
Structure analysis
Figures were generated using coordinates of molecule C with programs MOLSCRIPT [34], Raster 3D [35], ESPript [36], GRASP [37] and PyMOL . Sequences were aligned using Clustalx [38]. Secondary structure elements were assigned using the program DSSP [39]. The buried surface was calculated with CNS [32] and the program LSQMAN was used for superpositioning of C-alpha positions [40].
Authors' contributions
W.W. conceived of the study, and participated in its design, coordination and writing of the manuscript. W.W. expressed, purified and established initial crystallization conditions and participated in data collection. A.K.W. carried out data collection, structure solution and refinement and participated in writing of the manuscript. All authors read and approved the final manuscript.
Acknowledgments
We thank Drs. T. Muziol, C. Petosa, R. Ravelli and members of the ESRF/EMBL JSBG for support at the ESRF beamlines and Dr. J. Marquez and his team for high throughput crystallization analysis. This work was supported by EMBL and the Deutsche Forschungsgemeinschaft SFB grant 597 (W.W.).
==== Refs
Hierro A Sun J Rusnak AS Kim J Prag G Emr SD Hurley JH Structure of the ESCRT-II endosomal trafficking complex Nature 2004 431 221 225 15329733 10.1038/nature02914
Teo H Perisic O Gonzalez B Williams RL ESCRT-II, an Endosome-Associated Complex Required for Protein Sorting; Crystal Structure and Interactions with ESCRT-III and Membranes Dev Cell 2004 7 559 569 15469844 10.1016/j.devcel.2004.09.003
Palade GE A small particulate component of the cytoplasm J Biophys Biochem Cytol 1955 1 59 68 14381428
Katzmann DJ Odorizzi G Emr SD Receptor downregulation and multivesicular-body sorting Nat Rev Mol Cell Biol 2002 3 893 905 12461556 10.1038/nrm973
Kleijmeer M Ramm G Schuurhuis D Griffith J Rescigno M Ricciardi-Castagnoli P Rudensky AY Ossendorp F Melief CJ Stoorvogel W Geuze HJ Reorganization of multivesicular bodies regulates MHC class II antigen presentation by dendritic cells J Cell Biol 2001 155 53 63 11581285 10.1083/jcb.200103071
Strack B Calistri A Craig S Popova E Gottlinger HG AIP1/ALIX is a binding partner for HIV-1 p6 and EIAV p9 functioning in virus budding Cell 2003 114 689 699 14505569 10.1016/S0092-8674(03)00653-6
von Schwedler UK Stuchell M Muller B Ward DM Chung HY Morita E Wang HE Davis T He GP Cimbora DM Scott A Krausslich HG Kaplan J Morham SG Sundquist WI The protein network of HIV budding Cell 2003 114 701 713 14505570 10.1016/S0092-8674(03)00714-1
Babst M Katzmann DJ Estepa-Sabal EJ Meerloo T Emr SD ESCRT-III: an endosome-associated heterooligomeric protein complex required for MVB sorting Dev Cell 2002 3 271 282 12194857 10.1016/S1534-5807(02)00220-4
Babst M Katzmann DJ Snyder WB Wendland B Emr SD Endosome-associated complex, ESCRT-II, recruits transport machinery for protein sorting at the multivesicular body Dev Cell 2002 3 283 289 12194858 10.1016/S1534-5807(02)00219-8
Katzmann DJ Babst M Emr SD Ubiquitin-dependent sorting into the multivesicular body pathway requires the function of a conserved endosomal protein sorting complex, ESCRT-I Cell 2001 106 145 155 11511343 10.1016/S0092-8674(01)00434-2
Katzmann DJ Stefan CJ Babst M Emr SD Vps27 recruits ESCRT machinery to endosomes during MVB sorting J Cell Biol 2003 162 413 423 12900393 10.1083/jcb.200302136
Odorizzi G Babst M Emr SD Fab1p PtdIns(3)P 5-kinase function essential for protein sorting in the multivesicular body Cell 1998 95 847 858 9865702 10.1016/S0092-8674(00)81707-9
Pornillos O Alam SL Rich RL Myszka DG Davis DR Sundquist WI Structure and functional interactions of the Tsg101 UEV domain EMBO J 2002 21 2397 2406 12006492 10.1093/emboj/21.10.2397
Schmidt AE Miller T Schmidt SL Shiekhattar R Shilatifard A Cloning and characterization of the EAP30 subunit of the ELL complex that confers derepression of transcription by RNA polymerase II J Biol Chem 1999 274 21981 21985 10419521 10.1074/jbc.274.31.21981
Kamura T Burian D Khalili H Schmidt SL Sato S Liu WJ Conrad MN Conaway RC Conaway JW Shilatifard A Cloning and characterization of ELL-associated proteins EAP45 and EAP20. a role for yeast EAP-like proteins in regulation of gene expression by glucose J Biol Chem 2001 276 16528 16533 11278625 10.1074/jbc.M010142200
Yeghiayan P Tu J Vallier LG Carlson M Molecular analysis of the SNF8 gene of Saccharomyces cerevisiae Yeast 1995 11 219 224 7785322
Gajiwala KS Burley SK Winged helix proteins Curr Opin Struct Biol 2000 10 110 116 10679470 10.1016/S0959-440X(99)00057-3
Zheng N Fraenkel E Pabo CO Pavletich NP Structural basis of DNA recognition by the heterodimeric cell cycle transcription factor E2F-DP Genes Dev 1999 13 666 674 10090723
Holm L Sander C Dali: a network tool for protein structure comparison Trends Biochem Sci 1995 20 478 480 8578593 10.1016/S0968-0004(00)89105-7
Bowers K Lottridge J Helliwell SB Goldthwaite LM Luzio JP Stevens TH Protein-protein interactions of ESCRT complexes in the yeast Saccharomyces cerevisiae Traffic 2004 5 194 210 15086794
Howard TL Stauffer DR Degnin CR Hollenberg SM CHMP1 functions as a member of a newly defined family of vesicle trafficking proteins J Cell Sci 2001 114 2395 2404 11559748
Katoh K Shibata H Hatta K Maki M CHMP4b is a major binding partner of the ALG-2-interacting protein Alix among the three CHMP4 isoforms Arch Biochem Biophys 2004 421 159 165 14678797 10.1016/j.abb.2003.09.038
Xie W Li L Cohen SN Cell cycle-dependent subcellular localization of the TSG101 protein and mitotic and nuclear abnormalities associated with TSG101 deficiency Proc Natl Acad Sci U S A 1998 95 1595 1600 9465061 10.1073/pnas.95.4.1595
Burchett SA Flanary P Aston C Jiang L Young KH Uetz P Fields S Dohlman HG Regulation of stress response signaling by the N-terminal dishevelled/EGL-10/pleckstrin domain of Sst2, a regulator of G protein signaling in Saccharomyces cerevisiae J Biol Chem 2002 277 22156 22167 11940600 10.1074/jbc.M202254200
Carlson M Regulation of glucose utilization in yeast Curr Opin Genet Dev 1998 8 560 564 9794821 10.1016/S0959-437X(98)80011-7
Alam SL Sun J Payne M Welch BD Blake BK Davis DR Meyer HH Emr SD Sundquist WI Ubiquitin interactions of NZF zinc fingers EMBO J 2004 23 1411 1421 15029239 10.1038/sj.emboj.7600114
4 CCPN The CCP4 suite programs for protein crystallography Acta Cryst 1994 D50 760 763
Sheldrick GM Schneider TR SHELXL: High-resolution refinement Methods Enzymol 1997 277 319 343
Terwilliger TC Berendzen J Automated MAD and MIR structure solution Acta Crystallogr D Biol Crystallogr 1999 D55 849 861 10089316 10.1107/S0907444999000839
Terwilliger TC Maximum likelihood density modification Acta Crystallogr D Biol Crystallogr 2000 D56 965 972 10944333 10.1107/S0907444900005072
Jones TA Zou J-Y Cowan SW Kjeldgaard M Improved methods for building protein models in electron density maps and location of errors in these models Acta Crystallogr D Biol Crystallogr 1991 A47 110 119
Brünger AT Adams PD Clore GM DeLano WL Gros P Grosse-Kunstleve RW Jiang J-S Kuszewski J Nilges M Pannu NS Read RJ Rice LM Simonson T Warren GL Crystallography & NMR system: a new software suite for macromolecular crystallography Acta Crystallogr D Biol Crystallogr 1998 D54 905 921 10.1107/S0907444998003254
Laskowski RA McArthur MW Moss DS Thornton JM PROCHECK: a program to check the stereochemical quality of protein structures J Appl Cryst 1993 26 283 291 10.1107/S0021889892009944
Kraulis PJ MOLSCRIPT: a program to produce detailed and schematic plots of protein structures J Appl Cryst 1991 24 946 950 10.1107/S0021889891004399
Merritt EA Bacon DJ Raster3D: photorealistic molecular graphics Methods Enzymol 1997 277 505 524
Gouet P Courcell E Stuart DI Metoz F ESPript: multiple sequence alignments in PostScript Bioinformatics 1999 15 305 308 10320398 10.1093/bioinformatics/15.4.305
Nicholls A Sharp KA Honig B Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons Proteins 1991 11 281 296 1758883
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG The CLustalX 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
Kabsch W Sander C Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features Biopolymers 1983 22 2577 2637 6667333
Kleywegt GJ Jones TA A super position CCP4/ESF-EACBM Newsletter Prt Cryst 1994 31 9 14
| 15579210 | PMC539351 | CC BY | 2021-01-04 16:29:59 | no | BMC Struct Biol. 2004 Dec 4; 4:10 | utf-8 | BMC Struct Biol | 2,004 | 10.1186/1472-6807-4-10 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-5-1861557163410.1186/1471-2105-5-186Research ArticleConstraint Logic Programming approach to protein structure prediction Dal Palù Alessandro [email protected] Agostino [email protected] Federico [email protected] Dipartimento di Matematica e Informatica, Università di Udine. Via delle Scienze 206, 33100 Udine, Italy2 Dipartimento di Scienze e Tecnologie Biomediche, Università di Udine, P.le Kolbe 4, 33100 Udine, Italy2004 30 11 2004 5 186 186 9 7 2004 30 11 2004 Copyright © 2004 Dal Palù 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 protein structure prediction problem is one of the most challenging problems in biological sciences. Many approaches have been proposed using database information and/or simplified protein models. The protein structure prediction problem can be cast in the form of an optimization problem. Notwithstanding its importance, the problem has very seldom been tackled by Constraint Logic Programming, a declarative programming paradigm suitable for solving combinatorial optimization problems.
Results
Constraint Logic Programming techniques have been applied to the protein structure prediction problem on the face-centered cube lattice model. Molecular dynamics techniques, endowed with the notion of constraint, have been also exploited. Even using a very simplified model, Constraint Logic Programming on the face-centered cube lattice model allowed us to obtain acceptable results for a few small proteins. As a test implementation their (known) secondary structure and the presence of disulfide bridges are used as constraints. Simplified structures obtained in this way have been converted to all atom models with plausible structure. Results have been compared with a similar approach using a well-established technique as molecular dynamics.
Conclusions
The results obtained on small proteins show that Constraint Logic Programming techniques can be employed for studying protein simplified models, which can be converted into realistic all atom models. The advantage of Constraint Logic Programming over other, much more explored, methodologies, resides in the rapid software prototyping, in the easy way of encoding heuristics, and in exploiting all the advances made in this research area, e.g. in constraint propagation and its use for pruning the huge search space.
==== Body
Background
Notwithstanding the continuous improvement in predictive methods, witnessed every two years by the world wide CASP experiment [1,2], predicting the structure of a protein, given its sequence, is still in general beyond our capabilities. Brute force approaches, like exhaustive conformational searches or molecular dynamics simulations of the folding process, are precluded by the computing power available at present. Alternative, faster methods have been developed along two main lines:
1. assemblying the structure of a protein using structural fragments of similar sequences, available in the protein structure repository (the Protein Databank [3]), and later screening the feasibility of the resulting structures, using energetic criteria;
2. representing the protein chain by a highly simplified model which is, hopefully, treatable.
This second class of approaches is appealing in many respects [4]: first, the linkage between kinetics and thermodynamics of protein folding process and the basic intramolecular interactions is more easily addressable, because of the lesser number of variables. Second, the use of a simplified model agrees with the idea that details of atomic interactions between aminoacids are less important than the overall character of these interactions, because protein structure is flexible and can accommodate changes in the volume and shape of aminoacids much better than changes in their character (e.g. polar vs. hydrophobic [5]). Besides aiming at catching essential features of the protein folding process, simplified models have important computational advantages: generating and evaluating the energy of a conformation is efficiently done due to the reduced number of variables. A less evident benefit is that sampling (e.g. by molecular dynamics simulation or Monte Carlo methods) may be much more efficient due to the smoothness of energy surface due, once again, to the reduced number of degrees of freedom. Many lattice models have been used for simplified representation of proteins, up to date. Their capability of reproducing the secondary structure of proteins, as well as their relative arrangement has been reviewed by Godzik et al. [6]. A reasonable tradeoff between accuracy and the need to keep limited the number of base vectors is achieved by the face centered cubic () lattice studied by Toma and Toma [7]. In particular both α-helices and β-strands are modelled with a very low RMSD from standard regular structures. Lattice models have been used mainly for understanding general properties of proteins, rather than for real predictive tasks, although their use, especially in hierarchical protocols has been proposed and realized. In particular, the (210) lattice has been used successfully by Skolnick and Kolinski in prediction of a small beta protein [8] and many other useful applications have been reported since these earlier works (see e.g. for recent successful applications [9,10] and also the two recent reviews [4,11]). A deep analysis of realistic lattice models of proteins proposed so far is definitely out of the scope of the present work, but there are few aspects of lattice models of proteins which need to be mentioned. The successful application of a lattice model depends obviously on the efficiency in generating conformations and searching for local minima. This aspect is dealt in the present work using Constraint Logic Programming, and taking advantage of all theoretical and implementative developments that have been realized in this context. The approach (and related languages) has been very seldom applied in the context of protein modeling and it has not been used for realistic protein structural predictions, to the best of our knowledge. A different, but equally important, aspect concerns the reliability of the model itself and of the forcefield used to evaluate conformational free energy. This aspect will not be dealt with by this work. An appropriate forcefield must take into account both local propensities to adopt a particular secondary structure (which ultimately depend on aminoacids' covalent structure and bulkiness) and their tendency to be in contact (which ultimately depends on their physico-chemical character). Contact potentials have been derived by many groups (see e.g. [12,13]) based on the observed versus expected contacts stored in the database. A similar approach could be followed in order to derive a torsional potential in order to describe local conformational propensities. However, it is not obvious how these potentials should be derived for lattice models and how the two potentials are to be considered together. These problems are not investigated here. Rather we consider contact potentials previously derived by our group from statistical analysis of the database [13], which are expected not to be accurate for a lattice model, but nevertheless should be able to reproduce essential features of aminoacid interactions. The local propensity to adopt a particular secondary structure can be computed by predictive methods [14]. However, for the small peptides analyzed in this paper, the correct secondary structure is selected from the deposited structures for testing purposes.
Constraint Logic Programming (briefly, ) [15,16] is a declarative programming paradigm particularly well-suited for encoding combinatorial minimization problems. It is the natural merger of the two declarative paradigms known as Constraint Solving and Logic Programming.
One of the peculiar features of is the independence of the problem modeling and of the search's strategy. Problem modeling is based on traditional declarative programs in which one can use the built-in notion of constraint. Constraints are first-order formulas concerning variables that can assume values in some domains. The scheme is general. Various possible constraints and domains can be used. However, for combinatorial problems it is common to use finite domain constraints, namely arithmetic constraints between arithmetic expressions, where variables range over finite subsets of ℕ. Constraint Logic Programming over Finite Domains is known as (). We briefly introduce this programming paradigm with a simple example. Let us consider three variables X, Y, Z that denote the number of possible items of some kind.
domain([X, Y, Z], 1, 10)
is a constraint that states that the three variables X, Y, Z have (finite) domain {1, 2, ..., 10}. Suppose we wish to state that the weight of each item of X is 3, of Y is 4, and of Z is 5 and the total weight of selected items must be less than or equal to 40. Moreover, we wish to state that the number of items of X plus those of Y must be less than those of Z. This can be simply stated as:
3 * X + 4 * Y + 5 * Z ≤ 40, X + Y <Z
We have modeled a sort of knapsack problem using (). In general, in the modeling stage we can use constraints as well as declarative programs involving them.
Solution's search is performed by a constraint solver that is available in the language. The constraint solver uses constraints for sensibly pruning the search tree. One of the main capabilities is called constraint propagation. Constraint propagation reduces the domains of the variables eliminating those values that cannot lead to constraint solutions. For instance, in the considered example, constraint propagation reduces the domains of the variables X, Y, and Z to {1, ..., 4}, {1, ..., 4}, and {3, ..., 6}, respectively. For finding a possible solution, a further built-in capability – the labeling predicate – can be used. We can look for a generic solution as well as for a solution minimizing some function. In the example above, we could ask for minimizing the function -2X2 + Y + 4Z. This can be done by adding a constraint of the form:
F = -2 * X * X + Y + 4 * Z, labeling([minimize (F)], [X, Y, Z]).
The constraint solver then exploits the solution's search using constraint propagation and branch-and-bound techniques returning the answer:
F = 3, X = 3, Y = 1, Z = 5
The library clpfd of SlCStus Prolog [17] allows to effectively program in this framework. Let us observe that it is not required that F be a linear function.
The above described approach to optimization combinatorial problems is the so-called Constrain & Generate technique introduced as opposed to the Generate & Test technique of the classical Logic Programming approach (see, e.g. [18]). In the latter approach, a first phase generates non-deterministically a possible solution, and then the deterministic test-phase checks whether the solution is acceptable or not. If the search space is exponential, this technique is not applicable. In the former approach, a first deterministic phase introduces a number of constraints, then a non-deterministic phase starts the generation of the solutions' space. The constraints introduced allow to sensibly prune the solutions' space in order to make the procedure effective. Moreover, in this phase one can take advantage from language built-in strategies (such as constraint propagation, branch and bound) and it is possible to further drive the solution search by means of problem-dependent heuristics.
We have followed the Constrain & Generate programming style for encoding the protein structure prediction problem. As a matter of fact, the main predicate of our solution is of the form reported in Figure 1.
In the definition of the predicate constrain the protein structure prediction problem is modeled using constraints. In particular, the energy function is encoded in the Energy parameter, The predicate solution_search is aimed at looking for the solution minimizing the Energy parameter. The other predicates are auxiliary predicates. initialization resets some parameters, protein recovers the relevant input (see also Methods Section), writetime and print_results are output predicates. The constraint predicate is defined using several predicates each of them modeling one of the properties of the problem. For instance, the predicate next_constraints sets the distance between consecutive aminoacids (see Figure 2).
Briefly, next_constraints recursively calls the predicate next for each pair of consecutive aminoacids. Assume that <X1, Y1, Z1> and <X2, Y2, Z2> are the variables that will store the positions of a consecutive pair of aminoacids, then the predicate next states that |X1 - X2| + |Y1 - Y2| + |Z1- Z2| = 2 and that |X1 - X2| ∈ {0, 1}, |Y1 - Y2| ∈ {0, 1}, |Z1 - Z2| ∈ {0, 1}. This is exactly the notion of adjacency in the face-centered cubic lattice of size 2 that we have used (see also the Methods Section).
Results and discussion
Constrained optimization problem in ()
In Table 1 we report the results of the experiments with the () code described in the Methods Section. All tests are done using SICStus PROLOG 3.11.1 [17] and a PC P4, 3.06 GHz. The structures of the protein model systems analyzed are known and stored in the PDB [3]. In the protein model systems 1LE3, 1PG1, and 1ZDD terminal protecting groups have been neglected.
From left to right, the meaning of each column is as follows: the protein PDB identification code, the number N of aminoacids, the execution time, the energy of the best model found and its RMSD from the native structure for all the residues and for the core residues only. When there is not explicitly written "limit" it means that the program successfully terminated in the time reported; otherwise the program terminated due to time limit. We wish to observe that the results with time limit 10 h/24 h are typically computed in few hours. The rest of the time is used to further explore the solutions' space.
When a CF = η is reported a further constraint on the compactness ratio η is added before the search. CF = η bounds the linear distances |Xi - Xj|, |Yi - Yj|, and |Zi - Zj| between all pair of residues i and j to ηN where N is the length of the primary list. If η is low (e.g. 0.2), this constraint imposes a compact form to the protein and strongly reduces the running time.
One of the structural constraints considered is the presence of disulfide bonded residues (ssbonds). The rigid structure of the lattice is such that a low value of Euclidean distance (e.g., 2) between ssbonds often precludes all possible solutions. For this reason the default is chosen as 6. However, in some cases we tried computations with lower value. In these cases in the table the text ss = γ is reported.
The secondary structure, as computed from the deposited structure in PDB, has been input as constraint. As a unique exception, in the case of 1VII(*) we have instead predicted it using the GOR IV secondary structure prediction method [19].
The predicted structures have been also transformed into all atoms models as described in the Detailed models from lattice models Section. There is some improvement in general on RMSD from native structure. This is especially significant when the starting structure is already close to the native one, being not merely due to increasing compactness of the structure. It is moreover reassuring that the procedure we are discussing is able to recover realistic models starting from the very simplified lattice models. The RMSDs of the resulting detailed models from the corresponding native structures are reported in Table 2. In order to assess the quality of the detailed model, the trace of the native structure and the reconstructed and optimized all-atom model are shown in Figure 3 for the core residues (7 to 30) of the WW domain (PDB id.: 1E0M).
We conclude the section comparing some results of our prediction with those returned by the well-known HMMSTR/Rosetta Prediction System [20]. This program does not use a lattice as underlying model: aminoacids are free to take any position in ℝ3. For the sake of comparison, we have used it as an ab-initio predictor (precisely, we have disabled the homology and psi-blast options). The comparison is obviously not fair because in our case secondary structure is known and not predicted. Times are obtained from the result files, but it is not clear to which machine/CPU occupation they refer. Results are reported in Table 3. HMMSTR/Rosetta prediction runs presumably faster, but our predictions (which however include known secondary structure) improve the RMSD (except for one case).
Constrained molecular dynamics simulation
We have used secondary structure information in conjunction with the well-established methodology of molecular dynamics simulations in order to implement a procedure similar to the one implemented using on the lattice. Secondary structure elements have been imposed through a constraining potential as described in the Methods Section. In order to search the conformational space a simulated annealing procedure has been adopted. Globularity of the simulated proteins is forced by a harmonic constraint on the radius of gyration.
The simulation time, ranging approximately between one and four CPU days, required for folding each protein on a 1.533 GHz AMD Athlon processor is reported in Table 2. The columns (from left to right) in Table 2 report the PDB identification code of the protein, the number of residues, the RMSD from native structure computed on Cα atoms on the whole protein and only on core residues and the simulation time. The last column reports the RMSD from native structure for models obtained by after addition of all atoms and energy minimization as described in the Methods Section.
The simulation time needed for obtaining structures similar to native structures increases with the size of the protein both for the increasing size of the system and for the longer simulated annealing runs needed because of increasing complexity of the free energy landscape. Unfortunately a safer scheme would employ substantially longer simulation times.
This fact prompts for searching alternative ways to employ the same ideas.
The results in terms of RMSD from native structure support the idea that folding may be achieved, at least in simulation, by a hierarchical approach where local secondary structure elements are formed first and later their arrangement and contacts are optimized. A similar conclusion has been reached using a different model by Maritan and coworkers [21]. The RMSD on core residues is, in all but one case, less than 5.0 Å. In four out of six cases the RMSD on core residues is close to 4.0 Å. In the worst case, which is also the longest simulated chain, the RMSD on core residues is 7.1 Å.
Conclusions
The purpose of the present work was to demonstrate that the protein folding problem can be approached by a well-established programming paradigm like . With respect to the few applications reported in the literature so far using the same methodology [22], mainly on the HP protein model [23,24], the present work takes a step further towards more realistic modeling. Notwithstanding the use of a protein simplified lattice model with a simple contact potential realistic models for a few small proteins have been generated by using . In the present application the known secondary structure of the protein has been imposed as a constraint. has been applied on face centered cubic lattice models of proteins where every aminoacid is represented by a single point on the lattice that can take one out of six possible positions with respect to the previous three aminoacids. It is immediately seen that the time needed for a systematic space search for such model grows exponentially with the number of free aminoacids. is a programming paradigm that is suited for the solution of optimization combinatorial problems. In the problem and the related heuristics are extremely natural to be programmed. Moreover, the constraint propagation allows to control the search in the huge solution's space.
The results obtained using this approach and reported in Tables 1 to 3 show that for small proteins a solution for the optimization problem is obtained in less than few hours. For the larger proteins studied here the inaccuracies of both the lattice model and contact potential prevent finding a compact solution. These problems are more likely to appear with increasing size of the protein and when the length of non-constrained chain connecting two secondary structure elements is short, because the lattice allows a limited set of conformations.
Further work is being devoted towards a more realistic modeling representation of the protein, with at least two centers of interaction per residue, and towards refinement of the potential function by including a term for rotamer preferences. This term should map on the lattice the directional preferences of each unit with respect to the previous three units. Each of the six possible next positions for each unit should be weighted by an energy term derived from database analysis.
Also the optimal size of non constrained parts of the chain will be determined in order to allow more possible relative orientations among constrained secondary structure elements, possibly without increasing significantly the computation time. At present, however, when the positions of all atoms are reconstructed from the lattice Cα trace, the RMSD on core residues of the resulting models, after energy minimization, compared to native structures, is as low as 4.8 Å for the thermostable domain of villin headpiece (PDB id.: 1VII), 3.6 Å for the WW domain (PDB id.: 1E0M), 2.3 Å for the coat protein-binding domain of bacteriophage P22 (PDB id.: 2GP8).
It should be also noted that both the thermostable domain of villin headpiece and the WW contain three secondary structure elements that can be arranged in different ways in order to produce a compact structure. The low RMSD is therefore significant.
A comparable protocol employing a molecular dynamics simulated annealing procedure still leads to superior results for larger proteins, as expected because the protein representation is more accurate, but it takes longer execution times between one and four days on a 1.5 GHz P3 machine.
Recent results have shown that simplified models and more refined models can be employed successfully in hierarchical modeling procedures [9,10]. The results obtained in the present work suggest that could be useful for finding starting conformations for further refinement.
Methods
The protein structure prediction problem as a minimization problem
The sequence of aminoacids defining a protein is called primary structure. This structure uniquely determines the (3D) native conformation, also known as tertiary structure. The protein structure prediction problem is the problem of predicting the tertiary structure of a protein given its primary structure. The native tertiary structure minimizes the global free energy of the protein.
Abstraction level
We consider each aminoacid as a single sphere centered in its Cα atom; the distance between two consecutive Cα atoms is assumed to be 3.8 Å Recent results (see, e.g., [13]) show that a contact between two residues, when represented only by their Cα atoms, is optimally defined for Cα - Cα distances shorter than 6.4 Å The number is obtained as the sum of the radius of the two Cα carbon atoms we are dealing with (2 x 1.9 Å) and the value of 2.6 Å empirically determined in [13] for van der Waals surface contact. A table that points out the energy associated to pairs of aminoacids in contact has been developed [12,13]. Let us denote by Pot(x, y) the energy value associated to a contact between aminoacids x and y (the order is immaterial); this value can either be positive or negative, according to the pair x, y.
Lattice model
According to [25] we use the Face-Centered Cubic Lattice () that allows realistic angles between consecutive residues. The lattice is composed by cubes of size 2, where the central point of each face and the vertices are admitted. Thus, the domain consists in a set of triples <x, y, z> where <x, y, z ∈ >. We recall that given a point <x, y, z>, its 2-norm is: ||<x, y, z>|| = . Given two points p1 and p2, ||p1 - p2|| is known as their Euclidean distance.
Going back to the lattice, two points at Euclidean distance are linked together, forming a lattice unit, corresponding to the distance of 3.8 Å. In this lattice, each point is adjacent to 12 neighboring points. A contact is defined between two non adjacent residues placed on two vertices of a side of a cube (i.e. they have Euclidean distance equal to 2, corresponding to 5.4 Å). This number can be considered a good approximation of the limit of 6.4 Å described above.
Mathematical formalization
In this setting, it is possible to formalize the protein folding problem as an optimization problem. Given a sequence S = s1 ... sn, with si aminoacids, a fold of S is a function ω : {1, ..., n} → such that: ||ω(i) - ω(i + 1)|| = and ||ω(i) - ω(j)|| ≥ 2 for i ≠ j. The first constraint states that consecutive aminoacids have a fixed distance, corresponding to one lattice unit; the second that each aminoacid occupies a unitary sphere and that two spheres cannot overlap.
The protein folding problem can be reduced to the optimization problem of finding the fold ω of S such that the following energy is minimized [26,27]:
where contact(ω(i), ω(j)) is 1 if ||ω(i) - ω(j)|| = 2, 0 otherwise. To avoid solutions equivalent modulo simple symmetries, other constraints can be added on the first positions.
Complexity issues
The decision version of this problem (and even of its HP-abstraction) is proven to be NP-complete on various lattices [28,29]. However, we do not want to solve the problem for proteins of arbitrary length. Solving it for length N = 200–300 could be considered as an important contribution to biological sciences and there are yet such results using the HP-abstraction [30]. Thus, in spite of its NP-completeness, it is important to understand the size of the solution's space. The size of the solution's space is the number of self-avoiding walks on the lattice that can be approximated by the formula (cf., e.g., [31])
SAWfcc = 1.26N0.162(10.0364)N (2)
This formula should modify in the presence of additional constraints as mentioned later.
Main implementation issues
Our implementation of the protein folding minimization problem described in the above sections is based on the code briefly introduced in the Background Section. The complete program and related material can be found in [32]. The program consists of ~2000 lines and, once loaded in SICStus Prolog, one may call goals of the kind reported in Figure 4, where Protein_Name is a standard PDB identification code, such as 1ENH. Time is the maximum amount of time in seconds that we let to the runtime; the default is 10 hours. CompactFactor allows to impose an additional constraint on the maximal distance between every pair of aminoacids. The rationale behind this additional constraint stems from the observation that protein structures are more compact than expected based on a freely rotating chain model [33]. In particular, the average end-to-end distance for a freely rotating chain model is approximated by where ℓ is the length of each unit and α is the cosine of the angle made by each unit with the direction of the preceding unit. The average end-to-end distance is clearly related to the average maximal dimension of the chain. Based on a survey of protein structures Huang and Powers derived the following approximated formula for the radius of gyration (in Å): 2.2N0.38 [34]. Note that the exponent is less than 0.5 which is an underestimate of the exponent for a self avoiding walk. For a uniform density sphere the diameter is the radius of gyration. The default value for CompactFactor was therefore assumed to be approximately equal to times the radius of gyration which in turn was computed by the empirical formula 2.2N0.38 [34].
The auxiliary file data.pl stores the Primary and Secondary structures of the proteins that one wishes to test, as, for instance in the example reported in Figure 5. The output in standard PDB format [3] is printed either on the screen or in a file named output-Protein_name.pdb.
Constraints
The intrinsic complexity of the problem forces us to introduce several other constraints. For instance, we constrain the sum of the coordinates of each aminoacid in the lattice to be even (a property of the lattice) and we add some constraints for avoiding equivalent symmetric solutions. In what follows, we refer to predicate names as used in the code. avoid_symmetries removes redundant admissible conformations equivalent to others modulo some symmetries and/or rotations. The predicate assigns immediately three consecutive aminoacids positions (in the Tertiary list).
With distance_constraints, we also impose that two non consecutive residues must be separated by more than one lattice unit, to reflect the steric interaction between the Cαs modelling aminoacids.
As described above, compact_constraints imposes that, for every pair of aminoacids, the norm of the projection of their distance on each x, y, z coordinate, is smaller than CompactFactor × N.
Further constraints are related to angles. In the lattice, the angle between three consecutive residues can assume values in {60°, 90°, 120°, 180°}. In real proteins, steric occupancy and energetic potential show a clear distribution of bend angles in the range 90°–150° [7,35]. When transferring on lattice, it is a good approximation to exclude 60° and 180° angles, as unfeasible. This constraint allows us to restrict the search space from a number close to 10N (cf. formula (2)) to a number close to 5N.
As said in the Lattice model Section, a contact is generated by two non consecutive aminoacids with Euclidean distance less than or equal to 2. As a consequence of the constraints applied, it suffices to check for a contact when the lattice distance equals 2, since distance_constraints excludes from the domain the possibility to place two non consecutive aminoacids at one lattice unit.
We also impose constraints coming from secondary structure information. Secondary structure can be predicted with good approximation (e.g., [36]). In our set of data we have collected such information from the Protein Data Bank. We represent secondary structure information as helix(i, j): elements i, i + 1, ..., j of the input sequence form an α-helix; strand(i, j): elements i, i + 1, ..., j are in a β-strand; ssbond(i, j): there is a disulfide bridge between element number i and j.
We use an auxiliary list called Indexes that stores torsional angles defined by four consecutive aminoacid positions. Due to lattice structure and our constraints, every four consecutive aminoacids can form only 6 discrete angles. Thus, each variable in Indexes can assume a value i from {0, ..., 5}, representing torsional angles of 0°, 60°, 120°, 180°, 240°, 300°, respectively. With these conventions, helices are approximated by sequences of indexes of the form 5, 5, 5, ... while β-strands are associated to sequences of the form 3, 3, 3, .... Note that specifying the coordinates of three points (i.e. to place and orient the protein) and the indexes, uniquely determines the conformation, ssbond(i, j), introduces a maximum distance constraint between the aminoacids i and j. The predicate energy_constrain is developed using an auxiliary symmetric matrix M. The optimal fold is reached when the sum of M elements is minimal. During the labeling phase, the information stored in M is used to control the minimization process and to cut the search tree.
Labeling stage
To reduce the size of the solution's space visited during execution, we have replaced the built-in labeling predicate with an ad-hoc constraint-based solution search predicate, called solutions_search. We describe here briefly the main features of this predicate and of its auxiliary predicates.
solutions_search • If the Tertiary list or the Indexes list is ground (already computed), then it terminates the folding process (possibly, after a call to the built-in labeling).
• Otherwise, it calls choose_labeling. When this procedure terminates, it calls recursively solutions_search. Termination is guaranteed by the fact that each call to choose_labeling reduces the number of non-ground variables.
choose_labeling • If the number of variables to be instantiated is low (in our code less than 4), it calls the built-in labeling.
• Otherwise, it calls selection_strategy. This predicate computes several subsequences of the list of Indexes. Each subsequence consists of alternations of ground elements and non-ground variables. selection_strategy selects the most known subsequence, namely the one containing the smallest ratio of variable over ground indexes, preferring the ones that include a ssbond. If in the selected subsequence there are too many variables, an arbitrary subsequence cut is done. After the subsequence is selected, the procedure labeling_new_launch is called.
labeling_new_launch It calls the auxiliary predicate labeling_new but stops the solution search when the global runtime is greater than the input time limit. If this is the case, the best computed solution is returned.
labeling_new This procedure receives the chosen sublist to be folded. Each index variable in it, is assigned an admissible value between 0 and 5. The order of values that is tried for each index is described by a pre-computed auxiliary list. For each torsional index, a frequency statistics of the 6 indexes is pre-computed and extracted from the PDB, according to the specific aminoacid sequence involved locally. We use this information to direct the search and explore first the most common torsional angles, in the hope that this selection rule reflects nature's strategy.
Moreover, the energy associated to the fold is minimized. For doing that, after each instantiation of a fixed number t of variables in a phase, we collect the best known ground admissible solution, its energy and its associated potential matrix. We compare the current status to history and decide if it is reasonable to cut the search tree. In particular, we designed a heuristic that allows to control the effectiveness of the cut, adapting it dynamically to the status of the fold. Practically, when the protein is partially specified, we estimate the ratio between ground and non-ground variables in the potential matrix. If the ratio is low (i.e. the protein is poorly determined), we allow the current energy to be worse than the corresponding counterpart in the best fold so far reached. When the ratio is high (i.e. protein almost folded) we constrain the current energy to be slightly lower than the previous best known.
Molecular dynamics simulations
In order to have a fair comparison with a similar approach using all-atom protein models we built detailed all atom models for six proteins in the studied set (namely those with PDB id. code: 1VII, 1E0M, 2GP8, 1ENH, 2IGD, 1YPA) and imposed, through torsional constraints, the secondary structure geometry found in the native structure. The constraining potential was 100 * (θ - θ0)2 kcal/(mol rad2). The reference target angles (i.e. θ0 in the previous formula) were set to φ = -139 and ψ = 135 for residues in β-strand and to φ = -48 and ψ = -57 for residues in α-helices. For all constrained residues also the ω dihedral angle was constrained at 180 degrees.
The chain was first built fully extended and minimized by 400 steepest descent minimization steps and by 500 conjugate gradients minimization steps.
The protein was then heated in 10 ps up to 900 K in 20000 steps using a timestep of 0.0005 ps. Then the temperature was lowered down to 270 K in 20 steps. During each step molecular dynamics simulation was carried out for 100 ps for a total simulation time of 2 ns.
Simulations used the Generalized Born implicit solvent method [37] as implemented in the program CHARMM [38] with standard parameters for proteins. The forcefield used was CHARMM v.27 [39].
In order to obtain globular protein during simulation a constraint on the radius of gyration (computed only on Cα atoms) was imposed. The target radius was decreased during the simulation from a value proper of an extended conformation down to the value given by 2.2N0.38 [34] where N is the number of residues.
The potential used for enforcing compactness was: kcal/mol, where , n is the number of atoms, rcg is the center of geometry of the same group of atoms, and Rg0 is the target gyration radius which is decreased during simulated annealing down to the theoretical value based on the formula cited above.
Detailed models from lattice models
The models obtained by described here may be converted into all-atom models which are realistic models of proteins. As a test the structures of all the proteins tested by simulated annealing described above were converted using the Maxsprout server [40] into an all heavy atom model. Hydrogens have been added using the module HBUILD in the program CHARMM [38] and the resulting structure was relaxed by energy minimization (using a distance dependent dielectric constant). First a minimization was performed with all backbone atoms fixed, then only Cα atoms were fixed and finally a 100 ps molecular dynamics simulation (following a heating phase of 10 ps) using the Generalized Born implicit solvent model was performed. The resulting structure at the end of the simulation was energy minimized.
The initial minimizations required 1500 minimization steps each, because the starting structures were built from the lattice models. The final minimization, on the structure relaxed by molecular dynamics simulation, employed 900 minimization steps. During molecular dynamics simulation the radius of gyration and backbone torsion angles corresponding to residues constrained in the () procedure were constrained as described above.
Figures and Tables
Figure 1 Main program predicate.
Figure 2 Code for stating that consecutive aminoacids must be in adjacent lattice points.
Figure 3 Native (yellow) and model after all-atom reconstruction and optimization (red) for WW domain (PDB id. 1E0M). The trace of core residues (7–30) is shown.
Figure 4 Program execution modes.
Figure 5 Protein sequence and secondary structure representation
Table 1 Experimental results
Name N Time Energy RMSD
1LE0 12 1.3 s -9040 2.8 / 2.6 (2–11)
1KVG 12 7.3 s -14409 2.7 / 2.4 (3–11)
1LE3 16 2.3 s -13653 3.0 / 2.7 (2–1
5)
1EDP 17 20.4 s -19389 4.3 / 1.1 (9–15)
1PG1 18 14.6 s -10126 6.0 / 5.2 (4–17)
1ZDD 34 300 s (limit) -20412 5.6 / 5.6 (5–34)
17 m25 s -22350 4.0 / 4.0 (5–34)
1VII 36 300 s (limit) -20860 10.4 / 6.7 (4–32)
1000 s (limit) -22377 9.1 /6.3 (4–32)
7 h42 m -26408 10.2 / 7.8 (4–32)
CF = 0.3 3 h58 m -28710 8.0 / 7.4 (4–32)
1VII(*) 36 300 s (limit) -17948 9.2 / 7.3 (4–32)
1000 s (limit) -17948 9.2 /7.3 (4–32)
3 h20 m -21211 10.3 / 6.9 (4–32)
1E0M 37 300 s (limit) -13830 6.5 / 5.8 (8–29)
1200 s (limit) -24613 8.4 / 3.6 (8–29)
10 h (limit) -26592 8.8 / 3.4(8–29)
24 h (limit) -30163 8.9 / 4.4 (8–29)
2GP8 40 300 s (limit) -10303 10.5 / 8.9 (6–38)
1000 s (limit) -24748 4.1 / 3.5 (6–38)
10 h (limit) -26196 4.9 / 3.5 (6–38)
10 h39 m -26196 4.9 / 3.5 (6–38)
1ED0 46 300 s (limit) -29970 7.3 / 4.1 (3–40)
1000 s (limit) -32369 8.6 / 7.1 (3–40)
9 h38 m -38218 8.0 / 7.2 (3–40)
1ENH 54 300 s (limit) -12480 10.4 / 8.9 (8–52)
1000 s (limit) -12480 10.4 / 8.9 (8–52)
10 h(limit) -23373 9.9 / 8.6 (8–52)
24 h (limit) -23373 9.9 / 8.6 (8–52)
6PTI 58 300 s (limit) no sol.
1000 s (limit) -29709 10.0 / 9.7 (3–55)
10 h (limit) -37837 10.0 / 9.7 (3–55)
24 h (limit) -37837 10.0 / 9.7 (3–55)
CF = 0.25 48 h (limit) -42096 9.7 / 9.4 (3–55)
CF= 0.18 24 h (limit) -47451 10.9 / 10.7 (3–55)
2IGD 60 300 s (limit) -24158 19.3 / 16.3 (6–59)
1000 s (limit) -29178 19.0 /16.2 (6–59)
10 h (limit) -37479 16.9 / 15.0(6–59)
24 h (limit) -37479 16.9 / 15.0 (6–59)
CF = 0.17 4 h 59 m -40588 12.6 / 11.5 (6–59)
2ERA 61 300 s (limit) -28993 11.6 / 10.6 (3–55)
9 m28 s, -30746 12. 3/ 12.1 (3–55)
ss = 5 15 m13 s -31692 12.7/11.6 (3–55)
CF = 0.25, ss = 5 15 m12 s -33693 10.9/9.3 (3–55)
CF = 0.25, ss = 4 1000 s (limit) -32985 12.3/12.4 (3–55)
CF = 0.19, ss = 5 1000 s (limit) -34275 10.6/8.9 (3–55)
CF = 0.19, ss = 4 1000 s (limit) -38138 11.6/10.6 (3–55)
1SN1 63 300 s (limit) no sol.
1000 s (limit) -53888 13.0 / 10.5 (2–51)
10 h (limit) -57615 11.9/ 9.6 (2–51)
CF = 0.25, ss = 5 10 h (limit) -47121 8.6 / 8.1 (2–51)
1YPA 63 300 s (limit) -36626 16.1 / 9.4 (12–52)
1000 s (limit) -33886 17.1 / 10.9 (12–52)
10 h (limit) -33886 17.1 / 10.9 (12–52)
CF = 0.17 100 s (limit) -26297 12.5 / 10.5 (12–52)
CF = 0.17 10 h (limit) -60244 12.9 / 9.8 (12–52)
Table 2 Summary of molecular dynamics results
Name N RMSD (Å) Time RMSD ( + opt) (Å)
1VII 36 5.3 / 4.7 (4–32) 17.8 h (2 ns) 5.8 / 4.8 (4–32)
1E0M 37 5.5 / 4.0 (7–30) 26.3 h (4 ns) 8.7 / 3.6 (7–30)
2GP8 40 5.9 / 3.8 (6–38) 37.7 h (4 ns) 3.9 / 2.3 (6–38)
1ENH 54 5.9 / 5.0 (8–52) / 3.7 (8–36) 29.4 h (2 ns) 11.2 / 10.7 (8–52) / 4.7 (8–36)
2IGD 61 5.7 / 4.1 (6–59) 48.6 h (4 ns) 12.9 / 11.5 (6–59)
1YPA 64 9.2 / 7.1 (12–52) 116.9 h (8 ns) 11.8 / 9.4 (12–52)
Table 3 Comparison with Rosetta predictions
Name N Time RMSD Rosetta Time Rosetta RMSD
1ZDD 34 17 m.25 s. 4.0 5 m.35 s. 3.5
1VII 36 3 h.58 m. 7.4 (4–32) 5 m.35 s. 4.2
1E0M 37 10 h. 3.4 (8–29) 6 m.35 s. 7.7
2GP8 40 10 h. 3.5 (6–38) 6 m.35 s. 6.4
1ED0 46 10 h. 7.2 (3–40) 7 m.23 s. 8.9
==== Refs
Moult J Hubbard T Fidelis K Pedersen J Critical Assessment of Methods of Protein Structure Prediction (CASP): Round III Proteins: Struct Funct Genet 1999 2 6 10526346
Venclovas C Zemla A Fidelis K Moult J Assessment of progress over the CASP experiments Proteins: Struct Funct Genet 2003 585 595 14579350
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
Skolnick J Kolinski A Reduced models of proteins and their applications Polymer 2004 45 511 524
Bashford D Chothia C Lesk A Determinants of a protein fold. Unique features of the globin amino acid sequences J Mol Biol 1987 196 199 216 3656444
Godzik A Kolinski A Skolnick J Lattice representations of globular proteins: how good are they? J Comp Chem 1993 14 1194 1202
Toma T Toma S Folding simulation of protein models on the structure-based cubo-octahedral lattice with the Contact Interactions algorithm Protein Sci 1999 8 196 202 10210197
Skolnick J Kolinski A Simulations of the folding of a globular protein Science 1990 250 1121 1125
Xia Y Huang ES Levitt M Samudrala R Ab initio construction of protein tertiary structures using a hierarchical approach J Mol Biol 2000 300 171 185 10864507
Zhang Y Kolinski A Skolnick J TOUCHSTONE II: a new approach to ab initio protein structure prediction Biophys J 2003 85 1145 1164 12885659
L Mirny ES Protein folding theory: from lattice to all-atom models Annu Rev Biophys Biomol Struct 2001 30 361 96 11340064
Miyazawa S Jernigan RL Residue-residue potentials with a favorable contact pair term and an unfavorable high packing density term, for simulation and threading J Mol Biol 1996 256 623 644 8604144
Berrera M Molinari H Fogolari F Amino acid empirical contact energy definitions for fold recognition in the space of contact maps BMC Bioinformatics 2003 4
Rost B Protein secondary structure prediction continues to rise J Struct Biol 2001 134 204 218 11551180
Jaffar J Maher MJ Constraint Logic Programming: A Survey Journal of Logic Programming 1994 19–20 503 581
Marriott K Stuckey PJ Programming with Constraints 1998 The MIT Press, Cambridge, Mass
Swedish Institute for Computer Science SICStus Prolog Home Page
Sterling L Shapiro E The art of Prolog 1997 2 The MIT Press, Cambridge, Mass
Pôle Biolnformatique Lyonnais GOR IV secondary structure prediction method
Simons K Bonneau R Ruczinski I Baker D Ab initio protein structure prediction of CASP III targets using ROSETTA Proteins: Struct Fund Genet 1999 3 171 176
Hoang TX Seno F Banavar JR Cieplak M Maritan A Assembly of protein tertiary structures from secondary structures using optimized potentials Proteins: Struct Funct Genet 2003 52 155 165 12833540
Backofen R The protein structure prediction problem: A constraint optimization approach using a new lower bound Constraints 2001 6 223 255
Dill KA Dominant forces in protein folding Biochemistry 1990 29 7133 7155 2207096
Yue K Dill KA Sequence-structure relationships in proteins and copolymers Physical Review E 1993 48 2267 2278
Raghunathan G Jernigan RL Ideal architecture of residue packing and its observation in protein structures Protein Sci 1997 6 2072 2083 9336831
Clote P Backofen R Computational Molecular Biology: An Introduction 2001 John Wiley & Sons
Dal Palù A Dovier A Fogolari F Apt KR, Fages F, Rossi F, Szeredi P, Vancza J Protein Folding in CLP() with Empirical Contact Energies In Recent Advances in Constraints, of Lecture Notes in Artificial Intelligence 2004 3010
Crescenzi P Goldman D Papadimitrou C Piccolboni A Yannakakis M On the complexity of protein folding In Proc of STOC 1998 597 603
Hart WE Newman A The Computational Complexity of Protein Structure Prediction in Simple Lattice Models In Handbook on Algorithms in Bioinformatics CRC Press
Backofen R Will S A Constraint-Based Approach to Structure Prediction for Simplified Protein Models that Outperforms Other Existing Methods In Proceedings of the 19th International Conference on Logic Programming (ICLP 2003), of Lecture Notes in Computer Science, Springer 2003 2916 49 71
Schuster P Stadler PF Crabbe MJC, Drew M, Konopka A Discrete Models of Biopolymers In Handbook of Computional Chemistry 2001 Marcel Dekker, New York
Dovier A Protein Folding with Constraints-Based Methods
Cantor CR Schimmel PR Biophysical chemistry 1980 W H Freeman and Co
Huang X Powers R Validity of using the radius of gyration as a restraint in NMR protein structure detremination J Am Chem Soc 2001 123 3834 3835 11457122
Fogolari F Esposito G Viglino P Cattarinussi S Modeling of polypeptide chains as C-α chains, C-α chains with C-β, and C-α chains with ellipsoidal lateral chains Biophys J 1996 70 1183 1197 8785277
Rost B Sander C Prediction of Protein Secondary Structure at better than 70% accuracy J Mol Biol 1993 232 584 599 8345525
Qiu D Shenkin P Hollinger F Still W The GB/SA Continuum Model for Solvation. A Fast Analytical Method for the Calculation of Approximate Born Radii J Phys Chem 1997 101 3005 3014
Brooks BR Bruccoleri RE Olafson BD States DJ Swaminathan S Karplus M CHARMM: A Program for Macromolecular Energy Minimization and Dynamics Calculations J Comp Chem 1983 4 187 217
MacKerell ADJ Bashford D Bellott M Dunbrack RLJ Evanseck JD Field MJ Fischer S Gao J Guo H Ha S Joseph-McCarthy D Kuchnir L Kuczera K Lau FTK Mattos C Michnick S Ngo T Nguyen DT Prodhom B Reiher WEI Roux B Schlenkrich M Smith JC Stote R Straub J Watanabe M Wiorkiewicz-Kuczera J Yin D Karplus M All-atom empirical potential for molecular modeling and dynamics Studies of proteins J Phys Chem B 1998 102 3586 3616
European Bioinformatics Institute MaxSprout: Reconstruction of 3D coordinates from C (alpha) trace
| 15571634 | PMC539352 | CC BY | 2021-01-04 16:02:45 | no | BMC Bioinformatics. 2004 Nov 30; 5:186 | utf-8 | BMC Bioinformatics | 2,004 | 10.1186/1471-2105-5-186 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-5-1951558831710.1186/1471-2105-5-195SoftwareGECKO: a complete large-scale gene expression analysis platform Theilhaber Joachim [email protected] Anatoly [email protected] Anish [email protected] Jack [email protected] Dapeng [email protected] Robert [email protected] Michael [email protected] Christoph [email protected] Steven [email protected] Cambridge Genomics Center, Sanofi-Aventis, 26 Landsdowne Street, Cambridge, MA 02139, USA2 Sanofi-Aventis, Genomics and Scientific Computation, Route 202–206, Bridgewater, NJ 08807, USA3 Fast Gun Software, Inc., 180 Myrtle St., Wrentham MA, 02093, USA4 Sanofi-Aventis Tucson Selectide, 1580 E. Hanley Blvd., Tucson, AZ 85737, USA5 Center for Computational Genomics and Bioinformatics, University of Minnesota, 426 Church Street SE, Minneapolis, MN 55455, USA2004 10 12 2004 5 195 195 16 7 2004 10 12 2004 Copyright © 2004 Theilhaber 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
Gecko (Gene Expression: Computation and Knowledge Organization) is a complete, high-capacity centralized gene expression analysis system, developed in response to the needs of a distributed user community.
Results
Based on a client-server architecture, with a centralized repository of typically many tens of thousands of Affymetrix scans, Gecko includes automatic processing pipelines for uploading data from remote sites, a data base, a computational engine implementing ~ 50 different analysis tools, and a client application. Among available analysis tools are clustering methods, principal component analysis, supervised classification including feature selection and cross-validation, multi-factorial ANOVA, statistical contrast calculations, and various post-processing tools for extracting data at given error rates or significance levels. On account of its open architecture, Gecko also allows for the integration of new algorithms. The Gecko framework is very general: non-Affymetrix and non-gene expression data can be analyzed as well. A unique feature of the Gecko architecture is the concept of the Analysis Tree (actually, a directed acyclic graph), in which all successive results in ongoing analyses are saved. This approach has proven invaluable in allowing a large (~ 100 users) and distributed community to share results, and to repeatedly return over a span of years to older and potentially very complex analyses of gene expression data.
Conclusions
The Gecko system is being made publicly available as free software . In totality or in parts, the Gecko framework should prove useful to users and system developers with a broad range of analysis needs.
==== Body
Background
In recent years, in response to the needs of our scientific community we have developed a comprehensive, company-wide gene expression data analysis platform based on a centralized client-server architecture (Figure 1). This platform, named Gecko (Gene Expression: Computation and Knowledge Organization) addresses the problems of analyzing large volumes of continuously generated data (thousands of Affymetrix scans per year), provides a broad spectrum of analysis tools, and creates a single, collaborative view of data for a large, decentralized community of users.
Three organizing concepts have guided the construction of Gecko. The first is the use of the Analysis Tree (actually, a directed acyclic graph) which provides a complete historical and hierarchical display of all analyses conducted to date by the users. In particular, in support of this concept, Gecko permanently stores the results of all analyses performed. A second organizing concept is that of the agglomeration syntax, an "Erector Set" of operations for flexibly creating, combining and subsetting data matrices. The third organizing concept is the pervasive use of experimental designs, which are associated with each data matrix, and which enable the application of a wide range of statistical and pattern recognition tools.
It is the aim of this paper to give an idea of the user's view of Gecko and how one conducts analyses using the system, as well as to provide a software-level overview of the Gecko system architecture. Indeed, we believe that Gecko presents a number of innovative features well-worth presenting, and in connection with this publication, we are making available a public release of the Gecko software[1].
In what follows, we first go "behind the scenes", and present the system architecture in some detail, including overall data organization, database structure, computational engines, statistical tools and models, and finally utility programs. We then present a focused discussion of a specific analysis example, so as to give the reader a more immediate impression of the Gecko system.
Implementation
The Gecko architecture
Gecko is based on a client-server architecture, with a global structure shown in Figure 2. The Gecko users have remote access to the system through a client application, currently designed for the Windows operating system and running on any desktop or laptop computer (a prototype Java-based client has also been developed, but is not yet in production use). Overall, the Gecko client is a "thin" client, focused on handling user requests and server responses, with most of the actual computation and data organizational tasks handled by the Gecko server. As indicated in the figure, the client not only manages interaction with the Gecko server, but also allows for local connection to applications such as Microsoft Excel or the Spotfire visualization tools[2], which can be invoked for additional data analysis after the data has been automatically streamed to these applications from the client.
The exchanges between the client and the server occur through HTTP requests, which transit through a web server running on the server platform. A central aspect of the Gecko client is that it contains an embedded Internet Explorer browser, a feature which greatly simplifies the task of building user interfaces. Thus, forms for submitting parameters to the server are typically built in HTML dynamically generated by server-side Perl CGI or Java servlet programs, and displayed in the embedded browser.
The Gecko server itself runs on a UNIX platform, and consists of the four main components indicated in Figure 2: a database, that predominantly contains non-numerical, organizational data; a set of computational engines, written in C++, Java, or Perl; a set of request-handler programs (Perl CGI and Java servlet programs) that enable the client-server interaction; and a flat file repository, that contain files for both raw numerical expression data (scans) as well as for all derived data types (analyses).
The Gecko database
The set of tables in the Gecko database can be partitioned into three main groups, which we call the "Scan", "Chip" and "Analysis" groups in accordance to their functional roles.
The Scan group of tables stores attributes of the individual scans of microarray data entered into the system. These attributes include a unique scan identifier (the scan name), as well as many parameters (project name, experiment name, sample name, compound(s) applied and treatment duration, hybridization protocols, etc), which record the nature of the biological sample used and how it was processed, and place the scan in a tree with experimental and biological context.
While the Scan group of tables captures many items in common with the so-called MIAME (Minimum Information about a Microarray Experiment) annotation standards[3], it should be emphasized that its design antecedates the creation of the MIAME standards, and is neither as comprehensive, nor fully consistent with these standards. In current installations of Gecko, an independent laboratory information management system (LIMS), upstream of the Gecko analysis platform itself, provides considerably more detailed information about the samples. With our emphasis on Gecko as an analysis platform and not as a LIMS, we have so far deferred the question of how to best federate (under a MIAME-compliant heading) all of the experimental annotation information.
The Scan group of tables also records numerical data in the form of summary statistics for each scan, including several measures of chip brightness and measures of noise and saturation. However, the bulk numerical data for each scan is stored as a file in the flat file repository, with only a file pointer stored in the database.
The Chip group of tables stores the attributes of the Affymetrix chip designs currently known to the system. These include the names of the chip designs used, and for each chip design, all the qualifiers pertaining to it. The tables also store sequence annotation information on a qualifier-by-qualifier basis, including a short description line, as well as a URL that provides a link to more general annotation information for each qualifier. The annotation information is generated externally to the Gecko system, with periodic updates via flat files which can be automatically uploaded.
The Analysis group of tables is central to all the analysis functions available in Gecko. Any analysis object generated in the system has a set of attributes which are saved under five categories of information: 1) its parent/child relations, 2) an internal pointer to the machine-generated file which contains the bulk numerical data, 3) the parameters for the operation which created the analysis object, 4) general parameters (name of the analysis object, data type, number of rows and columns in the data matrix, etc), and 5) experimental design parameters. Knowledge of the experimental design [[4], p.93] [[5], p.214] underlying a dataset is essential to many types of analyses (e.g. ANOVA, contrast calculations [[5], p.214], supervised classification). In Gecko, the experimental design parameters (factors and levels) of each data matrix are thus stored in the database, and can be accessed or modified by the user at any time. They are automatically retrieved and used whenever a relevant analysis is invoked.
Finally we note that no access control is imposed in Gecko: any user can access any collection of scans, or visit any of the existing analyses created by other users. While this very open architecture has greatly fostered collaboration, it is conceivable that access control might eventually be required. To that end, limited architectural and programming modifications are needed. Modifications might consist of expansion of the current user tables, to include group definition and password fields, and addition of straightforward programming logic in both client and server, to mask access to data which is out of the scope of a given user.
Computational engines
Gecko incorporates a spectrum of computational tools, which enter into 5 major categories: 1) agglomeration, 2) statistical analysis, 3) clustering, 4) supervised classification and 5) transformation methods
Agglomeration tools
Data for a given experiment is typically distributed over many scans, numbering in some cases hundreds or even thousands. The ability to easily construct or modify the relevant data matrix, with appropriate normalization of scans with respect to each other, is thus critical to all downstream analysis, and this need is addressed by a suite of tools under the generic heading of agglomeration (Figure 3). For instance, the Gecko Concatenate tool enables assembly of large sets of scans (Figure 3a) through the submission of a simple spreadsheet containing the list of scans in an ordered format. The spreadsheet data entry optionally includes specification of the experimental design (factors and levels), which can also be modified or created de novo at any later time.
Once created, a data matrix then becomes accessible as a single object, to be used in higher-level agglomeration operations. For instance, Cat Ratio enables one to take ratios of two complete data matrices, on a element-by-element basis (Figure 3b). To achieve this, the user needs only to specify the two relevant datasets by selecting the corresponding nodes in the Analysis Tree. All subsequent aspects of the computation (matching numerators and denominators pairwise, actual ratio calculations, on-the-fly normalizations, etc) are achieved automatically, and the resulting data matrix, now containing ratios, is registered in Gecko as a child of the two input datasets.
The suite of agglomeration operations also includes concatenating data matrices to each other, merging replicates within an agglomerated dataset, subsetting on rows or columns, or performing join operations (Figs. 3c,3d,3e), altogether approximating an "Erector Set" for building data matrices out of smaller or larger blocks. These operations are routinely performed on large datasets (currently up to ~ 50000 rows × 500 columns). To indicate processing times for these operations, we note that on a 400 MHz Sun Enterprise server, concatenation requires about 2 second per scan, while the more complex ratio calculations require about 10 seconds per scan pair. Thus concatenating, say, 1000 scans, will require about 30 minutes of processing time, while computing the ratio of 1000 scans to another 1000 scans simultaneously, will require about 3 hours of processing time.
We finally note that users can bypass agglomeration of scan data altogether, and directly upload arbitrary data matrices into the system (see Data sources section below). This feature makes Gecko into a general analysis tool, for multivariate analysis in contexts quite different from that of gene expression.
Statistical analysis tools
The suite of statistical analysis tools includes application of both parametric and non-parametric tests to the agglomerated data matrices, on a qualifier-by-qualifier basis and using the associated experimental designs. Included are two-class comparison tests (Student t-tests, SAM[6], comparison of variances, Mann-Whitney [[5], p.265]), as well as multiple-class and multiple-factors tests (one and two-way ANOVA) and the ability to perform contrast calculations [[5], p. 241] of several different types.
The parametric tests are available with a "renormalization" option which corrects P-values in accordance to an intra-class correlation (icc) model (JT, manuscript to be submitted for publication). For instance, when applied to a one-way ANOVA across several classes, the icc model folds part of the class-dependent effects into the null hypothesis, by mathematically assuming that they have a random component already explained by the null hypothesis, with variance proportional to the variance of the residuals within each class. The proportionality constant is then computed on-the-fly, by requiring that the resulting distribution of the F statistic over all genes is non-significant up to its median value. This renormalization suppresses weak or biologically unremarkable class-dependent effects, while preserving significant data in the upper tail of the observed F distribution. It typically avoids the conundrum of "all genes are significantly regulated" which very often occurs as the number of samples becomes large.
Biased-variance versions of the parametric tests (where an additional, fixed variance term is introduced in the denominators of the t or F statistics so as to reduce noise) are also implemented, in a form where the icc model is combined with semi-parametric resampling to estimate accurate P-values.
Alongside these statistical location tests, which depend on samples being assigned to different classes, one can compute class-independent statistics, such as χ2, grand means or standard deviations on a qualifier-by-qualifier basis across all samples. These tests are frequently useful in ranking expression profiles on the basis of one or several of these test statistics, typically for subsequent filtering-out of noisy profiles, or for overall statistical assessment of the dataset.
Tests incorporating the calculation of the Pearson correlation coefficient are also implemented. These tests enable one to perform "nearest-neighbor" searches for the expression profiles most like those of single or multiple query profiles. As with the set of location tests, these correlation-based tests include options for renormalization, based on the icc model, and for biased-variance terms in the denominators of the equations for correlation coefficients.
As an indication of typical execution times for statistical tests, we note that on a 400 MHz Sun Enterprise server, a two-way ANOVA with associated contrasts, applied to a ~ 22000 rows × 100 columns data matrix, requires about 120 seconds of processing time.
In all cases, tests results are saved in the Gecko Analysis Tree and can be revisited a posteriori by use of the generic Get Stats tool, which internally computes receiver operating characteristics (ROCs) [[7], p. 48], generates graphics for the corresponding ROC plots, and allows for selection of qualifiers based on P-value or on false-discovery rate criteria[8].
Clustering and supervised classification tools
The types of clustering tools implemented in Gecko include self-organized maps (SOM)[9], average linkage hierarchical clustering [[10], p. 318], principal component analysis (PCA) [[11], p. 23], multidimensional scaling (MDS) [[11], p. 107] and the ability to build and display correlation or distance matrices. Supervised classification tools include a gene expression k-nearest-neighbor classifier(GENNC)[12], in conjunction with fully self-consistent feature selection, based on a number of cross-validation methods (leave-one-out, leave-one-group-out, v-fold) [[13], p. 219].
Transformations
Data transformations are frequently required in the course of analyses. Among those available in Gecko are point transformations, where each element of the data matrix is independently transformed (log-transformations, flooring of values to the noise standard deviation, and others), as well as more global transformations, including variance stabilization[14], standardization of rows and/or columns (by mean or median centering followed by division by the corresponding standard deviations) [[11], p. 8], and wholesale transposition of the data matrix. In Gecko, transformations usually appear as explicit steps in the Analysis Tree, rather than being "rolled into" other operations, such as clustering.
Adding new analysis methods
New analysis methods, if already available as executables or applications running from the UNIX command line (for instance, based on C++, Java, R, Matlab, or other languages), can be internally added to the Gecko system by straightforward programming steps. These steps include i) providing for a user interface, generated by server Perl CGI or Java servlet programs, and displayed as HTML in the client Browser window; and ii) constructing a server-based driver program, that will execute the UNIX command, using the parameters communicated by the user interface. We note that while an application programming interface (API) has not been formalized, a Gecko API is already well-approximated, by the existence of a modular set of methods for accessing the database, and for reading and writing to numerical flat files.
For external analysis using other applications, direct streaming of all internal Gecko types is currently implemented for Spotfire[2] and Microsoft Excel. For saving data to local disk, generic data export in tab-separated values format is also possible. Furthermore, specially formatted types of data export to disk have also been implemented, in particular for the Cluster and TreeView[15] clustering and visualization programs. Extending the number of specially formatted export options to other analysis packages (for instance, to create R "data frames" to be used in BioConductor R packages[16]), should be a straightforward programming task, consisting of adding an appropriate formatting function to the existing Perl/CGI module.
Data organization in Gecko: the Analysis Tree
A central concern in the design of Gecko was to enable the user to perform and especially to later recall complex analysis work flows (such as the cell line data analysis, described in detail below). In general, graphs of analyses conducted in Gecko, with nodes corresponding to datasets and edges to operations on these datasets, result in directed acyclic graphs (DAGs). A DAG is unlike a tree, in that each of its nodes can have multiple parents, whereas in a tree each node has a unique parent; for simplicity however, we refer to the DAG generated by Gecko as the Analysis "Tree". Furthermore, in the Gecko client the DAG is actually displayed as a tree: the DAG topology is correctly maintained by replicating, for nodes with multiple parents, the corresponding subgraphs under each of the parent nodes.
Once generated, the data file corresponding to a node in the Gecko Analysis Tree is permanently stored (unless the node is explicitly deleted by the user at some later time). This approach enables users to return at any time to potentially very large and complex panels of analysis results, without requiring them to regenerate all final and intermediate results on-the-fly, as might be required in an alternative real-time "dataflow" approach (in which only the sequence of operations is permanently stored, and in which data is recomputed every time a new session is started). We have found that the dataflow approach can entail a prohibitive computational cost and waiting time, whenever a large number of analyses are being simultaneously considered, as in the examples of Figure 4 (described in detail below). This situation is obviously exacerbated by the presence of individual lengthy computations, such as are required for instance for classifier cross-validation.
The permanent storage of all analysis results might seem an extravagant use of computer resources, but experience shows that it results in reasonable use of server memory over time. For an expression analysis community of roughly 100 scientific users, over a span of 5 years memory use has been limited to about 150 GB (corresponding to the disk space available on a couple of current generation personal computers), reached with slow linear growth over time. Furthermore, should it be absolutely required, implementing a file archival and retrieval system for the oldest analyses would be a straightforward task.
Noise model
The Gecko noise model is based on the so-called PFOLD joint noise model and ratio estimation algorithm[17]. This model includes both additive (background, cross-hybridization) and multiplicative (coefficient-of-variation effects) noise terms, within a Bayesian estimation framework. Expression ratios and related P-values and confidence limits are computed on the basis of a posterior distribution of ratios conditional on measured intensities and noise terms. The rigorous mathematical derivation of the posterior distribution results in a formulation that seamlessly connects high and low signal-to-noise regimes, and allows estimation of ratios even when recorded intensities are zero or negative.
Data sources
While currently all scans uploaded into Gecko are generated by Affymetrix technology, in the past the system has also been used with other types of expression data, for instance generated by two-color hybridizations on spotted arrays. This has been possible at low programming cost, because the internal representation of scan data in Gecko is independent of microarray technology, with a generalized storage of intensity and noise information for each chip qualifier or microarray spot. Programming modifications needed for a new technology thus primarily occur in the design of the new raw-file parser (automatically invoked on entry by the scan processing pipeline). Note that for the two-color technologies mentioned above, data for each channel is entered as a separate intensity scan. Channel-to-channel ratios between matched scans are then computed downstream by the users, using the Cat Ratio Agglomeration tool mentioned above.
An alternative and very flexible method for data entry into Gecko, which entirely bypasses scan entry, is to directly upload a tabular file through the Gecko client. In particular, this method enables one to upload gene expression data from the many public sources where it is provided only in spreadsheet format. Furthermore, as already stated, it also enables one to use the Gecko analysis tools in contexts unrelated to gene expression.
Utilities: the Gecko scan processing pipeline
As Gecko was designed as a centralized resource, but also for service of geographically remote sites (Figure 1), it was critical that the Affymetrix scan submission process be made as automatic and foolproof as possible. To that end, a two-step procedure was devised, described as follows.
First, users register scans through an interface provided in the Gecko client, using an appropriate submission window. This registration step stores the scan attributes in the Gecko database (project name, experiment name, sample name, and so forth), but does not transfer the scan numerical data (intensity values) itself. In the second, independent step, the users send the scan numerical data, in the form of Affymetrix CEL files[18], to a specific incoming directory on the Gecko server, typically using the file transfer protocol (FTP) utility (Figure 1).
The Gecko scan processing pipeline, run as a periodic "cron" job on the UNIX platform, automatically converts the Affymetrix CEL file data to Affymetrix MAS5[18] estimated values, using an emulator of the corresponding algorithm, and writes the results in a format specific to the Gecko system, finally setting a "processing pending" flag to off for each processed scan. Error statuses for files which exceptionally fail processing are written into the database and displayed in a client-based processing queue administration window. On a 400 MHz Sun Enterprise server, the processing time per scan is approximately 3 minutes, enabling upload of about 500 scans per 24 hour period.
The processing pipeline has proven to be very robust, and can be readily modified to accept other sources of gene expression data, as already mentioned above. Thus, it has also been used to process cDNA microarray data[19] in the past.
Results
An analysis example
As an example of an analysis workflow conducted in Gecko, we describe a study of a cancer cell line treated with a panel of compounds which are inhibitors of cell proliferation. Cultures of the A498 cell line (a cell line derived from kidney carcinoma and part of the NCI60 panel[20]) were treated with five different dimethyl sulfoxide(DMSO)-dissolved compounds (here named A1, A2, A3, B1 and B2) falling into two distinct classes (DNA replication inhibition or tubulin binding, A and B, respectively) depending on their mechanism of action. Control cell cultures, treated with the DMSO solvent alone, were also generated. Six biological replicates of the cell cultures were generated for each combination of compound and harvest time, with harvests occurring at 6 hours or 24 hours after the start of treatment. After processing of the cell extracts, the resulting cRNA samples were hybridized to HG_U133A Affymetrix chips[18,21], resulting in a total of 72 chip scans, which were submitted to the Gecko scan processing pipeline, and uploaded into the system.
The Gecko client user interface
Figure 4 shows the Gecko client as seen by the user. The client user interface consists of a list of menu items (top), with an associated list of icons (shortcuts to menu items, immediately below), under which are three large adjustable window panes, with content as follows.
The left-hand window pane (Tree window) provides a tree representation of the data objects existing in Gecko; in the figure, it currently displays the Analysis Tree, which provides a full and permanent record of analysis operations and resulting datasets executed so far. This window can also display the Scan Tree, a hierarchical display of all scans in the system, by selection of the corresponding Scan Tree tab (upper left-hand corner).
The right-hand window pane of the client (the Browser window) contains forms for submitting parameters to the analysis tools, and also displays analysis results. Currently selected is an input form for performing supervised classification of the compound-treated samples, using a k-nearest neighbor classifier[12].
The bottom window pane of the client (the Properties window) displays the properties of the object currently selected in the Analysis Tree. Here, the experimental design for the selected object, the data matrix compound-panel.AGG, is currently visible.
The Analysis Tree contains nodes at three types of levels. Nodes at the highest, most general level are named Projects: in Figure 4, the Analysis Tree is opened under the Project Oncology_compound_response. Nodes at the next, lower level, named Analyses, enable classification under more specific themes: in Figure 4, the Analysis Tree is opened under the Analysis A498-series, which contains results specific to the A498 cell line assays. The nodes at all levels below Projects and Analyses contain the actual results of analysis operations, and are arranged in a recursive, parent-child hierarchy of arbitrary depth. Thus in Figure 4, under A498-series, five generations of results are displayed. Note that each analysis result has a specific data type, indicated by the extension of its name and by a color-coded icon. A total of 33 data types are currently defined in Gecko.
The analysis workflow for the A498 cell line data
The analysis workflow of the A498 cell line data is indicated in an expanded view of the Analysis Tree (Figure 5). The analysis was started by creating a single data matrix out of the 72 independent scans which together constitute all data for the A498 series. The data matrix was created by a copy-and-paste submission of a spreadsheet containing the list of scans to the Concatenate tool, which then automatically assembled and normalized the relevant scan data. This operation resulted in two objects, a scan reference file, compound_panel.GPPL (grey square icon), containing the constitutive list of scans, and the data matrix itself, compound_panel.AGG (orange square icon). Note that these two objects were automatically inserted below the analysis node A498-series, with compound_panel.AGG inserted as a child of compound_panel.GPPL.
It is important to emphasize that the data matrix compound-panel.AGG is physically stored on the server platform. This centrality insures that all users have simultaneous access to ongoing analyses, and if desired, that they can collaborate in real-time, even when working from very different geographical locations (Figure 1).
The data matrix compound_panel.AGG has dimensions 22283 rows × 72 columns, with each row corresponding to a different Affymetrix qualifier on the HG_U133 chip (here the term "qualifier" is synonymous with Affymetrix "probe set"), and each column to a specific experimental sample. The associated experimental design [[4], p. 93] [[5], p. 219] of the A498 series, is also saved in the Gecko database in association with compound_panel.AGG, and is displayed in the Properties window (bottom window in Figure 4). The experimental design was originally specified in the spreadsheet submitted to the Concatenate tool, but can also be modified (or newly created) at any later time. It contains four factors, labeled dose, time_hr, compound and moa, corresponding to compound doseage, harvest time, compound name and compound mechanism of action, respectively.
Based on a general experimental design, one can then automatically define in Gecko simpler two-factorial designs, by selection of the factors in the appropriate client interface. For instance, Figs. 6a and 6b display the two-factorial designs for compound_panel.AGG which result from the combinations (compound × time_hr) and (moa × time_hr), respectively. The number of replicates for every combination of levels is indicated in each cell of the tables. The factorial design (compound × time_hr) is of particular interest for finding genes with expression differentially regulated by the treatments with the different compounds, with or without concommittant time variation. In the A498 analysis workflow, this design was used to generate a two-way analysis of variance (ANOVA) [[5], p. 214] of compound_panel.AGG, resulting in the dataset compound-panel_compound_time_hr.ANOVA2 (Figure 5, purple triangle icon), which was again automatically inserted as a child of its parent dataset. The two-way ANOVA is conducted on a qualifier-by-qualifier basis, and results in a file contains 22283 rows, each row consisting of the P-values (and associated statistics) for the compound, time_hr and compound × time_hr effects for the corresponding qualifier.
Once created, compound-panel_compound_time_hr.ANOVA2 can be revisited for selection of statistically significant data using a generic utility called Get Stats. In particular, Get Stats internally computes receiver operating characteristics [[7], p. 48] for all the effects considered in the factorial design, and permits selection of significant qualifiers at a specified false-discovery rate (FDR)[8]. For instance, for a threshold FDR ≤ 0.05 used in conjunction with the compound effects, one finds that 517 qualifiers out of 22283 exhibit compound-related changes in expression. In the Analysis Tree, the data subset corresponding to these 517 qualifiers, compound-panel_compound_time_hr-517.ANOVA2, is automatically inserted as a child of the parent file. The operation parameters (effect used for selection and threshold FDR) which generated the subset are also saved, and are displayed in the Properties window for reference.
Following the ANOVA operations, the original data matrix, compound_panel.AGG, was then filtered to the rows corresponding to the 517 signifi-cant qualifiers contained in compound-panel_compound_time_hr-517.ANOVA2, in preparation for down-stream clustering and supervised classification operations. This step, implemented by the subsetting tool Reduce on Qlist, results in the filtered data matrix compound_panel-517.AGG.
Note that for all of the datasets discussed above, prior to each operation, a tentative output name was automatically created (typically by a concatenation of the input dataset name and of the name of the operation to be applied), and then presented to the user in a preview page. The tentative name can then be modified, if desired, before final submission.
Clustering and supervised classification of the A498 cell line data
Several additional analysis steps were performed on the A498 series data, illustrating the use of complementary unsupervised (clustering) methods, as well as a supervised classification approach. Starting from the filtered data matrix compound_panel-517.AGG (Figure 5), and after row standardization [[11], p. 8] (compound_panel-517RmedNR.dat), three clustering methods were first applied, resulting in i) a self-organized map[9] of the data with 1 × 64 cluster geometry (compound_panel-517_1 × 64.SOM), ii) a hierarchical clustering using average linkage [[10], p. 318] (compound_panel-517.TREE), and iii), a principal component analysis (PCA) [[11], p. 23] (compound_panel-517.PCA).
Supervised classification of the samples was also performed, using the gene expression k-nearest-neighbor classifier[12] integrated into Gecko. The classification was done on the basis of mechanism of action of the compounds (excluding controls, and regrouping 6 hour and 24 hour samples), resulting in a two-class problem with class labels "DNA replication inhibition" and "tubulin binding". The Feature Selection tool was first used, to compute the misclassification error as a function of the number of features (qualifiers) retained in the dataset, using the Fisher interclass separation [[13], p. 135] as a feature selection criterion and with misclassification error computed using "leave-one-group-out" (LOGO) cross-validation [[13], p. 219]. In each step of the LOGO procedure, all instances corresponding to a given compound are simultaneously removed and cross-classified by the remaining instances in the training set. Applied to each compound in turn, this resulted in 5 separate cross-classifications, each applied to the 12 held-out samples, with a tally of all misclassifications errors applied at the very end. The results were saved in compound_panel-517RmedNR_feature_sel_SCAN.STAT. An explicit k-nearest-neighbor classification, using an optimal set of 60 qualifiers determined by the feature selection step was then performed. The final classification results, including an internally generated PCA representation of the data, were automatically saved in the data set compound_panel-517RmedNR_feature_sel_FILTER_60_moa.CVEC (pink circle icon with × pattern).
Visualization of analysis results
Gecko provides for flexible visualization of analysis results, with results either directly displayed in the client Browser window, or streamed to external visualization tools such as Spotfire[2]. In Figure 7, the receiver operating characteristic for the distribution of P-values according to compound effects in the two-way ANOVA (compound-panel_compound_time_hr.ANOVA2) is displayed in the Browser window. In Figure 8, after streaming to Spotfire, a PCA representation of data for the supervised classification compound_panel-517RmedNR_feature_sel_FILTER_60_moa.CVEC is displayed as a three-dimensional scatter plot.
Conclusions
Constructed around the three organizing concepts of the Analysis Tree, the agglomeration syntax, and the pervasive use of experimental designs, Gecko has proven to be a robust analysis platform for a large and distributed scientific community. Gecko has allowed for flexible incorporation of new analysis methods over time, and has insured intelligible access to older, complex analyses, successfully answering the question of "where is my data?".
It should be emphasized that the analysis framework afforded by Gecko is general and not limited to gene expression data. Data can be uploaded from many other sources, and the analysis methods relevant to the new data types can also be incorporated as needed. Thus, methodologies for the analysis of protein-protein interaction data[22], or for the analysis of Gene Ontology, categorical data[23] have been integrated into Gecko in the past. It is now hoped that with its public release, many other uses will be found for this general analysis platform.
Availability and requirements
All components of the Gecko software, including source code, are being made available as a package under SourceForge.net[1]. The Gecko project's home page will provide information regarding release schedules and availability. Interested parties may also directly contact the corresponding author (JT) for information.
Installation of the complete platform will require manual intervention as well as execution of the automated builds provided in the package. Manual intervention is required for installation of the required external software libraries (Perl modules, GNU software, graphics software, etc) as well as for setting up the run-time Gecko infrastructure (web server, servlet engine, Oracle data base). The automated builds provide for compilation of the C++ and Java source code, for creation of required flat-file directories, and for the creation of the database tables.
Existing installations of the Gecko platform are on Sun Enterprise UNIX servers running SunOS 2.8. Transposition to other operating systems, such as Linux, will thus require some additional "tuning" of components during installation. It should also be noted that the Gecko numerical analysis programs can be used in a standalone fashion (i.e. by execution from the command line), without requiring a complete installation of the platform.
Authors' contributions
JT designed and implemented the early browser-based version of Gecko; he then focused on algorithm and tool development and implementation, alongside giving overall scientific direction for the design of the production version. AU designed and implemented the database. AM designed and built the Java servlets which process user requests, as well as implement many analysis tools and utilities. JC designed and built the client application. DX designed and implemented several statistical analysis algorithms. RN and MH have been involved in running, packaging and in creating automated builds of the Gecko platform. CB had a central role in promoting the Gecko platform and in scientific design input. SB initiated and oversaw development of the production version of Gecko, and was closely involved with the design of the earlier browser-based versions.
Acknowledgments
The authors wish to thank Paul Giresi for scientific input and user feedback during development, Richard Goldman and Igor Krigman for their extensive programming contributions, Rainer Fuchs and Michael Rosenberg for support and scientific input during development, and Amanda Jackson, Teresa Garcia and Sergio Roman-Roman for scientific input and extensive user feedback in early design phases.
Figures and Tables
Figure 1 Sketch of the functional organization of a Gecko installation, emphasizing its distributed aspects. Research groups in France, Germany and the United States (sites #1 through #3) submit Affymetrix scan data (3 dark lines) to a server based in Cambridge, United States. Users can conduct analyses on any part of the stored data, using a client application (small Gecko icons) which enables two-way communication with the server (two-way arrows).
Figure 2 Overview of the Gecko software architecture, including the major components of the client and server. Note that the client allows for data streaming to other applications, such as Excel or Spotfire.
Figure 3 Examples of the Gecko agglomeration syntax. a) Concatenate: concatenation of a large number of scans (here arising from many samples, each profiled across three distinct chip designs) into a single data matrix; normalizations are computed on-the-fly; b) Cat Ratio: element-by-element ratios are computed for two data matrices, creating a new data matrix containing ratios and P-values; c) Combine on Columns: the columns of two data matrices are concatenated to create a larger data matrix; d) Reduce on Qlist and Reduce on Columns: a data matrix is subsetted on its rows (right-hand arrow) or on its columns (bottom arrow), respectively, to form reduced data matrices; e) Merge Replicates: replicates are merged by taking medians of intensities (with concomitant reestimation of noise terms); f) Join: two data matrices are joined using qualifiers (row indices) as the join key.
Figure 4 The Gecko user interface, showing Analysis Tree (upper left-hand-side), Browser (upper right-hand-side) and Properties windows (bottom). The Analysis Tree is opened on the A498 cell line data analyses described in the text (top), as well as on another independent study (bottom). An input form for a k-nearest-neighbor classifier is displayed in the Browser window. The experimental design for compound_panel.AGG (the selected object) is displayed in the Properties window.
Figure 5 Details of the analysis workflow for the A498 cell line data (this is an expanded view of the Analysis Tree shown in Figure 4).
Figure 6 The two-factorial designs for compound_panel.AGG, resulting from the factor combinations a) (compound × time_hr) or b) (moa × time_hr). The number of replicates for every combination of levels is indicated in each cell of the tables. The pink cells in a) are flagged as reference cells for contrast calculations with all other cells in the corresponding column.
Figure 7 Receiver operating characteristic for the selection of qualifiers according to compound effects in the two-way ANOVA data set compound-panel_compound_time_hr.ANOVA2, with plot generated in the Gecko Browser window. The number Nf (P0) of qualifiers found at a P-value less than or equal to P0 is plotted against P0 (note the log/-log scale used in the plot). The red line shows the trend expected under the null hypothesis of no compound effects (credits: pgplot graphics package).
Figure 8 Spotfire scatter plot for the three-dimensional principal component analysis of the training set used for supervised classification, in the data set compound_panel-517RmedNR_feature_sel_FILTER_60_moa.CVEC, generated after data streaming to Spotfire. Samples are color coded in accordance to mechanism of action of the compound treatment, as indicated in the plot.
==== Refs
Gecko open source release 2004
Spotfire, USA 212 Elm Street Somerville, Massachusetts, USA
Minimum information about a microarray experiment – MIAME standards MIAME 11 document April 1, 2002
Fisher RA The Design of Experiments 1971 New York: Hafner
Keeping ES Intoduction to Statistical Inference 1995 New York: Dover
Tusher VG Tibshirani R Chu G Significance analysis of microarrays applied to the ionizing radiation response Proc Natl Acad Sci USA 2001 98 5116 5121 11309499 10.1073/pnas.091062498
Duda RO Hart PE Stork DG Pattern Classification 2001 New York: John Wiley
Reiner A Yekutieli D Benjamini Y Identifying differentially expressed genes using false discovery rate controlling procedures Bioinformatics 2003 10 1093
Kohonen T Self-Organizing Maps 1997 New York: Springer
Ripley BD Pattern Recognition and Neural Networks 1996 Cambridge (UK): Cambridge University Press
Dillon WR Goldstein M Multivariate Analysis 1984 New York: John Wiley
Theilhaber J Connolly T Roman-Roman S Bushnell S Jackson A Call K Garcia T Baron R Finding Genes in the C2C12 Osteogenic Pathway by k-Nearest-Neighbor Classification of Expression Data Genome Research 2002 12 165 176 11779842 10.1101/gr.182601
Fukunaga K Introduction to Statistical Pattern Recognition 1990 2 New York: Academic Press
Durbin BP Hardin JS Hawkins DM Rocke DM A variance-stabilizing transformation for gene-expression microarray data In Proc 10th Int Conf on Intelligent Systems for Molecular Biology(ISMB 2002) Bioinformatics 2002 18 S105 S110
Eisen MB Spellman PT Brown PO Botstein D Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci USA 1998 95 14863 14868 9843981 10.1073/pnas.95.25.14863
Bioconductor: open source software for bioinformatics
Theilhaber J Bushnell S Jackson A Fuchs R Bayesian estimation of fold-changes in the analysis of gene expression: the PFOLD algorithm J Comput Biol 2001 8 585 614 11747614 10.1089/106652701753307502
Affymetrix Santa Clara, California, USA See for an overview of Affymetrix DNA chip technology
Schena M Shalon D Davis RW Brown PO Quantitative monitoring of gene expression patterns with a complementary DNA microarray Science 1995 270 467 470 7569999
NCI60 cell line panel Developmental Therapeutics Programs, NCI/NIH 2004
Lockhart DJ Dong H Byrne MC Follettie MT Gallo MV Chee MS Mittmann M Wang C Kobayashi M Horton H Brown EL Expression monitoring by hybridization to high-density oligonucleotide arrays Nat Biotechnol 1996 14 1675 1680 9634850 10.1038/nbt1296-1675
Schwikowski B Uetz P Fields S A network of protein-protein interactions in yeast Nat Biotechnol 2000 18 1257 61 11101803 10.1038/82360
Hosack DA Dennis G JrSherman BT Lane HC Lempicki RA Identifying biological themes within lists of genes with EASE Genome Biol 2003 4 R70 14519205 10.1186/gb-2003-4-9-r60
| 15588317 | PMC539353 | CC BY | 2021-01-04 16:02:46 | no | BMC Bioinformatics. 2004 Dec 10; 5:195 | utf-8 | BMC Bioinformatics | 2,004 | 10.1186/1471-2105-5-195 | oa_comm |
==== Front
BMC NeurolBMC Neurology1471-2377BioMed Central London 1471-2377-4-231558830910.1186/1471-2377-4-23Research ArticleMorphological brain differences between adult stutterers and non-stutterers Jäncke Lutz [email protected]änggi Jürgen [email protected] Helmuth [email protected] Institute of Psychology, Department Neuropsychology, University Zurich, Switzerland2 Department of Neurology, Johann-Wolfgang Goethe University Frankfurt am Main, Germany2004 10 12 2004 4 23 23 13 8 2004 10 12 2004 Copyright © 2004 Jäncke et al; licensee BioMed Central Ltd.2004Jäncke 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 neurophysiological and neuroanatomical foundations of persistent developmental stuttering (PDS) are still a matter of dispute. A main argument is that stutterers show atypical anatomical asymmetries of speech-relevant brain areas, which possibly affect speech fluency. The major aim of this study was to determine whether adults with PDS have anomalous anatomy in cortical speech-language areas.
Methods
Adults with PDS (n = 10) and controls (n = 10) matched for age, sex, hand preference, and education were studied using high-resolution MRI scans. Using a new variant of the voxel-based morphometry technique (augmented VBM) the brains of stutterers and non-stutterers were compared with respect to white matter (WM) and grey matter (GM) differences.
Results
We found increased WM volumes in a right-hemispheric network comprising the superior temporal gyrus (including the planum temporale), the inferior frontal gyrus (including the pars triangularis), the precentral gyrus in the vicinity of the face and mouth representation, and the anterior middle frontal gyrus. In addition, we detected a leftward WM asymmetry in the auditory cortex in non-stutterers, while stutterers showed symmetric WM volumes.
Conclusions
These results provide strong evidence that adults with PDS have anomalous anatomy not only in perisylvian speech and language areas but also in prefrontal and sensorimotor areas. Whether this atypical asymmetry of WM is the cause or the consequence of stuttering is still an unanswered question.
==== Body
Background
Persistent developmental stuttering (PDS) is a relatively severe disturbance characterized by involuntary, audible or silent, repetitions or prolongations of sounds or syllables. These are not readily controllable and often are accompanied by other movements and by negative emotions [1,2]. Developmental stuttering evolves before puberty without apparent brain damage or other known cause. Several authors suppose a hereditary component of PDS because of the relatively high concordance rate in family members of PDS subjects (70% for monozygotic twins, about 30% for dizygotic twins, and 18% for siblings of the same sex) [3-5]. Because of this hereditary component and the early onset of stuttering it has repeatedly been suggested that some kind of anatomical or neurophysiological predetermination increases the vulnerability for stuttering [for a summary of theories and findings related to stuttering research [6]].
Several experimental studies have shown that stutterers reveal prolonged manual and vocal reaction times to simple and complex verbal and nonverbal stimuli [7-10], reduced bimanual coordination measures [11-13], atypical functional lateralizations [14-18], abnormalities in the auditory system [19-21], or increased variability of time-critical speech parameters [9,22,23]. More recent neuroimaging studies have shown atypical hemodynamic responses in speech-related brain areas even during fluent utterances suggesting a dysfunctionally operating speech control circuit in stutterers [24-29]. However, although much research has been invested to understand the neurophysiological mechanisms and underpinnings of this disorder, none of the aforementioned studies provide a substantial breakthrough in understanding stuttering. Several researchers have hypothesized subtle but nevertheless crucial deficiencies in the anatomical and neurophysiological underpinnings of the speech and language system. One popular hypothesis is that stutterers would show an atypical lateralisation of the speech system (reversed or reduced laterality) thought to make the system more vulnerable to speech dysfluencies [17]. Although recent neuroimaging studies have shown atypical activation and deactivation of brain regions in adults with PDS [24,30-32] the anatomical underpinnings of stuttering have not been examined in detail so far.
According to the present literature four anatomical studies revealed brain abnormalities in stutterers compared to controls. The earliest study examined two left-handed stuttering siblings using CT and revealed an atypical (reduced) anatomical asymmetry of the occipital poles [33]. The first high-resolution MRI study investigating stutterers revealed a reduced volumetric asymmetry of the planum temporale (a brain area which is involved in higher order auditory processing) and other anatomical peculiarities in speech-related areas [34]. A more recent paper of the same group revealed that PDS is also associated with atypical (mostly reduced) prefrontal and occipital lobe asymmetries [35]. In addition, deficits in language processing were associated with some anatomic measures in the adults who stutter. Using a new MRI technique (diffusion tensor imaging: DTI), that allows the assessment of white matter ultrastructure, Sommer et al. [36] found an area of decreased white matter tract coherence in the left Rolandic operculum. This structure is adjacent to the primary motor representation of tongue, larynx, and pharynx and the inferior arcuate fascicle linking temporal and frontal language areas, which both form a temporo-frontal language system involved in word perception and production. Thus, there are indeed first strong hints that the brain of stutterers differ from non-stuttering subjects on a macroanatomical level suggesting that morphological predispositions determine stuttering.
Although the aforementioned anatomical studies have focussed on the anatomical foundations of stuttering, several questions are not answered yet. Therefore we re-examined the hypothesis of anatomical differences between stutterers and non-stutterers using voxel-based morphometry (VBM). This approach circumvents the problem of analyzing predetermined regions of interests by analyzing stereotactically normalized brains on a voxel-by-voxels basis with respect to differences in the volume of white matter (WM) or grey matter (GM) [37,38]. This approach has successfully been used in the last 8 years for several clinical populations and allows studying the morphology separately for GM and WM looking at the entire brain [39-41]. A further advantage of this method is the objectivity and thus rater independence. We hypothesized that beside the previously reported atypical anatomical asymmetries in perisylvian and frontal areas there should be additional differences in further brain areas also involved in speech motor control. Because several studies report that the auditory system in stutterers is dysfunctional [19,21,42-47] especially during speaking (thus, emphasizing the role of auditory feedback in the context of stuttering), we anticipated structural peculiarities in the auditory cortex (Heschl's gyrus and the planum temporale) in this group. In addition, we also anticipated anatomical peculiarities in frontal brain areas and in the somatosensory and motor system controlling the speech muscles.
Methods
Subjects
The sample included adults with PDS (n = 10) and controls (n = 10) matched according to sex, age and education. All subjects were consistent right-handers (CRH) according to the Annett handedness questionnaire (AHQ) [66]. Our sample contained the approximate sex distribution as those reported in population studies of adults who stutter; thus, there were more men (n = 8) than women (n = 2) in this sample. The dysfluent sample was limited to adults with PDS who had been diagnosed with developmental stuttering before the age of 8 years and had undergone treatment at some point, but continued to be dysfluent. None of the subjects was taking centrally acting medications that could have resulted in PDS, and all met the clinical criteria of developmental stuttering – not acquired stuttering. Of the adults who stuttered, 50% had a family history of stuttering; none of the controls had a family history of stuttering. Stuttering severity was determined using the Stuttering Severity Inventory (SSI) [67] with individuals in the sample ranging from mild (2), moderate (7) to severe (1). All participants were native German speakers with no reported history of dyslexia, specific language impairment, attention deficit disorder, traumatic brain injury, substance abuse, or other neuropsychiatric conditions. All participants gave informed consent before participating.
MRI scanning protocol and data analysis
We used a Siemens 1.5 T magnet and a 22-min fast-low-angle-shot MR sequence yielding 128 contiguous sagittal slices with 1 × 1 × 1.17 mm image voxel size [68–70]. Data were analyzed on a PC workstation using MATLAB 5.3 (MathWorks, Natick, MA) and SPM 99 (Wellcome Dept. Cogn. Neurol, London; ) [71]. Preprocessing were guided by the VBM method proposed by Godd et al. [40,41]. In short, the following steps were conducted: (1) Spatial normalization of each brain to the MNI space using the MNI template; (2) spatial smoothing with an 8-mm full-width at half-maximum (FWHM) isotropic Gaussian kernel; (3) creating of a mean anatomical image from these normalized and smoothed scans; (4) stereotactic normalisation of all MRI scans (in native space) using the newly developed template and non-linear smooth spatial basis functions; (5) these spatially normalized images were resliced with a final voxel size of 2 × 2 × 2 mm3. The normalized scans were then segmented into grey (GM) and white matter (WM), cerebro spinal fluid (CSF), and other non-brain partitions applying the algorithmus implemented in SPM99 based on the algorithms developed by Ashburner and Friston [37,38]. In order to sensitize our subsequent statistical analysis not only to differences in the GM (WM) proportions but also to differences in the true GM (WM) volumes a further processing step – known as the 'Jacobian Modulation – was incorporated. The partitioned images (GM and WM) were multiplied by the Jacobian determinants of the deformation field transforming the GM and WM density values into volume equivalents [38,40]. The normalized, segmented (and modulated) images are smoothed using a 10-mm FWHM isotropic Gaussian kernel to improve statistical quality of the data (e.g., normal distribution).
Statistical analysis of VBM data
The normalized, smoothed, segmented (and modulated) data were analyzed using statistical parametric mapping (SPM99) employing the framework of the General Linear Model. Regionally specific differences in GM (and WM) (both for the density and the volume equivalents) between groups were assessed statistically using a two-tailed contrast. Corrections for the search volume (and implicit multiple comparisons) in terms of the P values were made using Gaussian random field theory, which accommodates spatial correlations inherent in the data and is now established as the conventional approach to inference in smooth spatially extended data. We restricted the search volume to the GM or WM volume enabling us to increase the statistical power of statistical testing. Significance levels for two-sided T statistics were set at T = 5 (corrected for multiple comparisons across the WM or GM volumes) and a spatial extend criterion of k = 50. The spatial extend of k = 50 was introduced because this volume size roughly corresponds to the size of a meaningful anatomical area (0.4 cm3).
Results
Because there was no substantial difference between the results of our statistical tests for the density and volume equivalents, we only report the findings based on the analysis of the volume equivalents. We found increased WM volumes in stutterers within four clusters on the right hemisphere. The clusters are located in the superior temporal gyrus (STG) including the planum temporale, the precentral gyrus (PrCG), the inferior frontal gyrus (IFG) comprising the pars opercularis (POP), and the middle frontal gyrus (MFD) (Table 1). There was no significant difference between stutterers and non-stutterers with respect to the GM volumes.
Table 1 Regions of increased WM volumes in stutterers. Indicated are the peak differences (in t-values), their stereotactically coordinates, and the associated anatomical labels derived from the MNI standard brain. Please note, there were no areas with increased WM volumes in controls compared to stutterers.
Anatomic region Coordinates (X, Y, Z) t-Value
R Superior temporal gyrus (STG) 64 -34 21 7.45
R Inferior frontal gyrus (IFG) 66 8 21 6.58
R Middle frontal gyrus (MFG) 44 48 11 6.23
R Precentral gyrus (PrCG) 30 -28 63 7.25
R Precentral gyrus (PrCG) 62 -12 37 6.53
In order to understand the differences between stutterers and non-stutterers with respect to the WM volumes in these anatomical areas more precisely, we placed regions of interest (ROI) in these anatomical areas and the homotopic areas on the left hemisphere. For the auditory cortex we used a rectangular ROI including Heschl's gyrus (HG) and the planum temporale (PT) (size of the ROI on both hemispheres: 8.4 cm3). The placement pf these ROIs were guided by anatomical landmarks and published probability atlases of the HG and PT [48,49]. The other ROIs were defined according to the stereotactic coordinates found in the VBM analysis. For these ROIs, rectangular volumes (10 mm edge length resulting in a volume of 10 × 10 × 10 mm) were used. The mean WM measures were calculated for each ROI and subjected two-way ANOVAs with one repeated measurement factor (Hemisphere: left vs. right) and one grouping factor (Group: stutterers vs. non-stutterers). Because we found significant interaction effects for all ROIs we will only interpret these interactions. For the auditory cortex we found a strong main effect for the factor Hemisphere (F(1, 18) = 29.2, p <= 0.001, ETA2 = 0.62) and a significant interaction between both factors (F(1, 18) = 31.6, p <= 0.001, ETA2 = 0.64). Subsequent Scheffé contrasts and Figure 2 show that there is a strong between-hemisphere difference for non-stutterers (larger WM volume on the left hemisphere, p < 0.01) but not for stutterers (p > 0.4). For the IFG there were strong main effects (Hemisphere: F(1, 18) = 9.2, p = 0.007, ETA2 = 0.34; Group: F(1, 18) = 23.6, p <= 0.001, ETA2 = 0.58) and a significant interaction (F(1, 18) = 23.9, p <= 0.001, ETA2 = 0.57). The strong interaction is qualified by a between-hemisphere difference found for stutterers (with larger WM volumes on the right compared to the left IFG) while there is no between-hemisphere difference in non-stutterers. For the PrCG we found a significant between-group difference (F(1, 18) = 11.1, p = 0.004, ETA2 = 0.38) and a significant interaction (F(1, 18) = 31.0, p <= 0.001, ETA2 = 0.64). The pattern of this interaction resembles the interaction found for the IFG with larger WM volumes on the right hemisphere for stutterers than on the left while non-stutterers show similar values for both hemispheres. For the MFG all two main effects as well as the interaction were strongly significant (Hemisphere: F(1, 18) = 23.5, p <= 0.001, ETA2 = 0.56; Group: (F(1, 18) = 13.2, p = 0.002, ETA2 = 0.42; interaction: (F(1, 18) = 18.4, p <= 0.001, ETA2 = 0.50). The interaction is due to the fact that stutterers revealed larger WM volumes on the right compared to the left hemisphere.
Figure 2 ROI analysis Mean WM volumes (and standard errors of the mean as vertical bars) in the precentral gyrus (PrCG), middle frontal gyrus (MFG), inferior frontal gyrus (IFG), and the superior temporal gyrus (STG) broken down for the left (open bars, LH) and right (filled bars, RH) hemisphere. The STG comprises Heschl's gyrus and the planum temporale. The volume measures are expressed as arbitrary values because these measures were obtained from brains transformed into the MNI space.
In addition, we did not find any correlation between the stuttering severity measures (SSI measures) and the anatomical peculiarities neither in the context of the VBM nor the ROI analysis.
Discussion
This study was motivated by the question whether stutterers reveal morphological brain anomalies compared to non-stuttering controls. In fact, we found prominent increases of WM in stutterers within a right-hemispheric network including brain structures relevant for language and speech. These areas comprise the STG (including the auditory areas PT and HG), the IFG (including the pars opercularis which is part of Broca's right-sided homologue), the somatosensory area (including the face and mouth representation, as well as the mesial part of the hand representation), and the middle frontal gyrus (MFG). Our findings of regionally increased right-hemispheric WM in stutterers might suggest an increased and possibly atypical intrahemispheric communication within these areas via association fibres [50,51] possibly accompanying different processing strategies in the right hemisphere in stutterers.
Three of the brain areas with different WM composition in stutterers are known to be involved in different speech and language functions. For example, the right IFG (including the pars opercularis) is involved in the perception and generation of phonological or prosodic speech features [52-55] while the ventral part of the precentral gyrus is part of the somatosensory representation of the mouth and tongue. The MFG has been shown to be involved during rhyme and tone perception [34]. The core region of this network is the auditory cortex with neurons specialized for tone, pitch, and prosody perception [56,57]. Within this circuit the auditory cortex plays a pivotal role because speech production follows the ultimate goal to generate speech sounds others can understand. The auditory cues in speech production are either phonetic cues (such as voice onset times or formant transitions) or specific suprasegmental features like duration, intensity, linguistic or emotional stress. During speech production the auditory system controls whether the appropriate auditory cues have been generated by means of auditory feedback control of the own speech. Several studies have shown that the auditory cortex is strongly involved in the continuous control of self-generated suprasegmental speech features (duration, intensity, stress pattern) and that this auditory feedback control is detrimental in stutterers [19-21,45,46,58-60].
The auditory cortices in the two hemispheres are relatively specialized in normal subjects [56,57]. Thus, temporal resolution is better in left auditory cortical areas and spectral resolution as well as processing of prolonged auditory information is better in right auditory cortical areas. It is thought that this functional specialisation is based on cytoarchitectonic peculiarities (more heavily myelinated axons and greater interconnectivity) and the relative composition of WM and GM in this area. In fact, in addition to the present findings, two previous studies [61,62] found a leftward asymmetry of WM volume in the auditory cortex in healthy subjects. However, our findings show that stutterers do not reveal the typical leftward asymmetry; they rather show symmetry with an atypically enlarged WM volume in the right auditory cortex. This atypical symmetry of WM volume in the auditory cortex in stutterers might suggest different and perhaps deficient processing of slowly changing auditory cues necessary to control suprasegmental features. In fact several studies have shown that stutterers reveal substantial peculiarities with respect to various aspects of the auditory feedback of their own speech especially when they control suprasegmental speech features [19,60,63].
The reported morphological features complement previous morphological studies comparing stutterers and non-stutterers. Firstly, this study shows again that stutterers reveal atypical anatomical lateralisation in speech-relevant areas. In three areas (PrCG, MFG, and IFG) stutterers reveal more WM volumes on the right than on the left. For the auditory cortex (STG) we found symmetric WM volumes while non-stutterers typically show a leftward asymmetry for this measure. Thus, some kind of hemispheric imbalance seems to be related to persistent developing stuttering. Secondly, using a different method than Foundas et al. [34], we also found an atypical anatomical lateralisation in the auditory cortex expressed as an increased symmetry of WM volume. Thus, the different "hardware" composition of the auditory cortex in stutterers is a crucial peculiarity possibly determining the processing mode of the right auditory cortex and the interaction between both auditory cortices. Taken together, the present results and findings of previous behavioural and neuroimaging studies emphasize a specific role of the auditory cortex in stuttering. Thirdly, while Foundas et al. [35] found atypical anatomical lateralisation in prefrontal areas we detected increased WM volumes in the right anterior MFG which is part of the prefrontal cortex. Thus, the atypical prefrontal lateralization may be due to atypical lateralisation of the WM volume of the right MFG. Finally, our analysis also revealed an atypical asymmetry with respect to the WM volume in vicinity of the right sensorimotor cortex (including the face and mouth representation as well as parts of the hand representation) possibly suggesting that these areas use different processing strategies as compared to non-stuttering subjects. However, although we and others found large morphological differences between stutterers and non-stuttereres we cannot rule out the possibility that the anatomical differences are the consequence of stuttering rather than the cause. Persistent developmental stuttering commences early in life forcing the affected subject to cope with this annoying and detrimental situation. Thus, some kind of adaptation or cortical reorganisation might accompany this process. Indeed, several studies indicate that intensive practise of various skills might affect the brain even on the macroanatomical level [64,65]. Future studies, however, are clearly needed to disentangle whether the anatomical peculiarities in stutterers are the cause or the consequence of stuttering.
Conclusions
These results provide strong evidence that adults with PDS have anomalous anatomy not only in perisylvian speech and language areas but also in prefrontal and sensorimotor areas. These anatomical features might indicate a deficiently working speech system. Whether this atypical asymmetry of WM is the cause or the consequence of stuttering is still an unanswered question.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
L.J., J.H. and H.S. conceived the experiment and drafted the manuscript. L.J. and J.H. prepared the exact experimental setup. H.S. supervised data acquisition. J.H. and L.J. performed all data and statistical analyses. All authors read and approved the final manuscript.
Figure 1 VBM results Areas where stutterers show increased relative white matter (WM) volume superimposed onto the standard MNI template. The brain outline on the right indicates the four different anatomical regions showing increased WM volume in stutterers.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by a DFG grant awarded to LJ and HS.
==== Refs
Comments on "Stuttering: a review of research findings and theories circa 1982" by Andrews et al J Speech Hear Disord 1983 48 246 263 6621016
Wingate ME A standard definition of stuttering J Speech Hear Disord 1964 29 484 489 14257050
Felsenfeld S Finding susceptibility genes for developmental disorders of speech: the long and winding road J Commun Disord 2002 35 329 345 12160352 10.1016/S0021-9924(02)00088-6
Felsenfeld S Kirk KM Zhu G Statham DJ Neale MC Martin NG A study of the genetic and environmental etiology of stuttering in a selected twin sample Behav Genet 2000 30 359 366 11235981 10.1023/A:1002765620208
Buchel C Sommer M What Causes Stuttering? PLoS Biol 2004 2 E46 14966540
Borden GJ Baer T Kenney MK Onset of voicing in stuttered and fluent utterances J Speech Hear Res 1985 28 363 372 4046577
Borden GJ Initiation versus execution time during manual and oral counting by stutterers J Speech Hear Res 1983 26 389 396 6645464
Metz DE Conture EG Caruso A Voice onset time, frication, and aspiration during stutterers' fluent speech J Speech Hear Res 1979 22 649 656 502520
Jancke L Hefter H Kalveram KT Fast finger extensions are slower in stutterers than in nonstutterers Percept Mot Skills 1995 80 1103 1107 7478864
Vaughn CL Webster WG Bimanual handedness in adults who stutter Percept Mot Skills 1989 68 375 382 2717346
Webster WG Neuropsychological models of stuttering--II. Interhemispheric interference Neuropsychologia 1986 24 737 741 3785662 10.1016/0028-3932(86)90014-X
Webster WG Neuropsychological models of stuttering--I. Representation of sequential response mechanisms Neuropsychologia 1985 23 263 267 4000461 10.1016/0028-3932(85)90110-1
Moore WHJ Bilateral tachistoscopic word perception of stutterers and normal subjects Brain Lang 1976 3 434 442 949595 10.1016/0093-934X(76)90037-7
Sommers RK Brady WA Moore WHJ Dichotic ear preferences of stuttering children and adults Percept Mot Skills 1975 41 931 938 1215135
Wells BG Moore WHJ EEG alpha asymmetries in stutterers and non-stutterers: effects of linguistic variables on hemispheric processing and fluency Neuropsychologia 1990 28 1295 1305 2280837 10.1016/0028-3932(90)90045-P
Travis LE The cerebral dominance theory of stuttering: 1931--1978 J Speech Hear Disord 1978 43 278 281 357839
Orton ST A physiological theory of reading disability and stuttering in children N Engl J Med 1928 199 1046 1052
Kalveram KT Jancke L Vowel duration and voice onset time for stressed and nonstressed syllables in stutterers under delayed auditory feedback condition Folia Phoniatr (Basel) 1989 41 30 42 2759514
Natke U Grosser J Sandrieser P Kalveram KT The duration component of the stress effect in stuttering J Fluency Disord 2002 27 305 317 12506448 10.1016/S0094-730X(02)00163-8
Salmelin R Schnitzler A Schmitz F Jancke L Witte OW Freund HJ Functional organization of the auditory cortex is different in stutterers and fluent speakers Neuroreport 1998 9 2225 2229 9694204
Jancke L Kaiser P Bauer A Kalveram KT Upper lip, lower lip, and jaw peak velocity sequence during bilabial closures: no differences between stutterers and nonstutterers J Acoust Soc Am 1995 97 3900 3903 7790664
Jäncke L Bauer A Kaiser P Kalveram KT Timing and stiffness in the speech motor control of stuttering and nonstuttering adults J Fluency Disord 1997 22 309 321 10.1016/S0094-730X(97)00022-3
Neumann K Euler HA von Gudenberg AW Giraud AL Lanfermann H Gall V Preibisch C The nature and treatment of stuttering as revealed by fMRI A within- and between-group comparison J Fluency Disord 2003 28 381 409 14643071 10.1016/j.jfludis.2003.07.003
Preibisch C Neumann K Raab P Euler HA von Gudenberg AW Lanfermann H Giraud AL Evidence for compensation for stuttering by the right frontal operculum Neuroimage 2003 20 1356 1364 14568504 10.1016/S1053-8119(03)00376-8
Fox PT Ingham RJ Ingham JC Zamarripa F Xiong JH Lancaster JL Brain correlates of stuttering and syllable production. A PET performance-correlation analysis Brain 2000 123 ( Pt 10) 1985 2004 11004117 10.1093/brain/123.10.1985
Fox PT Ingham RJ Ingham JC Hirsch TB Downs JH Martin C Jerabek P Glass T Lancaster JL A PET study of the neural systems of stuttering Nature 1996 382 158 161 8700204 10.1038/382158a0
Weber-Fox C Neural systems for sentence processing in stuttering J Speech Lang Hear Res 2001 44 814 825 11521774
Wu JC Maguire G Riley G Fallon J LaCasse L Chin S Klein E Tang C Cadwell S Lottenberg S A positron emission tomography [18F]deoxyglucose study of developmental stuttering Neuroreport 1995 6 501 505 7766852
Ingham RJ Fox PT Ingham JC Zamarripa F Martin C Jerabek P Cotton J Functional-lesion investigation of developmental stuttering with positron emission tomography J Speech Hear Res 1996 39 1208 1227 8959606
Braun AR Varga M Stager S Schulz G Selbie S Maisog JM Carson RE Ludlow CL Altered patterns of cerebral activity during speech and language production in developmental stuttering. An H2(15)O positron emission tomography study Brain 1997 120 ( Pt 5) 761 784 9183248 10.1093/brain/120.5.761
Jeffries KJ Fritz JB Braun AR Words in melody: an H(2)15O PET study of brain activation during singing and speaking Neuroreport 2003 14 749 754 12692476 10.1097/00001756-200304150-00018
Strub RL Black FW Naeser MA Anomalous dominance in sibling stutterers: evidence from CT scan asymmetries, dichotic listening, neuropsychological testing, and handedness Brain Lang 1987 30 338 350 3567553 10.1016/0093-934X(87)90107-6
Foundas AL Bollich AM Corey DM Hurley M Heilman KM Anomalous anatomy of speech-language areas in adults with persistent developmental stuttering Neurology 2001 57 207 215 11468304
Foundas AL Corey DM Angeles V Bollich AM Crabtree-Hartman E Heilman KM Atypical cerebral laterality in adults with persistent developmental stuttering Neurology 2003 61 1378 1385 14638959
Sommer M Koch MA Paulus W Weiller C Buchel C Disconnection of speech-relevant brain areas in persistent developmental stuttering Lancet 2002 360 380 383 12241779 10.1016/S0140-6736(02)09610-1
Ashburner J Friston KJ Why voxel-based morphometry should be used Neuroimage 2001 14 1238 1243 11707080 10.1006/nimg.2001.0961
Ashburner J Friston KJ Voxel-based morphometry--the methods Neuroimage 2000 11 805 821 10860804 10.1006/nimg.2000.0582
Good CD Ashburner J Frackowiak RS Computational neuroanatomy: new perspectives for neuroradiology Rev Neurol (Paris) 2001 157 797 806 11677400
Good CD Johnsrude IS Ashburner J Henson RN Friston KJ Frackowiak RS A voxel-based morphometric study of ageing in 465 normal adult human brains Neuroimage 2001 14 21 36 11525331 10.1006/nimg.2001.0786
Good CD Johnsrude I Ashburner J Henson RN Friston KJ Frackowiak RS Cerebral asymmetry and the effects of sex and handedness on brain structure: a voxel-based morphometric analysis of 465 normal adult human brains Neuroimage 2001 14 685 700 11506541 10.1006/nimg.2001.0857
Natke U [Reduction of stuttering frequency using frequency-shifted and delayed auditory feedback] Folia Phoniatr Logop 2000 52 151 159 10782008 10.1159/000021530
Natke U Donath TM Kalveram KT Control of voice fundamental frequency in speaking versus singing J Acoust Soc Am 2003 113 1587 1593 12656393 10.1121/1.1543928
Natke U Kalveram KT Effects of frequency-shifted auditory feedback on fundamental frequency of long stressed and unstressed syllables J Speech Lang Hear Res 2001 44 577 584 11407562
Kalinowski J Saltuklaroglu T Choral speech: the amelioration of stuttering via imitation and the mirror neuronal system Neurosci Biobehav Rev 2003 27 339 347 12946686 10.1016/S0149-7634(03)00063-0
Kalinowski J Self-reported efficacy of an all in-the-ear-canal prosthetic device to inhibit stuttering during one hundred hours of university teaching: an autobiographical clinical commentary Disabil Rehabil 2003 25 107 111 12554385
Jancke L Kalveram KT [Control of on-time (duration of phonation) and voice onset time in right and left ear auditory feedback: different lateralization in stutterers and non-stutterers?] Z Exp Angew Psychol 1987 34 54 63 3577254
Annett M A classification of hand preference by association analysis Br J Psychol 1970 61 303 321 5457503
Riley GD A stuttering severity instrument for children and adults J Speech Hear Disord 1972 37 314 322 5057250
Jancke L Schlaug G Huang Y Steinmetz H Asymmetry of the planum parietale Neuroreport 1994 5 1161 1163 8080979
Schlaug G Jancke L Huang Y Steinmetz H In vivo evidence of structural brain asymmetry in musicians Science 1995 267 699 701 7839149
Steinmetz H Volkmann J Jancke L Freund HJ Anatomical left-right asymmetry of language-related temporal cortex is different in left- and right-handers Ann Neurol 1991 29 315 319 2042947
Friston KJ Holmes AP Poline JB Grasby PJ Williams SC Frackowiak RS Turner R Analysis of fMRI time-series revisited Neuroimage 1995 2 45 53 9343589 10.1006/nimg.1995.1007
Rademacher J Morosan P Schormann T Schleicher A Werner C Freund HJ Zilles K Probabilistic mapping and volume measurement of human primary auditory cortex Neuroimage 2001 13 669 683 11305896 10.1006/nimg.2000.0714
Westbury CF Zatorre RJ Evans AC Quantifying variability in the planum temporale: a probability map Cereb Cortex 1999 9 392 405 10426418 10.1093/cercor/9.4.392
Rauschecker JP Tian B Mechanisms and streams for processing of "what" and "where" in auditory cortex Proc Natl Acad Sci U S A 2000 97 11800 11806 11050212 10.1073/pnas.97.22.11800
Rauschecker JP Making brain circuits listen Science 1999 285 1686 1687 10523187 10.1126/science.285.5434.1686
Buchanan TW Lutz K Mirzazade S Specht K Shah NJ Zilles K Jancke L Recognition of emotional prosody and verbal components of spoken language: an fMRI study Brain Res Cogn Brain Res 2000 9 227 238 10808134 10.1016/S0926-6410(99)00060-9
Riecker A Wildgruber D Dogil G Grodd W Ackermann H Hemispheric lateralization effects of rhythm implementation during syllable repetitions: an fMRI study Neuroimage 2002 16 169 176 11969327 10.1006/nimg.2002.1068
Riecker A Ackermann H Wildgruber D Meyer J Dogil G Haider H Grodd W Articulatory/phonetic sequencing at the level of the anterior perisylvian cortex: a functional magnetic resonance imaging (fMRI) study Brain Lang 2000 75 259 276 11049668 10.1006/brln.2000.2356
Wildgruber D Hertrich I Riecker A Erb M Anders S Grodd W Ackermann H Distinct Frontal Regions Subserve Evaluation of Linguistic and Emotional Aspects of Speech Intonation Cereb Cortex 2004 14 1384 9 15217896 10.1093/cercor/bhh099
Zatorre RJ Belin P Penhune VB Structure and function of auditory cortex: music and speech Trends Cogn Sci 2002 6 37 46 11849614 10.1016/S1364-6613(00)01816-7
Zatorre RJ Belin P Spectral and temporal processing in human auditory cortex Cereb Cortex 2001 11 946 953 11549617 10.1093/cercor/11.10.946
Kalinowski J Dayalu VN A common element in the immediate inducement of effortless, natural-sounding, fluent speech in people who stutter: 'the second speech signal' Med Hypotheses 2002 58 61 66 11863399 10.1054/mehy.2001.1451
Kalinowski J Stuart A Wamsley L Rastatter MP Effects of monitoring condition and frequency-altered feedback on stuttering frequency J Speech Lang Hear Res 1999 42 1347 1354 10599617
Kalinowski J Armson J Roland-Mieszkowski M Stuart A Gracco VL Effects of alterations in auditory feedback and speech rate on stuttering frequency Lang Speech 1993 36 ( Pt 1) 1 16 8345771
Penhune VB Zatorre RJ MacDonald JD Evans AC Interhemispheric anatomical differences in human primary auditory cortex: probabilistic mapping and volume measurement from magnetic resonance scans Cereb Cortex 1996 6 661 672 8921202
Anderson B Southern BD Powers RE Anatomic asymmetries of the posterior superior temporal lobes: a postmortem study Neuropsychiatry Neuropsychol Behav Neurol 1999 12 247 254 10527109
Kalinowski J Stuart A Stuttering amelioration at various auditory feedback delays and speech rates Eur J Disord Commun 1996 31 259 269 8944848
Munte TF Altenmuller E Jancke L The musician's brain as a model of neuroplasticity Nat Rev Neurosci 2002 3 473 478 12042882
Draganski B Gaser C Busch V Schuierer G Bogdahn U May A Neuroplasticity: changes in grey matter induced by training Nature 2004 427 311 312 14737157 10.1038/427311a
| 15588309 | PMC539354 | CC BY | 2021-01-04 16:28:50 | no | BMC Neurol. 2004 Dec 10; 4:23 | utf-8 | BMC Neurol | 2,004 | 10.1186/1471-2377-4-23 | oa_comm |
==== Front
BMC Pregnancy ChildbirthBMC Pregnancy and Childbirth1471-2393BioMed Central London 1471-2393-4-251558832410.1186/1471-2393-4-25Study ProtocolThe DiAMOND trial protocol: a randomised controlled trial of two decision aids for mode of delivery among women with a previous caesarean section [ISRCTN84367722] Montgomery Alan A [email protected] DiAMOND Study Group [email protected] Academic Unit of Primary Health Care, Department of Community Based Medicine, University of Bristol, The Grange, 1 Woodland Road, Bristol BS8 1AU, United Kingdom2004 10 12 2004 4 25 25 1 12 2004 10 12 2004 Copyright © 2004 Montgomery and for the DiAMOND Study Group; licensee BioMed Central Ltd.2004Montgomery and for the DiAMOND Study Group; 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
Caesarean section (CS) has become an increasingly common method of delivery worldwide, rising in the UK from 9% of deliveries in 1980 to over 21% 2001. This increase, and the question of whether CS should be available to women on request, has been the subject of considerable debate, and national reports and guidelines have specifically highlighted the importance of patient choice in the decision making process. For women who have already experienced CS, the UK National Institute of Clinical Excellence recommends that the decision should consider maternal preferences and priorities in addition to general discussion of the overall risks and benefits of CS. Decision aids for many different medical treatment and screening decisions have been developed and evaluated, but there is relatively little evidence for the use of decision aids for choice of mode of delivery among women with a previous CS. The aim of the study is to evaluate two interventions to assist decision making about mode of delivery among pregnant women with one previous CS.
Methods/design
Women with one previous CS are recruited to the trial during their booking visit at approximately 12–20 weeks' gestation in participating maternity units in Bristol, Weston and Dundee. Using central randomisation, women are allocated to one of three arms: information programme and website; decision analysis; usual care. Both interventions are computer-based, and are designed to provide women with detailed information about the potential outcomes for both mother and baby of planned vaginal delivery, planned CS and emergency CS. The decision analysis intervention additionally provides a recommended 'preferred option' based on maximised expected utility. There are two primary outcomes (decisional conflict and actual mode of delivery), and five secondary outcomes (anxiety, knowledge, perceptions of shared decision making; satisfaction with decision making process, proportion of women attempting vaginal delivery). Primary follow up for the questionnaire measures is at 36–37 weeks' gestation, and a total of 660 women will be recruited to the study. The primary intention-to-treat analyses will comprise three pair-wise comparisons between decision analysis, information and usual care groups, for each of the two primary outcomes. A qualitative study will investigate women's experiences of the decision making in more depth, and an economic evaluation from the perspective of the NHS will be conducted.
Discussion
Provision of information to women facing this decision appears variable. The DiAMOND study aims to inform best practice in this area by evaluating the effectiveness of two interventions designed to aid decision making.
==== Body
Background
Over the last 20 years caesarean section (CS) has become an increasingly common method of delivery. The CS rate in the United Kingdom rose from 9% of deliveries in 1980 to 21% in 2001 [1]. This increase, and the question of whether CS should be available to women on request, has been the subject of considerable debate [2]. The optimal mode of delivery for women who have experienced a previous CS is complicated by the difficulty in balancing the risks of repeat CS with those of vaginal birth after caesarean section (VBAC). An evaluation of caesarean delivery by the American College of Obstetricians and Gynecologists reported that first time mothers with term singleton cephalic pregnancies and women with a previous CS account for two thirds of all caesarean deliveries in the US [3].
The Changing Childbirth report has emphasised the importance of patient choice when decisions need to be made in relation to the management of pregnancy and childbirth [4]. However the views of women who have experienced CS and their preferences for future deliveries have received little attention to date. Obstetricians tend to focus on the risks of uterine rupture and emergency caesarean section in labour [5], which may influence the advice they give to women uncertain about future mode of delivery. Others have focused on the increased morbidity following CS [6] and on the negative impact of operative delivery on first postnatal contact between the mother and her baby [7]. In Chile, where there is a very high rate of CS in the private sector, only a minority of women reported that they had wanted this method of delivery [8]. In a Scottish study, more women who delivered by elective CS reported they were satisfied with their involvement in the decision making process compared with women who underwent emergency CS [9]. The advantages and disadvantages of planned CS versus planned vaginal delivery has also been debated in north America in response to a growing number of requests for elective CS [10].
In an attempt to ensure appropriateness of CS in the UK, a set of evidence-based guidelines on indications for CS have recently been published [11]. The guidelines, commissioned by the UK National Institute for Clinical Excellence, make a specific recommendation that for women with a previous CS, the decision should consider maternal preferences and priorities in addition to general discussion of the overall risks and benefits of CS. It is essential that the process of decision making about future mode of delivery is evaluated and enhanced to achieve the safest and most satisfactory outcome for both mother and baby.
It has been proposed that the way in which clinical decisions are made lies on a continuum, from paternalistic (clinician decides) through partnership (clinician and patient share the decision) to informed (patient decides) [12]. Although proposed as the preferred approach of determining patients' treatments [13], some problems with the concepts, terminology and practice of shared medical decision making have recently been highlighted [14]. The appropriateness of the shared model may also depend on the clinical context as well as patients' and clinicians' preferences for involvement in decision making [15]. In addition, there is some evidence that patients and health professionals often have different treatment preferences, potentially making agreement on a treatment strategy more difficult [16].
Decision aids are designed to help people select between various treatment strategies by providing information on the options and outcomes relevant to a person's health. Decision aids for many clinical conditions have been developed [17], and evaluations of these decision aids have been the subject of systematic reviews [18,19]. A north American trial found no difference in terms of VBAC rate between written versus personal counselling interventions that actively promoted vaginal delivery [20]. An Australian trial of a paper-based decision aid for women who have previously experienced CS is currently underway [21].
The specific content of decision aids may vary, but in general they aim to present more than one strategy for clinical management, help people understand the probable outcomes of treatment choices and allow people to consider the personal value they place on benefits versus harms. Decision aids can take several formats, such as leaflets, interactive videodisks, individualised decision analysis, personal counselling sessions and audio workbooks. Interventions to assist patient decision making can improve knowledge about treatment options, make patients more realistic in their expectations, reduce decisional conflict and increase active involvement in decision making [18].
As an intervention to aid patient decision making, individualised decision analysis has so far received limited attention. By explicitly combining patients' values regarding treatment outcomes and individual probability information, decision analysis attempts to provide a rational framework to guide patient decision making. The use of individualised decision analysis has been debated [22], but empirical evidence demonstrates that it is feasible and acceptable and has value as an aid to patient decision making [23].
Aim
The aim of this paper is to describe the protocol for a randomised controlled trial of two interventions to aid decision making about mode of delivery among pregnant women with one previous CS. The interventions being assessed are (1) Decision analysis, and (2) Information programme and website. Both decision aids will be compared with usual care given by the obstetric team. The interventions will be assessed in terms of decisional conflict, planned and actual mode of delivery, anxiety, knowledge, perception of shared decision making and satisfaction with the decision making process. Development and piloting of the interventions took place in 2003–2004, and the main phase of the study started in May 2004 and will continue until December 2006. Ethical approval for the study was obtained from the UK South West Multi-Centre Research Ethics Committee.
Methods/design
Recruitment and allocation of participants
The sample will comprise pregnant women with one previous lower segment CS (all parities will be included, but the most recent delivery must have been CS), no current obstetric problems and delivery expected at ≥ 37 weeks. Women are being recruited to the study by research midwives during their initial booking visit at approximately 12–20 weeks' gestation. Recruitment takes place from maternity units in St Michael's and Southmead Hospitals in Bristol, Weston General Hospital, and Ninewells Hospital in Dundee. The current CS rates for these units range between 18 and 24% and are representative of rates for other units throughout England and Scotland. The women are informed that although both vaginal delivery and repeat CS carry their own benefits and risks, the best method of presenting this information in order to assist women in reaching a decision is not known. Women expressing an interest in participating in the trial at the booking visit are given an information sheet, a written consent form and a baseline questionnaire to take home. Following receipt of the baseline questionnaire and written informed consent to enter the trial, women are randomised to one of three arms as detailed below. Allocation is stratified by maternity unit and preferred mode of delivery and blocked (using random permuted blocks of sizes 6, 9, 12 and 15) to ensure reasonable balance between the trial groups through time. The randomisation sequence was generated by a member of the study team (AAM), and allocation of participants is performed by a staff member with no other involvement in the study.
Interventions
Both interventions are computer based. Women allocated to receive an intervention have an appointment with a researcher to allow the decision aid to be delivered using a laptop computer, usually in the woman's own home or workplace.
(i) Information programme and website
This intervention provides information about the outcomes associated with planned vaginal delivery, planned CS, and emergency CS. This includes descriptions of outcomes for both mother and baby, and the probabilities of these outcomes based on the best available evidence. Both the probabilities of having and not having the event are given, and all probabilities are presented in both numerical and pictorial format [24]. The programme easily allows women to choose the information that they view, and the information each women accesses is logged. At the end of the initial appointment with a researcher, women are given a password that allows them to access the information programme via the internet as often as they wish. Womens' use of the programme via the internet is also logged.
(ii) Decision analysis
There are generally four main steps involved in a decision analysis. The first is to draw up a decision tree that maps out the likely outcomes of the strategies in question [25]. These outcomes are then assigned utilities that represent how an individual values a particular outcome. A utility is a number between 0 and 1, often representing the outcomes 'death' and 'perfect health' respectively [26]. Probability information is then included in the tree to represent the chance of each outcome occurring [26]. Finally, strategies are compared by calculating the weighted sum of the utilities of all possible outcomes [27]. The recommended strategy is that with the highest expected utility value, or in other words, the one that gives an individual the best chance of achieving an outcome that is valued.
The decision analysis intervention in the trial proceeds according to the principles described above. First, women are given information about the outcomes associated with planned vaginal delivery, planned CS, and emergency CS. This includes descriptions of outcomes for both mother and baby, but not the probabilities of these events. These are embedded in the decision tree which is not visible to users. Second, women are required to rate (assign a utility value between 0 and 1) each possible outcome using a visual analogue scale. Finally, the programme combines the elicited utilities and the probabilities of each outcome in a decision tree to produce a recommended 'preferred option' based on maximised expected utility. Each woman is given a computer printout of the outcome of the decision analysis.
(iii) Usual care
This comprises care normally given by the obstetric and midwifery team. Women allocated to decision analysis or information programme receive these interventions in addition to usual care.
Women in both intervention arms are contacted by letter at 35 weeks' gestation. The purpose is to encourage discussion of the intervention with their obstetrician and/or midwife when they attend the clinic at 36–37 weeks to finalise their planned method of delivery. Participation in the study is recorded in the medical records of all women in the trial.
Outcome measures
There are two primary outcomes:
(1) Decisional Conflict Scale (DCS) [28,29]. This is a 16 item questionnaire that measures degree of uncertainty about which course of action to take and the main modifiable factors contributing to uncertainty. Previous research indicates that effect sizes of about 0.35 to 0.5 standard deviations can discriminate between individuals who make a decision and those who delay or are unsure [28]. Assuming a standard deviation of 15 points [18], this is equivalent to differences of 5.25 to 7.5 points on the total DCS 100 point scale.
(2) Actual mode of delivery (vaginal, elective CS, or emergency CS). Unlike a previous trial [20], we are not seeking to promote one method of delivery over another. However differences between the arms in the proportions of different modes of delivery may have important healthcare resource implications.
There are five secondary outcomes: anxiety; [30] knowledge; perception of shared decision making; satisfaction with decision making process; proportion of women attempting vaginal delivery.
Follow up
The primary and secondary outcomes are assessed in all three groups at baseline, and approximately two weeks after randomisation. The questionnaire at two weeks will constitute a secondary follow up, and will enable sufficient time for delivery of the appropriate interventions. As part of usual care, women in the trial normally attend the clinic at around 36–37 weeks' gestation to finalise plans for their preferred method of delivery. Questionnaire outcomes are measured again after this visit, and this constitutes the primary follow up for this trial. Actual and attempted mode of delivery (cross-checked with hospital records) and satisfaction with choice are assessed by a further postal questionnaire at approximately six to eight weeks after giving birth.
Justification of sample size
As noted above, differences in excess of 0.35 standard deviations have been considered as important for the total DCS score, and differences of this magnitude are feasible for interventions of this kind [23]. With regard to mode of delivery, UK data indicate that about 33% of women with one previous CS are delivered vaginally [1]. The sample size calculation for a previous trial of written versus verbal counselling in north America presumed a vaginal delivery rate of 30% for a minimal intervention, and in the event observed that 51% of women achieved vaginal delivery for the trial groups overall [20]. A change from 30–33% to 51% corresponds to an odds ratio of about 2.1–2.4, and this would certainly be considered as clinically important.
With two-sided 1% alpha, a total sample size of 600 provides 82–99% power to detect a standardised difference of 0.35–0.5 in total decisional conflict score between the groups, and 84–95% power to detect odds ratios of 2.1–2.4 in women achieving vaginal delivery. A pair-wise alpha of 1%, corrected for multiple comparisons using Tukey's procedure, maintains an overall study-wise alpha of 3.4%. In order to allow for pre-term deliveries, malpresentations, and losses to follow-up, we will therefore recruit 660 women to the trial.
Statistical analysis
Data analysis will proceed according to CONSORT guidelines for randomised controlled trials. The first stage of the analysis will be to use descriptive statistics to describe the group of individuals recruited to the trial in relation to those eligible, and to investigate comparability of the trial arms at baseline. The primary analyses will comprise three pair-wise intention-to-treat comparisons between decision analysis, information and usual care groups, for each of the two primary outcomes. These comparisons will use appropriate (that is, standard or logistic) multivariable regression models, adjusting for maternity unit, initial preference regarding mode of delivery, and value of the outcome variable at baseline. Full attention will be paid to the estimates and confidence intervals for these comparisons as well as the p-values, with the latter adjusted for multiple comparisons using Tukey's procedure. Secondary outcomes will then be analysed in the same way, using appropriate multivariable regression models depending on the nature of the outcomes.
Other secondary analyses will involve investigation of the short-term effects of the interventions using data from the two week follow up, and the effects at 36–37' weeks gestation adjusted for both baseline and two week follow up. Pre-planned subgroup analyses employing appropriate interaction terms in the regression models will be used to ascertain any differential effects of the interventions on the two primary outcomes across the following categories of women: previous caesarean section occurring before or after labour; previous successful vaginal delivery; stated preferred mode of delivery. Since the trial is powered to detect overall differences between the groups rather than interactions of this kind, the results of these essentially exploratory analyses will be presented using confidence intervals as well as p-values, and interpreted with due caution. Finally, we will investigate the effect of differential use of the information intervention via the internet using descriptive statistics and appropriate comparisons with the other groups.
Qualitative study
Qualitative research methods will be used in order to explore aspects of the interventions and women's experiences of the decision making in more depth. Specifically, semi-structured interviews will be conducted with a sample of women from each of the intervention arms (Decision Analysis and Information), across the research sites in Bristol and Dundee. These interviews will explore:
(1) Women's views and experiences of the intervention and its delivery – for example, the quality and relevance of the decision aid/information, what they felt about the risk information presented to them, and which particular aspects of each intervention were helpful or unhelpful.
(2) Which factors women felt had most influenced their preferences regarding method of delivery.
(3) Whether the women had prior preferences about method of delivery, whether/how these changed during their pregnancy and in the case of the decision analysis, what they felt about the method of delivery proposed by the intervention compared to any prior preferences.
(4) Any other information sources women sought and used to help them make a decision about method of delivery (for example, information from health professionals, partner/family/friends, internet, media, books).
(5) Women's views and feelings about their actual method of delivery compared with any prior preferences and the method of delivery suggested by the intervention (for decision analysis).
A small number of interviews may also be conducted with women in the usual care arm of the trial to explore what support and advice was normally provided during pregnancy to those who did not receive an intervention.
A sample of approximately 30 women across the two intervention arms will be interviewed in depth to ensure a thorough exploration of emergent themes and concepts. Within each arm, women will be purposefully chosen to include those with different parities/ages/socio-economic backgrounds, and where possible different methods of delivery/outcomes, following a maximum variation sampling strategy [31]. The interviews will take place a short time after the primary follow up, to avoid any influence of the interview on these measures. A subset of the women will be interviewed a second time six to eight weeks after birth in order that they can reflect upon their decision regarding preferred mode of delivery with the actual mode of delivery. In addition, a small number of interviews may take place closer in time to receipt of the intervention if this is deemed (a) feasible given recruitment and (b) worthwhile in terms of providing new information to that gleaned from qualitative interviews undertaken in the development phase of the trial.
Interviews will be conducted in the womens' homes or other suitable setting chosen by them. A check-list of topics will be used to ensure that the primary issues are covered, whilst allowing flexibility for new issues to emerge from each interview. Interviews will be recorded on mini-disc, fully transcribed and anonymised to protect confidentiality. Transcripts will be studied in detail and a list of common themes and concepts drawn up. Data collection and analysis will run in parallel and the coding index added to or refined and coded material regrouped as new themes and categories emerge from subsequent interviews [32]. Further analysis will employ the constant comparison method of grounded theory in which the textual data is scrutinised for differences and similarities within themes keeping in mind the context in which mention of these themes arose in each interview [32].
In addition to the interviews with women, a small number of focus groups with health professionals (e.g. obstetricians, midwives) in each research site may be conducted near the end of the trial, resources and time permitting. Focus groups are often used in evaluations of new services/interventions and are useful for exploring group views, concerns and preferences (e.g. consensus or disagreement about an issue) [33]. They may be valuable for exploring professionals' views about the intervention and issues around implementation into routine practice.
Economic evaluation
The aim of the economic evaluation is to compare the costs and benefits of the two interventions with usual care. The analysis will be from an NHS perspective and will be based on the costs incurred during pregnancy, delivery and 6–8 weeks following delivery. Incremental cost effectiveness ratios will be formed comparing (i) the cost per point improvement on the Decisional Conflict Scale (DCS) and (ii) the cost per extra patient able to make a decision (represented by a 0.5 standard deviation change on the DCS). We will also compare the average cost of care per patient in each arm of the trial with patient satisfaction. Differences in the cost of care of women receiving the two interventions will be compared with usual care from the point at which patients are randomised. The analysis will include all resources under the control of the NHS that may differ as a result of the interventions and will include resources used by both mother and baby. The costs identified as being of relevance are: antenatal appointments in addition to routine appointments, including both primary and secondary care; mode of delivery and related hospital stay for mother and baby; follow up care for mother and baby, including primary and secondary care appointments, outpatient appointments, inpatient stays, A&E visits, and prescribed medication.
Data on resource use will be collected principally by two questionnaires completed by the women. The first questionnaire, completed at 36–37 weeks' gestation, will provide information on all antenatal appointments, and the proportion of these that involved discussion about mode of delivery. Women will be asked how often mode of delivery was addressed at (i) routine appointments, and (ii) appointments initiated by them to discuss mode of delivery. A sample of 10 hand-held records from each arm at each main centre (St Michaels, Southmead and Ninewells, total 90) will be scrutinised to validate the information given in response to the questionnaire. The second questionnaire, completed six to eight weeks after delivery, will provide information on mode of delivery and all non-routine postnatal health service contacts and prescriptions for both mother and baby.
The principle of opportunity cost will underlie the valuation of resource use. In many cases, market rates will act as a proxy for opportunity cost. National data sets such as the Unit Costs of Health and Social Care [34] and the British National Formulary [35] will be used to value primary care consultations and prescribed medication. Secondary care contacts will be coded according to Healthcare Resource Group (HRG) and valued using the Department of Health National Reference Costs [36].
Costs and outcomes will not be discounted, as the economic evaluation will be limited to a period of 12 months. Sensitivity analysis will be conducted in areas where there is uncertainty around assumptions about resource use measurement and/or valuation. Variation in resource use and the effectiveness of the intervention is not captured by a cost-effectiveness/utility ratio. We will use bootstrapping to address this, and construct a cost effectiveness acceptability curve.
Discussion
Women with an uncomplicated pregnancy and expected term delivery who have previously delivered by CS face a choice between repeat elective CS or attempted trial of labour. Guidelines in the UK emphasise the importance of involving women in the decision making process and taking account of maternal preferences and priorities, but the type and amount of information available to women facing this choice appears variable. The DiAMOND study is a randomised trial that aims to inform best practice in this area, by evaluating the effectiveness of two interventions to assist decision making in terms of decision quality and actual mode of delivery.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Alan Montgomery, Deirdre Murphy, Ali Shaw and Sandra Hollinghurst drafted the manuscript, with input from the other members of the DiAMOND Study Group.
The Decision Aids for Mode Of Next Delivery (DiAMOND) Study Group comprises the following members: Clare Emmett, Tom Fahey, Peter Gregor, Sandra Hollinghurst, Claire Jones, Beverley Lovering, Alan Montgomery, Irene Munro, Deirdre Murphy, Roshni Patel, Tim Peters, Ian Ricketts, Anne Schlegelmilch, Ali Shaw, Kav Vedhara, Kate Warren.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgement
The DiAMOND study is funded by a grant from the BUPA Foundation.
==== Refs
Royal College of Obstetricians and Gynaecologists Clinical Effectiveness Support Unit The National Sentinel Caesarean Section Audit Report 2001 London, RCOG Press
Wagner M Choosing caesarean section Lancet 2000 356 1677 1680 11089840 10.1016/S0140-6736(00)03169-X
American College of Obstetricians and Gynecologists Evaluation of cesarean delivery 2000 Washington, DC, ACOG
Department of Health Changing Childbirth. Part 1: Report of the Expert Maternity Group 1993 London, HMSO
Lydon-Rochelle M Holt VL Easterling TR Martin DP Risk of uterine rupture during labor among women with a prior cesarean delivery NEJM 2001 345 3 8 11439945 10.1056/NEJM200107053450101
Murphy DJ Liebling RE Verity L Swingler R Patel R Early maternal and neonatal morbidity associated with operative delivery in second stage of labour: a cohort study Lancet 2001 358 1203 1207 11675055 10.1016/S0140-6736(01)06341-3
Rowe-Murray HJ Fisher JRW Operative intervention in delivery is associated with compromised early mother-infant interaction British Journal of Obstetrics and Gynaecology 2001 108 1068 1075 11702839 10.1016/S0306-5456(01)00242-X
Murray SF Relation between private health insurance and high rates of caesarean section in Chile: qualitative and quantitative study BMJ 2000 321 1501 1505 11118176 10.1136/bmj.321.7275.1501
Graham WJ Hundley V McCheyne AL Hall MH Gurney E Milne J An investigation of women's involvement in the decision to deliver by caesarean section B J Obstet Gynaecol 1999 106 213 220
Minkoff H Chervenak FA Elective primary cesarean delivery NEJM 2003 348 946 950 12621140 10.1056/NEJMsb022734
National Collaborating Centre for Women's and Children's Health Caesarean Section 2004 London, RCOG Press
Charles C Whelan T Gafni A What do we mean by partnership in making decisions about treatment? BMJ 1999 319 780 782 10488014
Coulter A Paternalism or partnership? BMJ 1999 319 719 720 10487980
McNutt RA Shared medical decision making - problems, process, progress JAMA 2004 292 2516 2518 15562133 10.1001/jama.292.20.2516
Robinson A Thomson R Variability in patient preferences for participating in medical decision making: implication for the use of decision support tools Qual Health Care 2001 10 (Suppl I) i34 i38 11533436
Montgomery AA Fahey T How do patients' treatment preferences compare with those of clinicians? Quality in Health Care 2001 10 (Suppl I) i39 i43 11533437
Edwards A Elwyn G The potential benefits of decision aids in clinical medicine JAMA 1999 282 779 780 10463715 10.1001/jama.282.8.779
O'Connor AM Stacy D Entwistle V Llewellyn-Thomas HA Rovner D Holmes-Rovner M Tait V Tetroe J Fiset V Barry M Jones J Decision aids for people facing health treatment or screening decisions The Cochrane Database of Systematic Reviews 2003
Estabrooks C Goel V Thiel E Pinfold P Sawka C Williams I Decision aids: are they worth it? A systematic review Journal of Health Services Research and Policy 2001 6 170 182 11467275 10.1258/1355819011927431
Fraser W Maunsell E Hodnett E Moutquin JM Randomized controlled trial of a prenatal vaginal birth after cesarean section education and support program Am J Obstet Gynaecol 1997 176 419 425
Shorten A Chamberlain M Shorten B Kariminia A Making choices for childbirth: development and testing of a decision-aid for women who have experienced previous caesarean Patient Education and Counseling 2004 52 307 313 14998601 10.1016/S0738-3991(03)00106-X
Elwyn G Edwards A Eccles M Rovner D Decision analysis in patient care Lancet 2001 358 571 574 11520546 10.1016/S0140-6736(01)05709-9
Montgomery AA Fahey T Peters TJ A factorial randomised controlled trial of decision analysis and an information video plus leaflet for newly diagnosed hypertensive patients Br J Gen Pract 2003 53 446 453 12939889
Edwards A Elwyn G Mulley A Explaining risks: turning numerical data into meaningful pictures BMJ 2002 324 827 830 11934777 10.1136/bmj.324.7341.827
Detsky AS Naglie G Krahn MD Redelmeier DA Naimark D Primer on Medical Decision Analysis: Part 2 - Building a tree Med Decis Making 1997 17 126 135 9107607
Naglie G Krahn MD Naimark D Redelmeier DA Detsky AS Primer on medical decision analysis: Part 3 - Estimating probabilities and utilities Med Decis Making 1997 17 136 141 9107608
Krahn MD Naglie G Naimark D Redelmeier DA Detsky AS Primer on medical decision analysis: Part 4 - Analyzing the model and interpreting the results Med Decis Making 1997 17 142 151 9107609
O'Connor AM Validation of a decisional conflict scale Med Decis Making 1995 15 25 30 7898294
O'Connor AM Decisional Conflict Scale 2004
Marteau TM Bekker H The development of a six-item short-form of the state scale of the Spielberger State-Trait Anxiety Inventory (STAI) Br J Clin Psychol 1992 31 301 306 1393159
Patton MQ Qualitative Research and Evaluation Methods 2002 3rd London, Sage
Hammersley M Atkinson P Ethnography Principles in practice 1996 London, Routledge 1 323
Kreuger RA Focus Groups: A Practical Guide for Applied Research 1994 2nd London, Sage
Personal Social Services Research Unit 2004
British National Formulary 2004
Department of Health 2004
| 15588324 | PMC539355 | CC BY | 2021-01-04 16:32:03 | no | BMC Pregnancy Childbirth. 2004 Dec 10; 4:25 | utf-8 | BMC Pregnancy Childbirth | 2,004 | 10.1186/1471-2393-4-25 | oa_comm |
==== Front
BMC GastroenterolBMC Gastroenterology1471-230XBioMed Central London 1471-230X-4-311559601810.1186/1471-230X-4-31Research ArticleSeroprevalence of hepatitis C and associated risk factors among an urban population in Haiti Hepburn Matthew J [email protected] Eric J [email protected] Department of Medicine, Brooke Army Medical Center, San Antonio, Texas, USA2 US Army Medical Research Institute of Infectious Diseases, MCMR-UIM-R (Hepburn), 1425 Porter Street, Fort Detrick, MD, 21702-5011, USA3 Alamo Medical Research, 621 Camden Suite 202, San Antonio, Texas, 78215, USA2004 14 12 2004 4 31 31 28 6 2004 14 12 2004 Copyright © 2004 Hepburn and Lawitz; licensee BioMed Central Ltd.2004Hepburn and Lawitz; 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 seroprevalence of hepatitis C varies substantially between countries and geographic regions. A better understanding of the seroprevalence of this disease, and the risk factors associated with seropositive status, supply data for the development of screening programs and provide insight into the transmission of the disease. The purpose of this investigation was to determine the seroprevalence of hepatitis C and associated risk factors in an urban population in Haiti.
Methods
A prospective survey for hepatitis C antibodies was conducted among an urban outpatient population in Cap-Haïtien, Haiti, with a sample size of 500 subjects. An anonymous 12 question survey, with inquiries related to demographic characteristics and risk factors for HCV acquisition, was concomitantly administered with testing. These demographic and behavioral risk factors were correlated with HCV antibody status using univariate and multivariate tests.
Results
The prevalence of positive HCV antibody was 22/500 (4.4%). Subjects that were anti-HCV positive had an average of 7 ± 8.6 lifetime sexual partners, compared to average of 2.5 ± 3.5 lifetime sexual partners among HCV-negative subjects (p = 0.02). In a multiple logistic regression model, intravenous drug use (OR 3.7, 1.52–9.03 95% CI) and number of sexual partners (OR 1.1, 1.04–1.20 95% CI) were independently associated with a positive HCV antibody result.
Conclusions
A substantial number of subjects with HCV antibodies were detected in this population in Haiti. Further investigation into the correlation between the number of sexual partners and testing positive for hepatitis C antibodies is indicated.
==== Body
Background
The seroprevalence of hepatitis C virus (HCV) varies substantially in different geographic regions throughout the world [1]. Prior studies have suggested a low prevalence of HCV antibodies among a sample of patients in rural Haiti [2]. No cases of positive HCV antibody were detected among 485 patients in a sexually transmitted infections clinic in Jamaica [3], but 41% of hemophiliacs in Jamaica were HCV antibody positive [4]. Our hypothesis was that a higher prevalence of HCV antibodies would be detected in an urban population in Haiti.
Risk factors associated with HCV serologic status may be specific to a country or region. In particular, the role of sexual contact in the transmission of HCV appears to be influenced by characteristics and location of the population studied [5]. Therefore, examination of the risk factors associated with the presence of HCV antibodies in this population can be utilized to guide screening procedures as well as provide insight into the transmission of HCV in the context of Haitian society. A better understanding of the transmission of HCV could enhance the effectiveness of prevention efforts. This research study provides analysis of risk factors associated with hepatitis C from a population that had not been previously studied. We have observed a seroprevalence rate of 4.4% (22/500) of HCV antibodies, with intravenous drug usage and the number of sexual partners being associated with positive HCV antibodies.
Methods
The study was approved by the Institutional Review Board at Brooke Army Medical Center (Fort Sam Houston, Texas). Subjects were recruited in the year 2000 from a healthy population utilizing hospital and clinic services in Cap-Haïtien, Haiti, which is the second largest city in the country. Subjects were recruited on presentation to the hospital laboratory for blood draws for routine laboratory tests. This laboratory was the only location in which subjects were recruited and samples obtained. Subject participation was not limited to a particular medical condition. The first 500 subjects that agreed to participate in the study were enrolled. These subjects were presenting to the hospital for services, and were not patients in a specific clinic. After informed consent was obtained, subjects completed a written 12 question survey in Creole, at the same location as where the blood was drawn. The subjects completed the survey by themselves. The questionnaire focused on demographic information and topic areas possibly associated with transmission of hepatitis C (intravenous drug use, intranasal cocaine use, blood transfusions, sexual history, number of tattoos). Intravenous drug use and cocaine use was measured on a 0–3 scale (never-rare-frequent-daily). Blood transfusions, number of sexual partners, age of first sexual intercourse and number of tattoos were quantified. Serum was obtained for testing for HCV antibody, utilizing the Abbott HCV EIA 3.0 kit™ (Abbott Laboratories, Abbott Park, Illinois). The survey information and serum results were identified only by a subject identification number, with all other identifying information removed.
Data were analyzed using univariate correlations with a Pearson's correlation coefficient. The number of sexual partners was compared between HCV-positive and negative subjects using an independent-sample t-test. Intravenous drug use between HCV-seropositive and seronegative subjects was compared using a Fisher's exact test. A multivariate logistic regression model was employed with stepwise backward elimination of non-significant variables, with HCV antibody status as the dependent variable. All of the variables from the survey were included in the model.
Results
A total of 500 subjects were recruited and had serum tested for HCV antibody. Only two of these subjects did not complete the survey. Most subjects who were informed of the study agreed to participate, but an exact number of subjects who refused participation in the study was not determined. The background characteristics of the patient population are displayed in Table 1. Of note, few subjects (12/496) admitted intravenous drug use, and only one subject noted having a tattoo.
Table 1 Characteristics and survey responses of recruited subjects
Characteristic Frequency
Sex- % males 332/498 (67%)
Age (mean ± SD in years) 33.7 ± 15.9
Years of Education (mean ± SD) 7.8 ± 5.7
Marital Status- % married 152/495 (31%)
Age of First Sexual Intercourse (mean ± SD in years) 19.1 ± 5.7
Number of Lifetime Sexual Partners (mean ± SD) 2.7 ± 4.0
Intravenous Drug Use- % with any history of usage 12/498 (2%)
Intranasal Cocaine Use- % with any history of usage 12/496 (2%)
Subjects with tattoos 1/498
The prevalence of positive HCV antibodies was 22/500 (4.4%, 95% CI 2.6–6.2%). Subjects that were anti-HCV positive had an average of 7 ± 8.6 lifetime sexual partners, compared to average of 2.5 ± 3.5 lifetime sexual partners among HCV-negative subjects (p = 0.02). There were no other statistically significant differences between the HCV antibody-positive and HCV antibody-negative groups in terms of demographic characteristics or topic areas associated with HCV infection/transmission. Among subjects with HCV antibodies, 4/22 subjects admitted intravenous drug use compared to 8/476 among HCV-antibody negative subjects (p < 0.001). A similar result was observed for rates of intranasal cocaine use among HCV-antibody positive (4/22) vs. HCV-antibody negative (8/474) subjects (p < 0.001). Of the 22 subjects with positive HCV antibodies, 12 subjects denied intravenous drug use, had no tattoos, and had either 1 or 2 lifetime sexual partners. Five subjects with positive HCV antibodies denied intravenous drug use and had >10 lifetime sexual partners.
Intravenous drug use (r = 0.26, p < 0.001), intranasal cocaine use (r = 0.29, p < 0.001) and the number of lifetime sexual partners (r = 0.24, p < 0.001) were the only three variables with a statistically significant correlation with the presence of HCV antibodies on univariate analysis. The number of lifetime sexual partners had some correlation with intravenous drug use (r = 0.12, p = 0.009) and intranasal cocaine use (r = 0.13, p = 0.004). There was a very close correlation between intravenous drug use and intranasal cocaine use (r = 0.99, p < 0.001). Therefore, intranasal cocaine use was not incorporated into the multivariate model due to concerns about co-linearity. In the multiple logistic regression model with backwards elimination, intravenous drug use (OR 3.7, 1.52–9.03 95% CI) and number of sexual partners (OR 1.1, 1.04–1.20 95% CI) were independently associated with a positive HCV antibody result.
Discussion
We observed a prevalence of 4.4% of HCV antibodies in an urban population in Haiti. Additionally, we found that HCV antibodies were associated with intravenous drug use as would be expected, but also with increasing number of lifetime sexual partners. It is conceivable that subjects who reported higher number of sexual partners were more likely not to admit past intravenous drug usage. However, it is also possible that this finding represents sexual transmission of hepatitis C among subjects with multiple prior sexual partners. Finally, we observed HCV antibodies were present is some subjects who denied intravenous drug use and had less than 3 lifetime sexual partners.
A recent survey of the influences on HIV preventive behaviors among youth in Haiti observed that 80% of males and 42% of females self-disclosed sexual activity [6]. This survey noted a mean age of first intercourse of 13.1 years, a number lower than our observations (19.1 years). This study also noted that condom use was infrequent in the surveyed population (18% of subjects reported always or sometimes using a condom), while 43% reported 3 or more lifetime sexual partners. The applicability of these data to our study population is limited because the age difference between this population and the population of our study. Although the results of this investigation do not directly describe the population of our study, we might extrapolate that our population would be unlikely to have a high rate of condom usage.
Only a single prior study in Haiti suggested that HCV antibodies were rare [2]. The most obvious difference between our results and this prior study was that the previous study was performed on rural subjects, while our data was collected in an urban setting. It is possible that intravenous drug abuse would be more prevalent in an urban environment, which could partially explain the observed differences in seroprevalence. The previous study was also conducted more than 10 years before our research.
Intravenous drug use does not exclusively explain the prevalence of HCV antibodies in the studied population. Although intravenous drug use was associated with an increased likelihood of having HCV antibodies (as has been thoroughly documented [1]), only four of the 22 subjects with HCV antibodies admitted drug use. Additionally, none of the HCV positive subjects had tattoos. These findings suggest either that self-reported drug use underestimates the prevalence of drug use in the population, or another mode of transmission of HCV has occurred.
The degree to which sexual transmission of HCV occurs is exceedingly controversial [5,7]. Studies in monogamous relationships suggest that sexual transmission of hepatitis C occurs very rarely [8]. Seroprevalence studies from sexually-transmitted diseases clinics describe a variable amount of HCV positive subjects, which tends to be low when injection drug users are excluded [9,10]. The number of sexual partners has been previously associated with increasing risk of HCV exposure [11]. We also observed the association between increasing number of sexual partners and the likelihood of having HCV antibodies. This association was observed even when controlling for other variables in a multivariate model. Our data suggest that sexual transmission of hepatitis C may occur more frequently in persons with multiple sexual partners. However, additional larger studies directed at evaluating HCV-infected persons with multiple sexual partners are needed. All studies on the sexual transmission of hepatitis C (including our study) are limited by the potential of the confounding variable of shared toothbrushes, razor blades and other items among sexual partners.
Our findings are limited by the lack of information regarding active HCV infection. The presence of HCV antibodies only indicates prior exposure, and definitive documentation of active HCV infection requires detection of virus in the bloodstream utilizing HCV RNA polymerase chain reaction testing. However, since most patients exposed to hepatitis C develop chronic infection [12], HCV antibody testing provides a reasonable estimate of the amount of HCV infection in a population. Other limitations of the study are linked to the difficulties inherent in self-reporting of behaviors such as sexual activity and drug use. Additionally, we were unable to obtain information on subjects that refused participation in the study, which may limit the representativeness of this population. It is also possible that the patients receiving blood draws in the Cap-Haïtien health clinic may not be representative of the population of Haiti, or even the urban population of Haiti. Finally, there are other possible alternative sources of percutaneous exposure to HCV, such as medical injections by alternative practitioners or other medical, surgical or dental procedures.
In conclusion, we provide seroprevalence data of HCV antibodies in an urban population in Haiti. These data are useful for understanding the risks of transfusion in Haiti if the blood has not been previously screened for HCV. This information contributes to our understanding of the worldwide prevalence of hepatitis C, which allows for informed decisions regarding the priorities of funding for the treatment and prevention of this infection. Additionally, we observed the number of sexual partners may be related to a greater likelihood of having HCV antibodies, but some subjects who denied intravenous drug use and had few lifetime sexual partners still had HCV antibodies. These findings could be utilized to foster consideration of new studies into some of the risk factors that are not clearly understood (such as procedures by medical, dental or alternate practitioners). These results suggest that further study into the mode of transmission of hepatitis C should focus on patients with a high number of lifetime sexual partners but no evidence of intravenous drug use.
Competing interests
Matthew J. Hepburn, MD: no competing interests to declare.
Eric J. Lawitz, MD: Dr. Lawitz has received research grants to conduct investigator-initiated research from Schering-Plough Corporation (Kenilworth, New Jersey).
Author's contributions
MH was involved in study design, data analysis, and manuscript preparation. EL was involved in study design, data collection, data analysis, and manuscript preparation.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors are very grateful for the assistance of Bernes E. Chalumeau MD, from the Hôpital Universitaire Justinien, Cap-Haïtien, Haiti, who provided crucial assistance with the data collection portion of this study. The funding source for this study was an unrestricted grant, Schering-Plough Corporation (Kenilworth, New Jersey). 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.
==== Refs
Memon MI Memon MA Hepatitis C: an epidemiological review J Viral Hepat 2002 9 84 100 11876790 10.1046/j.1365-2893.2002.00329.x
Allain JP Hodges W Einstein MH Geisler J Neilly C Delaney S Hodges B Lee H Antibody to HIV-1, HTLV-1, and HCV in three populations of rural Haitians J Acquir Immune Defic Syndr 1992 5 1230 1236 1333530
Smikle M Dowe G Hylton-Kong T Hepatitis B and C viruses and sexually transmitted disease patients in Jamaica Sex Transm Infect 2001 77 295 296 11463935 10.1136/sti.77.4.295
Wharfe G Smikle M Dowe G Buchner L Choo-Kang E Graham S King D Seroprevalence of hepatitis C virus in haemophiliacs in Jamaica Hum Antibodies 2002 11 61 4 12454365
Brook MG Sexually acquired hepatitis Sex Transm Infect 2002 78 235 40 12181458 10.1136/sti.78.4.235
Holschneider SO Alexander CS Social and psychological influences on HIV preventive behaviors of youth in Haiti J Adolesc Health 2003 33 31 40 12834995 10.1016/S1054-139X(02)00418-4
Terralt NA Sexual activity as a risk factor for hepatitis C Hepatology 2002 36 S99 S105 12407582 10.1053/jhep.2002.36797
Wejstal R Sexual transmission of hepatitis C virus J Hepatol 1999 31 92 5 10622568 10.1016/S0168-8278(99)80382-1
Balogun MA Ramsay ME Parry JV Donovan L Andrews NJ Newham JA McGarrigle C Harris KA Teo CG A national survey of genitourinary medicine clinic attenders provides little evidence of sexual transmission of hepatitis C virus infection Sex Transm Infect 2003 79 301 306 12902580 10.1136/sti.79.4.301
Thomas DL Zenilman JM Alter HJ Shih JW Galai N Carella AV Quinn TC Sexual transmission of hepatitis C virus among patients attending sexually transmitted diseases clinics in Baltimore. An analysis of 309 sexual partnerships J Infect Dis 1995 171 768 775 7535827
Daikos GL Lai S Fischl MA Hepatitis C virus infection in a sexually active inner city population. The potential for heterosexual transmission Infection 1994 22 72 76 8070933
National Institutes of Health Consensus Development Conference Panel Statement: Management of Hepatitis C: 2002 – June 10–12, 2002 Hepatology 2002 36 S3 S15 12407572 10.1053/jhep.2002.37117
| 15596018 | PMC539356 | CC BY | 2021-01-04 16:29:55 | no | BMC Gastroenterol. 2004 Dec 14; 4:31 | utf-8 | BMC Gastroenterol | 2,004 | 10.1186/1471-230X-4-31 | oa_comm |
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-5-951559834210.1186/1471-2164-5-95Research ArticleHomopolymer tract length dependent enrichments in functional regions of 27 eukaryotes and their novel dependence on the organism DNA (G+C)% composition Zhou Yue [email protected] Jeffrey W [email protected] Kenneth A [email protected] Division of Endocrinology, Gerontology, and Metabolism, Stanford University School of Medicine, Stanford, CA, USA2 Center for Intelligent Biomaterials, Department of Chemistry, University of Massachusetts, Lowell, MA, USA2004 14 12 2004 5 95 95 22 2 2004 14 12 2004 Copyright © 2004 Zhou et al; licensee BioMed Central Ltd.2004Zhou 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
DNA homopolymer tracts, poly(dA).poly(dT) and poly(dG).poly(dC), are the simplest of simple sequence repeats. Homopolymer tracts have been systematically examined in the coding, intron and flanking regions of a limited number of eukaryotes. As the number of DNA sequences publicly available increases, the representation (over and under) of homopolymer tracts of different lengths in these regions of different genomes can be compared.
Results
We carried out a survey of the extent of homopolymer tract over-representation (enrichment) and over-proportional length distribution (above expected length) primarily in the single gene documents, but including some whole chromosomes of 27 eukaryotics across the (G+C)% composition range from 20 – 60%. A total of 5.2 × 107 bases from 15,560 cleaned (redundancy removed) sequence documents were analyzed. Calculated frequencies of non-overlapping long homopolymer tracts were found over-represented in non-coding sequences of eukaryotes. Long poly(dA).poly(dT) tracts demonstrated an exponential increase with tract length compared to predicted frequencies. A novel negative slope was observed for all eukaryotes between their (G+C)% composition and the threshold length N where poly(dA).poly(dT) tracts exhibited over-representation and a corresponding positive slope was observed for poly(dG).poly(dC) tracts. Tract size thresholds where over-representation of tracts in different eukaryotes began to occur was between 4 – 11 bp depending upon the organism (G+C)% composition. The higher the GC%, the lower the threshold N value was for poly(dA).poly(dT) tracts, meaning that the over-representation happens at relatively lower tract length in more GC-rich surrounding sequence. We also observed a novel relationship between the highest over-representations, as well as lengths of homopolymer tracts in excess of their random occurrence expected maximum lengths.
Conclusions
We discuss how our novel tract over-representation observations can be accounted for by a few models. A likely model for poly(dA).poly(dT) tract over-representation involves the known insertion into genomes of DNA synthesized from retroviral mRNAs containing 3' polyA tails. A proposed model that can account for a number of our observed results, concerns the origin of the isochore nature of eukaryotic genomes via a non-equilibrium GC% dependent mutation rate mechanism. Our data also suggest that tract lengthening via slip strand replication is not governed by a simple thermodynamic loop energy model.
==== Body
Background
DNA homopolymer tracts are the simplest of simple sequence repeats (SSRs); the two types being poly(dA).poly(dT) and poly(dG).poly(dC). They are present in all genomes, but in some eukaryotes they are found at high frequencies indicating that the tracts are highly enriched relative to their random occurrence within a random sequence DNA genome of similar base composition. Homopolymer tracts were previously examined systematically in the coding, intron and flanking regions of the slime mold D. discoideum [1]. Only long (N>10 bp) homopolymer tracts were observed at high frequencies in this AT-rich genome. The non-coding regions were found to be highly over-represented in the large poly(dA).poly(dT) tracts compared to random sequences of equivalent base composition containing tracts at frequencies expected for random occurrence. At shorter sequence lengths (2 bp<N<6 bp), poly(dG).poly(dC) tracts were over-represented somewhat more than poly(dA).poly(dT) tracts of comparable length.
Although the elongation of SSR tracts may be due to more than one mechanism [2], most often the phenomenon has been attributed to slip-strand replication errors, which occur from the slippage and re-annealing of the nascent strand during DNA replication [3-5]. This is a type of mutation that can be affected by the proofreading function of DNA polymerases [6-8]. For example, it has been shown that the proofreading and repair function for DNA polymerase epsilon is efficient for short homopolymer tracts, but that only the mismatch repair system can prevent frameshift mutations in tracts of length 8 nucleotides or greater [8]. These slip-strand errors, which lead to the formation of longer homopolymer tracts, can have deleterious effects. In coding regions, the mutation can cause frame-shift errors, leading to transcription errors and aberrant protein translation. As would be expected from the triplet codon constraints, there appears to be selection against long homopolymer tracts in coding regions. Marx et al [1] demonstrated that long homopolymer tracts were not present at frequencies higher than expected in the coding regions of D. discoideum DNA.
Compared to nonhomopolymer random B-DNA sequences, poly(dA).poly(dT) tracts have a shorter turn, a smaller axial rise, a narrower and deeper minor groove [9], a wider and shallower major groove, and are straighter and more rigid over longer lengths. These characteristics are due to the high propeller twist of the base pairs [10], the maximal overlap/stacking between bases on the same strand, and non-Watson-Crick cross-strand H-bonds between base pairs [9-13]. The result is that long tracts tend to be energetically excluded from nucleosomes [14,15]. In some definitive studies, it was first shown that tracts of critical and longer lengths are excluded from the reconstituted nucleosome [16]. Other investigators, using native nucleosomes derived from native chicken chromatin, demonstrated that long poly(dA).poly(dT) tracts are excluded from the central core regions [17]. This conclusion received confirmation from a study of the GenBank sequences of D. discoideum, where it was demonstrated that long poly(dA).poly(dT) tracts (N>10 bp) are preferentially spaced at sequence lengths corresponding to the average nucleosome DNA spacing in D. discoideum nucleosomes [18]. In this study, adjacent long poly(dA).poly(dT) tracts, combined with their short nonhomopolymer spacer sequences, exhibited average total lengths that correspond to D. discoideum nucleosomal linker lengths, suggesting their in vivo localization in these chromatin regions and avoidance of the nucleosomal core regions.
Certain natural DNA sequences possess tertiary structures exhibiting a significant amount of curvature that is associated with short homopolymer lengths (4 – 6 bp) of poly(dA).poly(dT) [19,20]. Also, bending occurs at the junction of these and nonhomopolymer tracts [12]. When short bent poly(dA).poly(dT) tracts are distributed 10 bp apart, they produce additive long range in-plane bending in the axis of the DNA helix [20,21]. The exact molecular mechanism of this bending behavior is still the subject of considerable experimentation and speculation [22]. DNA bending patterns resulting from spaced poly(dA).poly(dT) tracts have been shown to occur in replication origins and in transcriptional regulatory regions, where a bent configuration is required for activity [23-25].
Poly(dG).poly(dC) tracts form an A-form double-helix. In contrast to poly(dA).poly(dT), the minor groove of these tracts is broad and shallow, while the major groove is deep. But, as with poly(dA).poly(dT) tracts, the tracts are rigid, which leads to the energetic exclusion of poly(dG).poly(dC) from nucleosomes [26,27]. These characteristics are due to the overlap of adjacent guanine bases and the invariant roll angle between them [28].
Beyond their structural properties, studies of homopolymer tracts have revealed some biological functions. The poly(dA).poly(dT) tracts can serve as protein binding sites [29], particularly as upstream promoter elements in the initiation of transcription [30-32] and in recombination [33]. And poly(dG).poly(dC) tracts have been found in certain eukaryotic promoter regions where they are postulated to form 4 stranded G-quartet structures [34].
As the number of DNA sequences publicly available increases, the representation (over and under) of homopolymer tracts in different genomes can be compared. Qualitative comparisons have been made between five eukaryotic and two prokaryotic genomes: P. falciparum (also very AT-rich), H. sapiens, S. cerevisiae, C. elegans, A. thaliana, E. coli and M. tuberculosis [35]. As with D. discoideum (1), it was shown in that study that homopolymer tracts occur in the non-coding regions at over-represented frequencies for poly(dA).poly(dT). Poly(dG).poly(dC) tracts were found to be over-represented in some but not all organisms at short lengths. However, over-representation was observed only in the eukaryotic genomes, not the prokaryotes.
In the present study, we have carried out a broad survey of non-overlapping homopolymer tract frequencies in the genomic sequences of 27 eukaryotic organisms across the base composition range from 20–60% (G+C). Within the coding, intron and flanking DNA functional compartments of largely single copy genes from these organisms, we compared the observed poly(dA).poly(dT) and poly(dG).poly(dC) tract frequencies in two size ranges to the tract occurrence frequencies expected for random tract occurrence in DNA compartments of the same base compositions. A large fraction of the 27 eukaryotes exhibited significant over-representations (enrichment) of longer length (N ≥ 9 bp) poly(dA) and poly(dT) tracts in their intron and flanking sequences, but not their coding sequences. This occurred in a novel base composition dependent fashion. A much smaller number of the 27 organisms exhibited significant over-representations of longer length (N ≥ 9) poly(dG) and poly(dC) tracts. For P. falciparum and S. cerevisiae single gene containing sequences as well as whole chromosomes, all homopolymer tracts were found to have similar length dependent frequencies and therefore over-representation in their functional compartments.
Results
The purpose of this study was to reveal similarities and differences in the frequency of occurrence of homopolymer tracts of varying lengths in different eukaryotic sequences across the biological range of base compositions from 20 – 60% (G+C). Primarily single gene containing sequences of 27 organisms were investigated. This was done on purpose since most of the organisms had largely only single gene containing sequences available in the public databases. Restricting our comparative analysis in this fashion ensured that the results for all organisms could be easily compared. A total of 25,109 sequence documents were collected. As we mentioned earlier, there are often a significant level of redundancies in the sequences found in the public databases. CleanUP is a program we used to remove those redundancies, and following its application, 5.2 × 107 bases from 15,560 cleaned sequence documents were analyzed and compared [36]. As shown in Table 1 comparing columns before and after application of CleanUP, there are varying levels of redundancies in some of the original collected files for different organisms in the public database. In Table 1, we list the complete name of each organism as well as an abbreviation that is used throughout the following discussion of our Results.
Table 1 Summary of the sequence files of the 27 organisms studied
ORGANISM ABBR. DNA type DOCUMENTS BEFORE CLEAN-UP DOCUMENTS AFTER CLEAN-UP TOTAL (bp) TOTAL GC%
Dictyostelium discoideum Dd single genes 492 440 966261 25.70
Plasmodium falciparum Pf single genes 1652 790 1065133 27.11
chromosome II, III 2 2 2007209 19.82
Tetrahymena thermophila Tt single genes 109 97 214676 28.89
Candida albicans Ca single genes 439 378 959035 35.21
Manduca sexta Ms single genes 54 46 119657 35.62
Caenorhabditis elegans Ce single genes 234 221 1319875 36.92
Schizosaccharomyces pombe Spo single genes 813 720 1346439 37.54
Arabidopsis thaliana At single genes 1908 1520 3139637 38.09
Schistosoma mansoni Sm single genes 98 74 125605 38.37
Danio rerio Dr single genes 339 266 603534 38.60
Saccharomyces cerevisiae Sc single genes 2249 928 3815906 38.68
chromosome I-XVI 16 16 11426263 38.45
Drosophila melanogaster Dm single genes 1883 968 3459297 39.88
Strongylocentrotus purpuratus Spu single genes 153 113 116647 41.78
Xenopus laevis Xl single genes 568 411 805354 41.99
Oryza sativa Os single genes 605 507 1514656 44.61
Trypanosoma brucei Tb single genes 457 368 1039208 46.17
Fugu rubripes Fr single genes 216 181 1737132 46.30
Zea mays Zm single genes 629 480 1225027 46.92
Mus musculus Mm single genes 9288 5179 11191148 47.36
Anopheles gambiae Ag single genes 78 43 73592 48.20
Gallus gallus Gg single genes 1868 1061 1910331 50.00
Toxoplasma gondii Tg single genes 195 117 284187 50.70
Emericella nidulans En single genes 72 62 165648 51.07
Aspergillus niger An single genes 215 160 386447 52.65
Neurospora crassa Nc single genes 252 217 494046 53.37
Leishmania major Lm single genes 89 81 129155 59.11
Chlamydomonas reinhardtii Cr single genes 136 114 349624 61.84
The base compositions of the total sequence population from each organism ranges from a low of 25.70% (G+C) for D. discoideum (Dd) to a high of 61.84% for C. reinhardtii (Cr). From the standpoint of (G+C)%, the organisms we investigated are not evenly distributed. If we take as the midpoint, G. gallus (Gg), whose (G+C)% is exactly 50%, there are only 6 organisms over 50%, while the rest (19 of 27) are all below 50%. This is due to the fact that there are more available sequenced eukaryotic organisms that are AT-rich than GC-rich. In this study, we dissected DNA sequences into coding, intron and flanking functional compartments as shown in Table 2. In every instance, the non-coding regions (intron and flanking) were found to be significantly more AT-rich than coding sequences.
Table 2 The (G+C)% in different compartments of the 27 organisms
ORGANISM ABBR. FLANK (bp) GC% INTRON (bp) GC% CODING (bp) GC%
Dictyostelium discoideum Dd 207722 15.26 22407 11.14 621748 30.32
Plasmodium falciparum Pf 143401 16.27 18597 12.64 840733 29.41
Tetrahymena thermophila Tt 60095 20.99 8624 20.14 101889 34.37
Candida albicans Ca 302347 31.26 3035 33.05 631769 36.93
Manduca sexta Ms 43331 33.38 17299 30.87 21979 47.79
Caenorhabditis elegans Ce 737951 32.87 121316 33.07 384627 45.45
Schizosaccharomyces pombe Spo 432364 32.67 22638 30.44 813101 40.41
Arabidopsis thaliana At 1256492 33.20 270913 32.05 1015490 45.75
Schistosoma mansoni Sm 37773 36.12 11760 35.94 39913 40.05
Danio rerio Dr 232819 35.07 41465 34.21 147459 47.90
Saccharomyces cerevisiae Sc 1124634 36.06 17273 32.90 2356815 40.14
Drosophila melanogaster Dm 1238086 37.56 227141 39.54 747774 52.85
Strongylocentrotus purpuratus Spu 38732 36.85 11450 34.43 24405 53.22
Xenopus laevis Xl 266034 39.32 183822 38.36 164675 48.08
Oryza sativa Os 567768 40.35 133898 36.74 394283 55.89
Trypanosoma brucei Tb 120949 43.19 1621 44.97 415684 50.73
Fugu rubripes Fr 624550 43.85 230616 43.22 342435 54.13
Zea mays Zm 611822 43.30 114505 41.21 305091 55.88
Mus musculus Mm 4571933 46.05 1708894 46.75 1475920 53.15
Anopheles gambiae Ag 29527 42.93 8028 44.33 24151 56.93
Gallus gallus Gg 563319 50.08 457259 46.15 369315 55.13
Toxoplasma gondii Tg 100004 50.01 32167 49.49 94260 54.51
Emericella nidulans En 75257 48.72 2888 47.13 83665 53.12
Aspergillus niger An 136216 48.46 15832 45.95 188982 56.48
Neurospora crassa Nc 154422 50.44 18685 48.88 254177 56.07
Leishmania major Lm 34029 57.27 54 50.00 70014 60.71
Chlamydomonas reinhardtii Cr 163614 59.70 49607 62.03 106939 66.55
Long homopolymer tracts are over-represented in non-coding sequences of AT-rich eukaryotes
The observed frequencies of non-overlapping base i tracts of length N, , in different DNA regions were analyzed as a function of tract length N in all 27 organisms. In Figure 1, we present the results of analyzing all 4 base tracts from Pf, Dm and Nc sequences as representative examples. The total (G+C)% of the DNA analysed from these organisms is 27.11%, 39.88% and 53.37%, respectively, representing typical low, median and high base composition eukaryotes. Very long tracts (high N) are rare, leading to low counts and large fluctuations in log (). Therefore, for each organism, any tract count observed to be less than 4 for a given tract length N was excluded from the data analysis in order to eliminate noise in the data. Getting rid of the interference caused by noisy data enhanced and clarified the comparisons we made from slope determinations. Some of the points in Figure 1 are not connected because they did not present contiguous data along the x-axis (tract length N).
Figure 1 Comparison of the observed length N dependent poly(dA), poly(dT), poly(dC) and poly(dG) tract frequencies found in sequences from different DNA functional regions from the organisms Pf, Dm and Nc A. flanking sequences; B. intron sequences; C. coding sequences
From Figure 1A, it is clear that the frequency of poly(dA).poly(dT) tracts in the very AT-rich Pf flanking regions are much higher than Dm and Nc tract frequencies. This becomes more pronounced when the tract length N becomes larger (N ≥ 7 bp) and reaches a maximum at around N ≥ 10 bp. On the other hand, the frequencies of poly(dG) or poly(dC) tracts in Pf are significantly lower and no differential N dependence is observed. Meanwhile, no significant difference can be observed between Dm and Nc, except for the higher values at longer N in Nc sequences. Similar behavior is also observed in intron sequences in Figure 1B. The over-representation of poly(dA) or poly(dT) tracts in intron regions is also evident at higher N values. However, this behavior is almost non-existent in coding regions (Figure 1C), except for Pf, which is the most AT-rich eukaryote in all the 27 of our survey. This organism exhibited over-representation of poly(dA) tracts in its coding region as well as poly(dT) tracts as we shall see below.
It is also very clear that the longer poly(dA) and poly(dT) tracts, usually of length larger than 20 bp, are only detected in flanking regions. In all cases, the plotted curves exhibited a transition region of changing slope, points falling between tract lengths 6 bp and 9 bp. This behavior, as we have described previously in D. discoideum DNA, leads one to conclude that the long poly(dA) and poly(dT) tracts are over-represented relative to random tract occurrence in random DNA sequences of equivalent base composition [1]. This is a fact that we illustrate and quantitate later in this study. By contrast, the nearly linear relationship of points in Figure 1, for all the organisms' tracts of all types at lengths N ≤ 6 bp indicates a similarity that differs for each organism only by the individual linear relationships being offset from each other. This is a trivial consequence of the different base compositions of the DNAs, giving rise to frequencies of any given tract at levels near those expected based on random occurrence in that base composition.
Comparing tract frequencies from single genes with those from whole chromosomes
In order to confirm that there is consistency in the homopolymer tract frequency levels between single gene data and whole chromosome data in any given organism, we collected single gene data and whole chromosome data separately for representative organisms – Pf and Sc, where fully annotated whole chromosome sequences were available. For Pf, we collected chromosomes II and III. For Sc, we collected sequences for all 16 chromosomes. The single gene data were compared to the chromosome data of each organism respectively. For both organisms, the comparison results are similar and, therefore, we only display in Figure 2 representative data here for Sc single gene data compared with Sc chromosome IV, the largest of the 16 chromosomes. Since the whole chromosome data is only annotated with coding and non-coding regions, we combined the results from single gene data for intron and flanking regions, which were previously separated in our Figure 1 analysis, in order to make a consistent comparison with the whole chromosome data. Aside from poly(dT) tracts in coding sequences, the analyses showed no consistent significant differences. Therefore, we judge that our conclusions using single gene data are representative of whole chromosome data for the 27 eukaryotes we analyzed in this survey.
Figure 2 Comparison of the four tract frequencies from Sc chromosome 4 sequences and Sc single gene sequences as a function of N, the tract length, calculated from: A. coding sequences; B. non-coding sequences. In the legend, "sg" represents "single gene" and "chr" represents "chromosome".
Quantitating the over-representation of tracts
We next wished to quantitatively compare for all 27 eukaryotes, the differences between the length N dependent frequencies of short tracts (N ≤ 6 bp) and long tracts (N ≥ 9 bp) of the type that we presented in Figure 1. We designed the data analysis method by separating the Figure 1 x-axis into two regions of different tract behavior: N ≤ 6 bp and N ≥ 9 bp. The data points in the short tract range and those in the long tract range were treated separately. For short and long tract point regions separately, the average frequency, fslope, of the tract base i in the particular genome compartment were calculated. The fslope parameter is the "effective" base i frequency for the sequences in that DNA compartment that would give rise to the observed log () vs. N dependent tracts frequency behavior in that region based upon the eqn. [1b] random model. The fslope is obtained from the inverse of P', where the slope [-log(P')] is obtained from the log () vs. N type plots in Figure 1, fit by eqn. [1b], as we present in Methods and have previously described [1]. This is a model that assumes random occurrence of tracts. Although it is known that DNA sequences do not occur randomly and that 1st Order Markov chain behavior can describe some of the behavior of eukaryotic sequences, we have chosen here to compare the occurrence of tracts in real sequences to that of tracts in random DNA of equivalent base composition because the comparison is intuitively easy to grasp. The results from all 27 organisms are presented here in Figures 3, 4 and 5 for flanking, intron and coding sequences respectively. The data are plotted as frequency ratios (fslope/fseq: where fseq is the actual base frequency tabulated from all bases comprising the sequences in the real sequence compartment) plotted versus the overall real (G+C)% of each individual DNA sequence compartment studied. The higher the frequency ratio in Figures 3, 4 and 5, the higher is the enrichment or over-representation of the tract. It is a common feature for all the organisms that when N ≤ 6 bp, the ratio is near 1. Therefore, in all the Figures 3,4,5, the trend lines developed by the linear regression fit of only the N ≤ 6 bp data have slopes close to zero and intersect the y-axis at a ratio near 1 to 2. The regression lines extrapolating to a ratio near 1.0 (Figure 3A, Figure 4A and Figure 5A) indicate that N ≤ 6 bp tracts occur at frequencies expected for the base compositions found in each of the organisms' sequence compartments. Interestingly, this behavior occurs for the poly(dA).poly(dT) tracts in all three functional compartments – coding, intron and flanking DNAs. On the other hand, for N ≤ 6 bp poly(dG).poly(dC) tracts of all organisms, the regression lines all have a slightly negative slope, with flanking and intron sequences (Figure 3B and Figure 4B, respectively) exhibiting an intercept ratio of 2 or greater. This clearly indicates a trend to greater over-representation of short poly(dG).poly(dC) tracts in organisms of higher (A+T)% base composition.
Figure 3 Comparison of the frequency ratio, fslope/fseq, to the real (G+C)% of the particular organisms' flanking DNA. The fslope is calculated from the slopes of Figure 1 types of graphs for short (N ≤ 6 bp) and long (N ≥ 9 bp) tract data found in flanking sequences from 27 organisms. A. poly(dA).poly(dT) tracts. The straight solid lines are linear regression fits for short tracts (N ≤ 6 bp) of each type and the dashed line (R2 = 0.5591) demonstrates the trend in long (N ≥ 9 bp) tracts; B. poly(dG).poly(dC) tracts. The straight lines are linear regression fits for short tracts of each type (N ≤ 6 bp).
Figure 4 Comparison of the frequency ratio, fslope/fseq, to the real (G+C)% of the particular organisms' intron DNA. The fslope is calculated from the slopes of Figure 1 types of graphs for short (N ≤ 6 bp) and long (N ≥ 9 bp) tract data found in intron sequences from 27 organisms. A. poly(dA).poly(dT) tracts. The straight lines are linear regression fits for short tracts (N ≤ 6 bp) of each type and the dashed line (R2 = 0.5474) demonstrates the trend in long (N ≥ 9 bp) tracts; B. poly(dG).poly(dC) tracts. The straight lines are linear regression fits for short tracts of each type (N ≤ 6 bp).
Figure 5 Comparison of the frequency ratio, fslope/fseq to the real (G+C)% of the particular organisms' coding DNA. The fslope is calculated from the slopes of Figure 1 types of graphs for short (N ≤ 6 bp) and long (N ≥ 9 bp) tract data found in coding sequences from 27 organisms, A. poly(dA).poly(dT) tracts. The straight lines are linear regression fits for short tracts (N ≤ 6 bp) of each type; B. poly(dG).poly(dC) tracts. The straight lines are linear regression fits for short tracts (N ≤ 6 bp) of each type.
For N ≥ 9 bp tracts in coding sequences (Figure 5A &5B), there were not enough tracts to allow calculation of fslope values from the Figure 1 type data. However, in flanking and intron sequences (Figure 3A and Figure 4A), the ratio was determined and is much higher than 1 for all the organisms, indicating that the poly(dA).poly(dT) tracts are significantly over-represented. The behavior of poly(dA).poly(dT) tracts in both flanking and intron sequences are similar and demonstrate a novel and interesting dependence of over-representation on the base composition of the organism's DNA. Starting at 30 %(G+C) and increasing to 50 %(G+C) (note linear fit trend line), these tracts are increasingly over-represented as the ratio trends from around 1.5 up to 4.0.
For lengths N ≥ 9 bp in Figures 3B and 4B, we observed similar behavior for poly(dG).poly(dC) tracts. In a few organisms between 35%–50% (G+C), there is a high ratio of frequencies indicating a high level of over-representation. However, longer poly(dG).poly(dC) tracts are not over-represented and do not occur at long lengths as we show later in nearly as many organisms as we observed for poly(dA).poly(dT) tracts. Therefore, even though there appears to be evidence for a trend in these figures, we have not indicated with a negative slope linear trend line, mirror-image trend behavior to that exhibited by the poly(dA).poly(dT) tracts.
The over-representation of long poly(dA).poly(dT) tracts exhibit exponential frequency increases compared to predicted values
In order to show more clearly the genomic over-representation of the long poly(dA) and poly(dT) tracts, we introduced a variable, , representing the predicted frequency of base i at length N based on random tract occurrence in DNA of equivalent base composition. Eqn. [2] is used to calculate . The ratio of equals R, the Threshold. In Figure 6, we plot log R vs. N for only the poly(dA) tract data from Dd, Os, Cr, representing genomes of low, median and high (G+C)% base composition, respectively. Similar comparative data for all 27 organisms was determined but is not shown here. For comparison purposes, we include in Figure 6 the tract frequency results determined for a random sequence of 106 nucleotides of 50% (G+C) composition generated with a random number generator. Each base position in that random sequence was picked from all 4 bases having equal probability (0.25) of being selected. In the randomly generated sequence, there were no tracts longer than 9 bp. The small inset panel in the upper left corner of the figure presents an enlarged view of N from 0 to 10. In this panel, as expected, the random sequence exhibits points with values closely centered around 0 on the y-axis. The only exceptions are for the points N = 6 and higher that are noisy and exhibit fluctuations as high as 0.15, due to the low number of tracts occurring at those sizes. Thus, there are no significant differences between and and . In contrast, the results from real organisms show very different behaviors. There are two regions, a linear part with slope around 0 when N is relatively small and an exponentially increasing frequency ratio when N increases beyond a certain value. In Figure 6, two of the three organisms exhibit this exponentially increasing ratio. Os is the first to go above the 0 line when N is around 3 bp. A similar change to an exponentially increasing ratio was observed for Dd at a different N value of about 7 bp. This is the same tract size where we previously observed that poly(dA) and poly(dT) tracts begin to exhibit over-representation [1]. Of all the organisms, the Dd data exhibited one of the most significant over-representation levels- a 1013-fold enrichment of these tracts occurring at lengths about 45 bp.
Figure 6 Comparison of fobs vs. fexp calculated for eukaryotes Dd, Os, Cr and a randomly generated sequence as a function of the tract length N. The and values for D. discoideum are presented Due to the small differences exhibited between the organisms when N is small, we present the small inset figure for the region from N = 0–10 bp enlarged for clarity.
Another significant feature of the Dd data, is the large difference between and . This indicates that a high over-representation of long poly(dA).poly(dT) tracts likely occurs in this organism, as we saw in Figure 2. It also makes clear that this organism utilizes poly(dA).poly(dT) tracts to sizes at least 40 bp longer in length than would be expected, = 14 bp, based upon the random tract occurrence calculated from its base composition. It is also the highest over-proportional tract size we observed, as we present later.
Poly(dA).poly(dT) tracts show inverse correlation between (G+C)% composition and threshold value
The concept of a threshold value was introduced to provide a description of the N dependence of the observed frequency of the tracts, , as it begins to rise significantly above that of the calculated . The threshold is a particular value of the log(R), where R is defined by eqn.3. In this study, we chose not to attempt a universal definition of the over-representation criterion. Rather, we decided to examine various thresholds that defined different over-representations beginning at 0.3, where the is 2 times greater than the , with increasing values up to 1.0, where the is exactly 10 times that of the . We present these data in Figure 7A for poly(dA) tracts, where the N values achieved at the different thresholds are plotted vs. the (G+C)% composition of the DNA. A negative slope between (G+C)% and N at threshold for poly(dA) tracts in flanking sequences was observed for all thresholds. Poly(dT) tracts displayed similar behavior (data not shown). The changing slopes of the linear correlation lines shown in Figure 7A exhibit a progression from highest negative slope at threshold 0.3 to lower negative slope at threshold of 1.0. Thus, the lower the (G+C)% base composition of the genome, the higher the N at which over-representation of poly(dA) tracts occurs. For poly(dC) and poly(dG) tracts in flanking sequences, a positive linear correlation between the (G+C)% base composition and N at threshold was observed (data not shown). Interestingly, the slopes of the correlation lines also resulted in a progression of slope values. Thus, for poly(dC) tracts in flanking sequences, the lower the genome (G+C)% base composition, the lower the N at which over-representation occurs.
Figure 7 The (G+C)% dependence of a series of calculated threshold values for enrichment of each homopolymer tract type In panel A. data is presented for the length N observed at the given series of threshold values for poly(dA) tracts from all 27 organisms. Slopes determined for each threshold from the type of representative data presented in A. were then calculated from all 27 organism to provide the values for poly(dG).poly(dC) and poly(dA).poly(dT) tracts within: B. coding; C. intron; D. flanking regions. The legend in panel D applies as well to panels B and C.
For homopolymer tracts of each type in coding, intron and flanking DNAs, data of the type shown for poly(dA) tracts in Figure 7A were calculated and the linear fit slopes are presented in Figure 7 panels B-D, respectively. For all the data, the poly(dA).poly(dT) tracts exhibit uniformly negative slopes between -4 and -10, while the poly(dG).poly(dC) tracts all exhibit positive slopes between 4.5 and 7. In coding sequences, poly(dA) tracts showed a sharp drop between 0.7 and 0.8 while poly(dT) tracts exhibited a slow increase. No poly(dG) or poly(dC) tracts of significant length occur in coding DNA, which did not allow over-representation to be exhibited at these threshold values. Therefore, no slope points are shown. For intron sequences in panel C, poly(dG).poly(dC) tracts exhibited no significant consistent slope trend. However, poly(dC) tracts have overall greater slopes than poly(dG) tracts. Likewise for poly(dA).poly(dT) tracts, no trend is evident but the latter possesses significantly greater slopes than the former. For flanking sequences, all four tract types exhibited increasing slopes as threshold values increased. As was true for intron sequences, poly(dC) again had overall somewhat greater slopes than poly(dG) tracts. Similar behavior was observed for poly(dA).poly(dT) tracts in both intron and flanking DNA sequence types, with poly(dA) again occurring at greater slopes than poly(dT).
The highest over-representation and over-proportional length of homopolymer tracts appear in median GC% organisms
We next used the proportion, P, eqn.5 for all the 4 homopolymer tracts to compare the maximum observed tract size with the maximum tract size expected for random tract occurrence within that (G+C)% base composition DNA. If the P quantity is greater than 1, tracts are over-proportional in length and if P is less than 1, tracts are under-proportional in length. We present P, / , for coding, intron and flanking DNAs from all 27 organisms in Figure 8 panels A-C, respectively. The values are calculated from eqn.4. Large differences are obvious between coding and non-coding sequences. It is clear that tracts in coding regions, being mostly less than 1, are under-proportional in length for all base types. However, poly(dA).poly(dT) tracts are slightly under-proportional in length in GC-rich organisms, a fact that agrees with our previous observation of over-representation in tract frequencies in Figure 5.
Figure 8 The relationship of the (G+C)% of the DNA analyzed to the calculated / (P) for all the sequences of 27 organisms A. coding; B. intron; C. flanking.
By contrast, the average behavior of intron and flanking regions in Figure 8 is that tracts of all types, but especially poly(dA).poly(dT) tracts, occur at significantly over-proportional lengths. This fact is consistent with their significant over-representation levels that we previously presented in Figures 3,4,5. Very long poly(dA).poly(dT) tracts are observed in non-coding regions of some organisms, at lengths greater than 20 bp in excess of the expected length. Interestingly, the highest over-representation levels of tracts are found in organisms between 30% – 50 % (G+C) base composition. The only exception to this was found in Dd, the most AT-rich organism we studied, where the longest poly(dA) tracts were 71 bp. Higher poly(dG).poly(dC) tract frequencies than expected for organisms greater than 40 % (G+C) base composition were observed in Figure 3B and Figure 4B. The same was true for the most AT-rich ones – Dd and Pf. Figure 8 panels B and C correspondingly exhibit significant levels of over-proportional lengths of poly(dG).poly(dC) tracts, consistent with over-representation, for organisms greater than 30% (G+C) base composition and exhibit moderate over-representation of poly(dG).poly(dC) tracts for Dd and Pf.
Discussion
As a result of recent progress in the rate of DNA sequencing, the amount of sequenced DNA from many organisms has grown significantly. This has allowed our systematic study of the behavior of non-overlapping homopolymer tract frequencies in the 27 eukaryotes in this study spanning the 20% – 60 % (G+C) base composition range. Pre-processing of each of the 27 eukaryotes' largely single gene containing sequence files eliminated sequence redundancies that would introduce bias into the frequency calculations [39-41], that would not be representative of the biological genomes. In most organisms, well over 10% of the documents obtained were judged to contain redundant sequences and were removed by the CleanUP program (Table 1).
From our results in this study, it is clear that long homopolymer tracts are over-represented in non-coding sequences, but not coding sequences, within eukaryotic genomes of all base compositions. This is perhaps not surprising considering that the coding sequence populations must satisfy the constraints of the triplet genetic code. In addition, organisms might minimize the numbers of tracts in coding regions to avoid the severe, even fatal frame-shift mutations that might be introduced by slippage-replication events at tracts [3-5,42]. In nearly all the organisms we studied, poly(dA).poly(dT) tracts were very much over-represented, beginning to be significantly enriched at lengths around 4–10 bp. These tracts also occurred at over-proportional lengths. This was particularly the case for organisms between 30% – 50% (G+C) composition, where over-proportional lengths were pronounced. By contrast, poly(dG).poly(dC) tracts, somewhat over-represented, do not occur at over-proportional lengths. This extends the findings of our previous D. discoideum DNA study that first described the tract over-representation transition region occurring at around 8–10 bp for poly(dA).poly(dT) sequences and their high over-proportional lengths [1]. Somewhat similar observations were made in a subsequent study of five eukaryotic organisms [35]. In general studies of repetitive sequences, poly(dA).poly(dT) tracts have been observed to be over-represented within eukaryotic genomes while poly(dG).poly(dC) tracts are significantly rarer [2,43]. Specific human repetitive sequences, such as the Alu elements, have been shown to contain long poly(dA).poly(dT) tracts, representing a significant repetitive sequence location for some of the over-represented tracts we observed in this study [44].
It has been suggested in a previous study that the over-representation occurring around 7–10 bp represented the minimum thermodynamic length required for any simple sequence repeat such as homopolymer tracts to undergo expansion by slip strand replication [35]. However, in our current study of 27 eukaryotes of widely varying base composition, we present more extensive results, especially those in Figure 7, that demonstrate this is not the case. Depending upon the threshold tract size value chosen to express over-representation of the tracts, the N value where over-representation occurs for A tracts can be seen in Figure 7A to range for all the organisms from as low as 4–6 bp for 0.3 threshold (2× enrichment) to 8–11 bp for 1.0 threshold (10× enrichment). Furthermore, for poly(dA).poly(dT) tracts, the (G+C)% base composition vs. N slopes are negative for all thresholds, while for poly(dG).poly(dC) tracts the slopes are positive for all thresholds. This means that the base composition of the organism is the most important determinant of the particular threshold N value where over-representation begins and argues against an absolute solely thermodynamic determinant to the N value where over-representation begins via slip strand replication. In fact, our observed negative slopes for the poly(dA) tracts in Figure 7A, means that in higher (G+C)% composition organisms, the poly(dA) tracts become enriched at shorter N values than in (G+C)% poor organisms. This result is counter-intuitive to a thermodynamic argument, since the high (G+C)% base composition in neighboring sequences around a short poly(dA) tract in a high (G+C)% organism would be expected to resist the tract looping out to allow for slip strand replication because of the higher level of base stacking stabilization energy in the (G+C)-rich neighboring sequences. We believe that these (G+C)% composition dependent variable threshold N values we observed here are describing a complex mechanism that determines successful tract lengthening, rather than a single thermodynamic criterion for successful DNA looping during slip strand replication.
The reason why poly(dG).poly(dC) tracts occur only at short lengths in eukaryotes may have to do with some interesting structural and energetic polymorphisms of these sequences. Even short tracts of this type have the ability to rearrange from the right-handed double helix to form G-quadraplex structures. These structures have been implicated in biological function in systems as diverse as eukaryotic immunoglobulin switch regions [45], telomeric repeats on chromosome ends [46] and promoter regions [34]. Therefore, eukaryotes may select against these tracts at any significant length in order to minimize problems resulting from the significant structural plasticity of these tracts. Another potential problem with these tracts is the fact that they represent potential reservoirs of oxidative damage. Recently, long-range electron transfer has been demonstrated to occur through the delocalized molecular orbitals of the stacked bases in the DNA double helix [47]. The electron transfer energy in these studies is insensitive to distance along the helix, but is sensitive to the level of base stacking. Therefore, these electron transfer events ultimately cause oxidative damage at GG dinucleotides, a base pair doublet that has high stacking levels. Even greater intensities of photo-damage were observed for GGG triplets. Therefore, eukaryotic organisms have a second compelling reason to mostly avoid the use of these homopolymer tracts at any significant length-a fact reflected in the data we have presented here.
It must be kept in mind in these discussions of homopolymer tract over-representation, that these tracts represent only a subset of the larger sequence class of polypurines and polypyrimidines that exist in and are over-represented within all eukaryotes. In a study of over 700 sequenced chromosomes or long sequences contained in plasmids [48], a bias toward longer polypurine and polypyrimidine tracts in eukaryotes was reported as a function of length N, similar to the homopolymer poly(dA).poly(dT) tract frequency behavior we have reported here.
We have previously observed that the long (N>10 bp) poly(dA).poly(dT) tracts over-represented in D. discoideum DNA (1) were not randomly distributed within the sequences from that organism. In fact, they are arrayed with an average spacing that corresponds to the repeating nucleosome DNA length found experimentally in D. discoideum chromatin [18]. And in that study, adjacent long pairs of tracts plus the intervening non-tract DNA were found to occur within a length corresponding to the internucleosomal linker DNA size found in D. discoideum chromatin. These results suggest that the long tracts only occur in restricted locations in chromatin. This supposition is supported by more recent experimental studies in D. discoideum chromatin compared to calculations of poly(dA).poly(dT) tract spacings in D. discoideum DNA (Marx, K.A., Zhou, Y. and Kishawi, I. unpublished results). That long poly(dA).poly(dT) tracts avoid being located within nucleosome core regions was experimentally determined from sequencing studies of native chicken erythrocyte chromatin [17]. In agreement with this line of reasoning, recent studies have shown that the nucleosome structure readily incorporates DNA containing short tracts, such as the sequence A5TATA4, but longer tracts such as those found in the sequence A15TATA16, completely disrupt the phasing of nucleosomes [49]. Short tracts not only are incorporated into nucleosomes, but they actually represent more stable than average nucleosome positioning sequences when they occur in-phase with the helical turn at roughly every 10 bp [50]. In human NF1-the Alu repeat element is blocked in vitro from forming a nucleosome by the presence of a bipartite T14A11 tract sequence [51].
Different investigators have postulated two additional functions for tracts. The first is their use as promoters. This function may be synergistic with the long poly(dA).poly(dT) tracts preventing the formation of nucleosome structures. The second is as DNA binding sequences for specific poly(dA).poly(dT) tract binding proteins that possess some as yet unknown function. There are a number of reports that poly(dA).poly(dT) tracts function in eukaryotes as promoter sequences. In D. discoideum DNA, the actin genes contain a remarkably long (45 bp) promoter upstream of the TATA box [52]. In this study, the length of the tract was shown to correlate with the transcriptional level of these genes. A number of studies have demonstrated similar long tract promoter activity in yeast promoter regions [32,53,54] and in various mammalian [55] and human promoters [56]. In many of these studies, it was demonstrated that the promoter activity of the long tracts was correlated with this sequence being nucleosome free or not complexed with a protein.
In the case of potential tract function where proteins bind to long poly(dA).poly(dT) tracts, there are a few investigated examples. The small protein datin, 13 kD, has been isolated from S. cerevisiae cells [57]. It has a required tract-binding site that is 9–11 bp in length and its function upon tract binding is unknown. Two high affinity poly(dA).poly(dT) tract-binding proteins, 70 and 74 kD species of unknown function, have been identified in D. discoideum [58]. Another example of a tract binding protein has been discovered in D. discoideum, where some 200 copies of terminal repeat retrotransposons are under transcriptional control by a 134 bp DNA control element [59]. Within this control element, a nuclear protein called CMBF binds to two almost homopolymeric 24 bp poly(dA).poly(dT) sequences. This CMBF protein contains so-called 'A.T hook' regions that interact with a 5–6 contiguous A:T base pair tract. These 'A.T hooks' are found in a number of other (A+T)-rich sequence binding proteins such as HMG-I, DAT1 from yeast, D1 from D. melanogaster and human UBF. In summary, it is unclear what functions these various pure poly(dA).poly(dT) tract binding proteins serve, and how their binding occurs at specific tracts while other tracts remain free of protein. The one point that can be stated with certainty is the correlation between tract binding site size (8–11 bp) of the proteins and the upper threshold (8–11 bp) size where tracts become significantly over-represented or enriched in our study. We believe that this similarity is not coincidental and is a consequence of some functional linkage.
A novel aspect of our study was that for both flanking and intron sequences the over-representation of the poly(dA) and poly(dT) tracts were actually more pronounced in less (A+T)-rich organisms as compared to the most (A+T)-rich, Dd and Pf. Also novel was that poly(dA) and poly(dT) tracts showed negative slopes between the organisms' DNA (G+C)% composition and the threshold value. Thus, the higher the (G+C) base composition, the lower the tract length at which over-representation occurs. In fact, the highest over-representations of homopolymer tracts were observed in median (G+C)% organisms from 30–50%. Also, the distribution was almost symmetric with respect to the different organisms' (G+C)%. We believe that these results could be explained as a result of the insertion of retrotransposon elements into DNA. Eukaryotic transposons are known to be a widely occurring class of repeated DNA sequences ranging in size from about 1 kb to 8 kb. They contain inverted sequence repeats at their termini. The most common transposon class is comprised of retrovirus-like transposons [60], thought to arise from the integration of retroviral RNA sequences into a given eukaryotic genome. The resulting retrotransposon elements do not represent infectious viral DNA and are not transcribed since they lack accompanying promoter sequences. These DNA sequences do possess a poly(dA).poly(dT) tract that resulted from the 3' poly A tail on the original viral mRNA that formed the retrotransposon. Typical retrovirus-like retrotransposons, such as copia in D. melanogaster and IAP in M. musculus, are known to occur in thousands of copies in their respective genomes [60]. Therefore, inserted retrotransposon elements represent the likely origin of the excess over-representation of poly(dA).poly(dT) sequences that we observed in the majority of eukaryotes in this study, irrespective of their overall base composition.
Methods
The single copy gene DNA sequences from 27 eukaryotic organisms were retrieved from GenBank, EMBL, and DDBJ, the members of the tripartite, international collaboration of sequence databases [61]. Every search excluded: ESTs (expressed sequence tag), STSs (sequence-tagged sites), and GSSs (genomic survey sequence), and were limited to organism and genomic DNA only. Moreover, sequences designated: "mitochondrion", "chloroplast", and "chromosome" were also excluded in the search query via these keywords using Boolean operators. In addition, the whole chromosome sequences from 2 of these organisms were also retrieved. The eukaryotic organisms covered are tabulated in Table 1.
The GenBank documents were processed by the program "CleanUP", kindly provided by the Department of Biochemistry and Molecular Biology, University of Bari, Italy [62]. Our purpose in using the program was to get rid of redundancy in our sequence collections so that no bias would be introduced into the homopolymer tract distributions we calculated [36]. The settings for the program are: precision factor (0), different adjacent nucleotides (2), threshold similarity percentage for searching (95.0), overlapping percent for searching (50.0), local similarity percent (70.0), percent ambiguous symbols (e.g. N) to skip matches (10), overlapping percent for cleaning (90.0), minimum length for overlapping (1), minimum length for overlapping segment (20), sequence minimum length so that a sequence is processed, otherwise is cleaned (30). This application of "CleanUP" results in eliminating all the sequences less than 30 bp in length, with more than 20 bp overlapping with the primary sequence (the sequence used use as a basis for comparison), and possessing over 90% similarity with the primary sequence.
Then the redundancy cleaned sequence files were input into the "Compile" program. Compile is part of the "MeltSim" program for the Windows suite of applications [37,63]. This program was used to extract raw sequences from the GenBank-formatted documents. Sequences of the functional categories, coding, intron and flanking were extracted according to their location tags. Respectively, "CDS" is for coding sequences, "intron" is for intron sequences, and "5'UTR", "3'UTR" and any other sequences excluding "CDS" and "intron" are all included into flanking sequences. The sequences were then concatenated into ASCII text files, one each for coding, intron and flanking. The ends of the individual sequences, as they appeared in the individual GenBank-formatted documents, were tagged to prevent the artifactual joining of those sequences that could result in the creation of artifactual long tracts. The basic characteristics of the coding, intron and flanking files that we used as computational start points for homopolymer tract frequency determination are summarized in Table 2.
Each file was subsequently analyzed using the program "Poly" [38,64], which calculates parameters for non-overlapping homopolymer tracts, including the total base count for each file, GC composition, and the numbers and the frequencies of the homopolymer tracts of different lengths. Poly uses a moving window of 1 bp in length to differentiate tracts and spacers, taking into account the tags used to prevent the artifactual concatenation. These data and additional information are kept as data objects in the program and can be manipulated in various ways.
Poly calculates the observed tract frequency of base i, , of length N by the formula:
where is the number of observed tracts of base i at length N contained in each sequence and lseq is the total length of the sequence (total base count) in which those tracts were counted.
Using the relationship:
the of tracts of length N can be related to N and P'. The parameter P' is the inverse of the frequency, fslope, of the tract base i in the particular genome compartment and is determined from the slope [-log(P')] of an eqn. [1b] plot. The fslope quantity, which represents an effective base frequency for that DNA, can be determined for a given set of data, and then compared to the real frequency for that base occurring in the sequences being examined, as we have previously described [1].
The expected frequency, , of a homopolymer tract of length N randomly occurring is calculated by the formula:
where base frequency is the fractional base composition of the tract base i within the DNA for that file, and N is the tract length.
The level of tract representation for base i is then calculated as the ratio of observed to predicted frequencies, defined as "Representation"(R):
Values larger than 1 indicates base i tracts are "over-represented", while values less than 1 indicate tracts are "under-represented". The log(R) verses N plots presented in Figure 6 were generated using the program Gnuplot [38,65]. The values of N at 0.3 to 1.0 of the log(R) were found using linear interpolation of the data. We term these N values, thresholds, corresponding to particular enrichments of tract occurrence.
The maximum expected length of a homopolymer tract of base i, , given the base composition of the entire sequence, , is calculated by the formula:
The length of a given homopolymer tract can then be compared to its expected length by taking the ratio of the longest observed length, , to the longest expected length, . This parameter is defined as "Proportion (P)". Thus, we have the formula:
P values larger than 1 are called "over-proportional" and P values less than 1 are "under-proportional". P represents the tract length comparison on the x-axis of Figure 8, which is complementary to the parameter R for tract frequency comparisons on y-axis. They are both important parameters for the evaluation of the frequency and length distributions of any tracts.
Authors' contributions
YZ carried out the sequence data collection, clean-up, and analysis, and drafted the major part of the manuscript. JWB contributed both major programs Meltsim and Poly in this study, and also participated in analyzing data and drafted part of the manuscript. KAM conceived the study, participated in its design and coordination and revised and finalized the manuscript in all its revisions. All authors read and approved the final manuscript. The authors acknowledge the contribution from a reviewer of the concepts of retroelement insertion and differential polymerase selectivity as potential origins for the higher A & T tract frequencies than G&C tract frequencies that we describe here.
==== Refs
Marx KA Hess ST Blake RD Characteristics of the large (dA).(dT) homopolymer tracts in D. discoideum gene flanking and intron sequences J Biomol Struct Dyn 1993 11 57 66 8216948
Toth G Gaspari Z Jurka J Microsatellites in different eukaryotic genomes: survey and analysis Genome Res 2000 10 967 981 10899146 10.1101/gr.10.7.967
Streisinger G Okada Y Emrich J Newton J Tsugita A Terzaghi E Inouye M Frameshift mutations and the genetic code Cold Spring Harb Symp Quant Biol 1966 31 77 84 5237214
Kunkel TA Soni A Mutagenesis by transient misalignment J Biol Chem 1988 263 14784 14789 3049589
Fujii S Akiyama M Aoki K Sugaya Y Higuchi K Hiraoka M Miki Y Saitoh N Yoshiyama K Ihara K Seki M Ohtsubo E Maki H DNA replication errors produced by the replicative apparatus of Escherichia coli J Mol Biol 1999 289 835 850 10369765 10.1006/jmbi.1999.2802
Strand M Prolla TA Liskay RM Petes TD Destabilization of tracts of simple repetitive DNA in yeast by mutations affecting DNA mismatch repair Nature 1993 365 274 276 8371783 10.1038/365274a0
Kroutil LC Register K Bebenek K Kunkel TA Exonucleolytic proofreading during replication of repetitive DNA Biochemistry 1996 35 1046 1053 8547240 10.1021/bi952178h
Tran HT Keen JD Kricker M Resnick MA Gordenin DA Hypermutability of homonucleotide runs in mismatch repair and DNA polymerase proofreading yeast mutants Mol Cell Biol 1997 17 2859 2865 9111358
Alexeev DG Lipanov AA Skuratovskii I Poly(dA).poly(dT) is a B-type double helix with a distinctively narrow minor groove Nature 1987 325 821 823 3821870 10.1038/325821a0
Aymami J Coll M Frederick CA Wang AH Rich A The propeller DNA conformation of poly(dA).poly(dT) Nucleic Acids Res 1989 17 3229 3245 2726457
Arnott S Chandrasekaran R Birdsall DL Leslie AG Ratliff RL Left-handed DNA helices Nature 1980 283 743 745 7354864 10.1038/283743a0
Nelson HC Finch JT Luisi BF Klug A The structure of an oligo(dA).oligo(dT) tract and its biological implications Nature 1987 330 221 226 3670410 10.1038/330221a0
Lipanov AA Chuprina VP The structure of poly(dA):poly(dT) in a condensed state and in solution Nucleic Acids Res 1987 15 5833 5844 3615203
Rhodes D Nucleosome cores reconstituted from poly(dA-dT) and the octamer of histones Nucleic Acids Res 1979 6 1805 1816 450714
Simpson RT Kunzler P Chromatin and core particles formed from the inner histones and synthetic polydeoxyribonucleotides of defined sequence Nucleic Acids Res 1979 6 1387 1415 450700
Kunkel GR Martinson HG Nucleosomes will not form on double-stranded RNA or over poly(dA).poly(dT) tracts in recombinant DNA Nucleic Acids Res 1981 9 6869 6888 7335494
Satchwell SC Drew HR Travers AA Sequence periodicities in chicken nucleosome core DNA J Mol Biol 1986 191 659 675 3806678
Marx KA Hess ST Blake RD Alignment of (dA).(dT) homopolymer tracts in gene flanking sequences suggests nucleosomal periodicity in D. discoideum DNA J Biomol Struct Dyn 1994 12 235 246 7848570
Marini JC Levene SD Crothers DM Englund PT A bent helix in kinetoplast DNA Cold Spring Harb Symp Quant Biol 1983 47 279 283 6574847
Wu HM Crothers DM The locus of sequence-directed and protein-induced DNA bending Nature 1984 308 509 513 6323997 10.1038/308509a0
Koo HS Wu HM Crothers DM DNA bending at adenine.thymine tracts Nature 1986 320 501 506 3960133 10.1038/320501a0
Hagerman PJ Straightening out the bends in curved DNA Biochim Biophys Acta 1992 1131 125 132 1610891
Bossi L Smith DM Conformational change in the DNA associated with an unusual promoter mutation in a tRNA operon of Salmonella Cell 1984 39 643 652 6096016 10.1016/0092-8674(84)90471-9
Snyder M Buchman AR Davis RW Bent DNA at a yeast autonomously replicating sequence Nature 1986 324 87 89 3785376 10.1038/324087a0
Zahn K Blattner FR Direct evidence for DNA bending at the lambda replication origin Science 1987 236 416 422 2951850
Jayasena SD Behe MJ Competitive nucleosome reconstitution of polydeoxynucleotides containing oligoguanosine tracts J Mol Biol 1989 208 297 306 2769759 10.1016/0022-2836(89)90390-2
Puhl HL Gudibande SR Behe MJ Poly(dA:dT) and other synthetic polydeoxynucleotides containing oligoadenosine tracts form nucleosomes easily J Mol Biol 1991 222 1149 1160 1762148 10.1016/0022-2836(91)90598-Z
McCall M Brown T Kennard O The crystal structure of d(G-G-G-G-C-C-C-C). A model for poly(dG).poly(dC) J Mol Biol 1985 183 385 396 4020865 10.1016/0022-2836(85)90009-9
Travers AA DNA conformation and protein binding Annu Rev Biochem 1989 58 427 452 2673015 10.1146/annurev.bi.58.070189.002235
Struhl K Naturally occurring poly(dA-dT) sequences are upstream promoter elements for constitutive transcription in yeast Proc Natl Acad Sci USA 1985 82 8419 8423 3909145
Boettcher B Transcription initiation and nuclease-sensitive sites upstream of the epsilon-globin gene in K562 cells are related to poly(dA).poly(dT) sequences J Theor Biol 1990 146 333 339 2259207
Iyer V Struhl K Poly(dA:dT), a ubiquitous promoter element that stimulates transcription via its intrinsic DNA structure Embo J 1995 14 2570 2579 7781610
Schultes NP Szostak JW A poly(dA.dT) tract is a component of the recombination initiation site at the ARG4 locus in Saccharomyces cerevisiae Mol Cell Biol 1991 11 322 328 1986228
Siddiqui-Jain A Grand CL Bearss DJ Hurley LH Direct evidence for a G-quadruplex in a promoter region and its targeting with a small molecule to repress c_myc transcription Proc Natl Acad Sci 2002 99 11593 11598 12195017 10.1073/pnas.182256799
Dechering KJ Cuelenaere K Konings RNH Leunissen JAM Distinct frequency-distributions of homopolymeric DNA tracts in different genomes Nucleic Acids Res 1998 26 4056 4062 9705519 10.1093/nar/26.17.4056
Attimonelli M Liuni S Pesole G CLEANUP : a fast computer program for removing redundancies from nucleotide sequence database Comput Appli Biosci 1996 12 1 8
Blake RD Bizzaro JW Blake JD Day GR Delcourt SG Knowles G Marx KA SantaLucia J Statistical Mechanical Simulation of Polymeric DNA Melting with MELTSIM Bioinformatics 1999 15 370 375 10366657 10.1093/bioinformatics/15.5.370
Bizzaro JW Marx KA Poly: a quantitative analysis tool for simple sequence repeat (SSR) tracts in DNA BMC Bioinformatics 2003 4 22 27 12791171 10.1186/1471-2105-4-22
Lamperti ED Kittelberger JM Smith TF Villa-Komaroff L Corruption of genomic databases with anomalous sequence Nucleic Acids Res 1992 20 2741 2747 1614861
Paricio N Terol J Artero RD Perez-Alonso M Poly(dC) segments and cloning artifacts in databases Genetics 1995 139 487 488 7705649
van Hintum TJL van Treuren R Molecular markers: Tools to improve Genebank efficiency Cell Mol Biol Lett 2002 7 737 744 12378233
Kirchner JM Tran H Resnick MA A DNA polymerase ε mutant that specifically causes +1 frameshift mutations within homonucleotide runs in yeast Genetics 2000 155 1623 1632 10924461
Katti M Ranjekar K Gupta VS Differential Distribution of Simple Sequence Repeats in Eukaryotic Genome Sequences Mol Biol Evol 2001 18 1161 1167 11420357
Roy AM Carroll ML Nguyen SV Salem AH Oldridge M Wilkie AO Batzer MA Deinenger PL Potential Gene Conversion and Source Genes for Recently Integrated Alu Elements Genome Res 2000 10 1485 1495 11042148 10.1101/gr.152300
Sen D Gilbert W Formation of parallel four-stranded complexes by guanine-rich motifs in DNA and its implications for meiosis Nature 1988 334 364 366 3393228 10.1038/334364a0
Sundquist WI Klug A Telomeric DNA dimerizes by formation of guanine tetrads between hairpin loops Nature 1989 342 825 829 2601741 10.1038/342825a0
Hall DB Holmlin RE Barton JK Oxidative DNA damage through long-range electron transfer Nature 1996 382 731 735 8751447 10.1038/382731a0
Ussery D Soumpasis DM Brunak S Starfeldt HH Worning P Krogh A Bias of purine streches in sequenced chromosomes Comput Chem 2002 26 531 541 12144181 10.1016/S0097-8485(02)00013-X
Shimizu M Mori T Sakurai T Shindo H Destabilization of nucleosomes by an unusual DNA conformation adopted by poly(dA).poly(dT) tracts in vivo. Embo J 2000 19 3358 3365 10880448 10.1093/emboj/19.13.3358
Widlund H Cao H Simonsson S Magnusson E Simonsson T Nielsen PE Kahn JD Crothers DM Kubista M Identification and Characterization of Genomic Nucleosome-positioning Sequences J Mol Biol 1997 267 807 817 9135113 10.1006/jmbi.1997.0916
Englander EW Howard BH A naturally occurring T14A11 tract blocks nucleosome formation over the human neurofibromatosis type 1 (NF1)-Alu element J Biol Chem 1996 271 5819 5823 8621451 10.1074/jbc.271.10.5819
Hori R Firtel RA Identification and Characterization of multiple A/T-rich cis-acting elements that control expression from Dictyostelium actin promoters: the Dictyostelium actin upstream activating sequence confers growth phase expression and has enhancer-like properties Nucleic Acids Res 1994 22 5099 5111 7800506
Suter B Schnappauf G Thomas F Poly(dA:dT) sequences exist as rigid DNA structures in nucleosome-free yeast promoters in vivo. Nucleic Acids Res 2000 28 4083 4089 11058103 10.1093/nar/28.21.4083
Filetici P Aranda C Gonzalez A Ballario P GCN5, a yeast transcriptional coactivator, induced chromatin reconfiguration of HIS3 promoter in vivo Biochem Biophys Res Commun 1998 242 84 87 9439614 10.1006/bbrc.1997.7918
Koch KA Thiele DJ Functional analysis of a homopolymeric (dA-dT) element that provides nucleosome access to yeast and mammalian transcription factors J Biol Chem 1999 274 23752 23760 10446135 10.1074/jbc.274.34.23752
Fashena S Reeves R Ruddle NH A poly(dA:dT) upstream activating sequence binds high-mobility group I protein and contributes to lymphotoxin (tumor necrosis factor-β) gene regulation Mol Cell Biol 1992 12 894 903 1732752
Winter E Varshavsky A A DNA binding protein that recognizes oligo(dA).oligo(dT) tracts Embo J 1989 8 1867 1877 2670564
Garreau H Williams JG Two nuclear binding proteins of Dictyostelium discoideum with a high affinity for poly(dA)-poly(dT) Nucleic Acids Res 1983 11 8473 8484 6324086
Horn J Dietz-Schmidt A Zundorf I Garin J Dingermann T Winckler T A Dictyostelium protein binds to distinct oligo(dA)×oligo(dT) DNA sequences in the C-module of the retrotransposable element DRE Eur J Biochem 1999 265 441 448 10491202 10.1046/j.1432-1327.1999.00768.x
Moran LA Scrimgeour KG Horton HR Ochs RS Rawn JD Biochemistry Neil Patterson, Prentice Hall, Engelwood Cliffs, NJ 1994 2
| 15598342 | PMC539357 | CC BY | 2021-01-04 16:32:42 | no | BMC Genomics. 2004 Dec 14; 5:95 | utf-8 | BMC Genomics | 2,004 | 10.1186/1471-2164-5-95 | oa_comm |
==== Front
BMC Chem BiolBMC Chemical Biology1472-6769BioMed Central London 1472-6769-4-21560358710.1186/1472-6769-4-2Research ArticleAn informatics search for the low-molecular weight chromium-binding peptide Dinakarpandian Deendayal [email protected] Vincent [email protected] Shveta [email protected] Kambiz [email protected] Brian [email protected] Horn J David [email protected] Division of Computer Science and Electrical Engineering, School of Computing and Engineering, University of Missouri-Kansas City, Kansas City, MO 64110, USA2 Department of Chemistry, University of Missouri-Kansas City, 5110 Rockhill Road, Kansas City, MO 64110, USA3 National Biomedical EPR Center, Medical College of Wisconsin, Milwaukee, WI 53226 USA2004 16 12 2004 4 2 2 10 8 2004 16 12 2004 Copyright © 2004 Dinakarpandian et al; licensee BioMed Central Ltd.2004Dinakarpandian 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 amino acid composition of a low molecular weight chromium binding peptide (LMWCr), isolated from bovine liver, is reportedly E:G:C:D::4:2:2:2, though its sequence has not been discovered. There is some controversy surrounding the exact biochemical forms and the action of Cr(III) in biological systems; the topic has been the subject of many experimental reports and continues to be investigated. Clarification of Cr-protein interactions will further understanding Cr(III) biochemistry and provide a basis for novel therapies based on metallocomplexes or small molecules.
Results
A genomic search of the non-redundant database for all possible decapeptides of the reported composition yields three exact matches, EDGEECDCGE, DGEECDCGEE and CEGGCEEDDE. The first two sequences are found in ADAM 19 (A Disintegrin and Metalloproteinase domain 19) proteins in man and mouse; the last is found in a protein kinase in rice (Oryza sativa). A broader search for pentameric sequences (and assuming a disulfide dimer) corresponding to the stoichiometric ratio E:D:G:C::2:1:1:1, within the set of human proteins and the set of proteins in, or related to, the insulin signaling pathway, yields a match at an acidic region in the α-subunit of the insulin receptor (-EECGD-, residues 175–184). A synthetic peptide derived from this sequence binds chromium(III) and forms a metal-peptide complex that has properties matching those reported for isolated LMWCr and Cr(III)-containing peptide fractions.
Conclusion
The search for an acidic decameric sequence indicates that LMWCr may not be a contiguous sequence. The identification of a distinct pentameric sequence in a significant insulin-signaling pathway protein suggests a possible identity for the LMWCr peptide. This identification clarifies directions for further investigation of LMWCr peptide fractions, chromium bio-coordination chemistry and a possible role in the insulin signaling pathway. Implications for models of chromium action in the insulin-signaling pathway are discussed.
==== Body
Background
Type II diabetes continues to grow as a worldwide epidemic; it is expected that this disease will surpass 300 million cases by 2025 [1-5]. Understanding the complex signaling mechanisms of insulin receptor and downstream factors, e.g. IRS1 and IRS2 [6], is crucial in the design of effective therapeutic strategies. Clarification of such signaling mechanisms is expected to lead to discoveries to cure or prevent diabetes and other metabolic conditions [1]. Current strategies include the search for small synthetic or naturally-derived molecules to act upstream of IRS1, to adequately compensate for insulin dysregulation [7,8]. One unknown facet in this complex problem is the role that chromium may play in the regulation of glucose metabolism; the molecular basis of chromium action in biological systems has not been definitively explained [9,10]. While the toxicology Cr(VI) has been well characterized [11], an understanding of the biochemistry and action of Cr(III) continues to elude researchers within this field [9,11-18]. The lack of good mechanistic models and experiments currently limits researchers' ability to assess the relative importance of Cr(III) and its possible roles in glucose metabolism, obesity or Type II diabetes. Recently, a Cr-peptide fraction isolated from bovine liver [19-23] was shown to potentiate the action of insulin. The amino acid composition of this "low molecular weight Cr" complex (LMWCr, containing 3–4 chromic atoms) was identified to be approximately E:G:C:D::4:2:2:2 [18,21,24], though the sequence remains unknown (Figure 1) and there appears to be some difficulty in exactly repeating this work. Although not specifically related to Cr(III) metabolism, Ramasami and co-workers report that Cr(III) induces structural changes and long range ordering in collagenous tissues and suggest the formation of multinuclear Cr-oxo and hydroxo clusters (2, 3, or 4 Cr atoms) holding the proteinaceous molecular assembly together [25,26]. In this study, we report on a bioinformatics search for the LMWCr peptide, and propose that the peptide that binds to Cr(III) may be the pentameric sequence (-EECGD-) as it is found within the insulin receptor (INSR; Swiss-Prot: P06213) (-NKDDNEECGD-, residues 175–184) and conforms to the predicted stoichiometry. Further, visible absorption data and electron paramagnetic measurements of the synthesized disulfide linked dimeric Cr-peptide complex are identical to that of the biologically isolated fractions.
Figure 1 Hypothetical models of low-molecular weight Cr peptide complexes.
Results
Genomic search for LMWCr
Recent reports [18,21,24], using extracts from bovine liver, suggest that LMWCr is of peptide origin, having an approximate stoichiometric ratio consisting of E:D:G:C::4:2:2:2. In the absence of the exact sequence of this peptide, we generated all possible permutations matching the reported stoichiometry of LMWCr (10!/4!2!2!2! = 18,900) and performed an exact match search of the non-redundant (nr) protein database for their occurrence. Based on a simple model where all amino acids are equally possible and about 1.5 million sequences of average length 200 present in the nr database, the E-value for an independent exact match is about 0.55 [(0.05)10 * 200 * 1.5 * 106 * 18,900].
We found perfect matches in the nr database corresponding to 3 unique permutations: EDGEECDCGE, DGEECDCGEE and CEGGCEEDDE. One match, CEGGCEEDDE, was found only in one sequence annotated as a putative protein kinase in rice (Oryza sativa). The peptide sequences, EDGEECDCGE and DGEECDCGEE, both matched the same region of the disintegrin domain of the human and mouse versions of ADAM 19 (A Disintegrin and Metalloproteinase domain 19; GeneID: 8728) [27]. Since the bovine form of ADAM 19 has yet to be sequenced, we were unable to determine if this sequence is also present in bos taurus. This acid-rich sequence region (Table 1) is present at the beginning of the disintegrin domain of ADAM19 [27]. We note that the DNA coding location for this subsequence is at the very end of an exon, giving rise to the speculation that the decameric peptide could possibly be an alternative splicing product. Arguments against this sequence giving rise to LMWCr and the complex acting intracellulary are that ADAM 19 is a membrane protein with extracellular domains. This argument is not needed if Cr might act extracellularly; the conserved nature of the motif leads one to suspect that it plays an important role, but it is at best a highly tentative candidate for being LMWCr. From these results, LMWCr appears unlikely to be a contiguous peptide sequence, unless an as-yet undiscovered peptide is found.
Table 1 Multiple sequence alignment showing conservation of the EDGEECDCGE motif in all known mammalian forms of ADAM19 and a homologous putative ADAM in fission yeast (top). Multiple sequence alignment showing proposed Cr(III) binding sequence, NKDDNEECGD, conserved in the insulin receptor (INSR) across species, but not in the insulin-like growth factor receptor (IG1R) (bottom).
SwissProt Protein Species Sequence
Q9H013 ADAM19 Human GNGYLEDGEECDCGEEEECNNP
O35674 ADAM19 Mouse GNGYLEDGEECDCGEEEECNNP
ADAM19 Rat GNGYLEDGEECDCGEEDECKNP
O13766 Homolog S. Pombe GNGIVEDGEECDCGEDCENNPC
P15208 INSR Mouse NKDDNEECGDVCPGT
P15127 INSR Rat NKDDNEECGDVCPGT
P06213 INSR Human NKDDNEECGDICPGT
Q9PVZ4 Homolog Xenopus NRDNKEECGDVCPGT
Q60751 IG1R Mouse NKPPKECGDLCPGTL
P24062 IG1R Rat NKPPKECGDLCPGTL
P08069 IG1R Human NKPPKECGDLCPGTL
O73798 Homolog Xenopus NKPPKECVDLCPGA.
Noting that the reported stoichiometry consists of even numbers of amino acids, we further considered the possibility that LMWCr might be a disulfide-linked dimer, with each monomeric unit having the composition, E:D:G:C::2:1:1:1. Such a hypothesis is consistent with the limited resolution of the experimental data used to derive the stoichiometry [24] and yields 60 sequence permutations. The expected number of random matches in the nr database is much higher, ~104. Restricting this search to the human proteome gave rise to 439 (expected: ~102) matches, with no reason to prefer one over the other.
Insulin signaling pathway mapping
Considering that LMWCr might be a subsequence of a protein related to "insulin" or known to be involved in the insulin signalling pathway, we compiled a set of 96 such sequences. This set comprised two components: 1) proteins playing a role in the insulin signaling pathway (23 protein sequences were selected) derived from pathway charts, and 2) the set of all protein sequences derived from SwissProt for the search, "insulin + human" (78 entries). The two components were compared for redundancy and the duplicates removed (5 instances).
A cross comparison of the resulting 439 entries using the BLAST query method versus the set of 96 pathway and insulin-related proteins results in a unique match for one of the 60 pentapeptides, EECGD, within the insulin receptor (INSR), residues 180–184 (expected matches: ~10-1). Comparison of the pentapeptide set to a more detailed insulin signaling pathway construct [28] yielded the same result.
This sub-sequence lies in the extracellular α-subunit of INSR in an acid-rich region (-NKDDNEECGD-) towards the end of the L1 domain, and at the start of the "cysteine-rich region". A BLAST query [29] for all INSR homologs in the nr database shows that this sub-sequence and acidic region is conserved in mouse and rat. Given the location of this acidic sequence within a molecule central to glucose homeostasis, the correspondence with experimentally measured stoichiometry and conservation across multiple species, we speculate that this sequence, or a fragment from this acid rich region may give rise to Cr-peptide fractions isolated from tissue. This suggestion implies that such fractions may not be homogeneous, discrete Cr(III) complexes, i.e. proteolysis may lead to a group of similar peptides that differ by one or more amino acid residues on either side of the Cr binding site. This interpretation differs significantly from reports on the isolation of LMWCr, which attempt to avoid a proteolytic product.
A crystal or solution structure of this region of the insulin receptor has not been determined. However, the crystal structure of a homologous molecule, the insulin-like growth factor receptor (IGR1, GeneID: 124240) has been published [30,31]. Sequence alignment shows that this molecule exhibits a conserved difference in this region, having the subsequence -KECGD- in the same region instead of -EECGD- as found in INSR. The decameric "acidic region" found in INSR is not present in this growth factor receptor – cf. -NKPPKECGDLCPGTL-. The cysteine in -KECGD- forms a disulfide bond with another cysteine residue 28 positions away in the crystal structure. However, we do not know if this is due to the crystallization conditions or representative of the natural form. Further, the insulin receptor possesses 3 additional cysteine residues compared to the insulin-like growth factor receptor, thus the pattern of pairing of cysteine residues to form disulfide bonds may be different in the two receptor types. The observed difference may be great enough to preclude Cr(III) acting on the insulin-like growth factor receptor.
Chromium-peptide complex synthesis
The identified sequence in hand, we synthesized the pentapeptide (AcEECGD-CONH2) and generated the disulfide-linked dimer via air oxidation. This peptide was subjected to conditions for the reconstitution of apo-LMWCr to generate a holo-peptide complex. Incubation of the peptide with fresh solutions of CrCl3 results in a clear gray-green solutions of a Cr-peptide complex with visible electronic spectra (Figure 2) consistent with those reported for LMWCr [24]. Further, EPR experiments with Cr-peptide complexes generated from (AcEECGD-CONH2)2 and (AcNEECGD-CONH2)2 indicate that Cr exists in trinuclear arrays in these complexes. A preliminary spectrum is shown in Figure 2 consistent with that of isolated LMWCr fractions [32], further characterization is necessary to identify the exact nature of the complexes in these experiments.
Figure 2 Schematic of proposed Cr-peptide complex. The EPR spectrum (top) is comparable to that of isolated LMWCr with g ~2.0; the spectrum also includes additional hyperfine coupling. Visible absorption spectra of complex formation from the reconstitution of apo-peptide with fresh solutions of chromium chloride (bottom). These data are qualitative and may comprise more than one discreet species.
Discussion
Barring the discovery of a novel, unsequenced or unidentified protein or peptide, these data point to the possible sequence of LMWCr fractions and may point to new strategies in therapeutic design. In addition, the question of sequence specificity in Cr(III)-peptide complexes must be fully addressed, along with thermodynamic and kinetic aspects of Cr(III) binding and transfer.
Models of non-toxicological action of Cr(III) in biological systems may broadly fall into three categories: structural, redox, and iron homeostasis. The earliest models [12], and those advanced by Vincent [18], fall into a structural category and focus on the interactions of Cr(III) with peptides and proteins to affect insulin signaling and glucose metabolism, either directly or indirectly. There is an important redox model recently advanced [33] that suggests higher oxidation states of Cr interact with tyrosine phosphatases to inhibit the down-regulation of the insulin receptor. Finally, the chemical similarity of the Cr(III) and Fe(III) cations, and various in vitro studies suggest that Cr(III) replacement in the physiological iron transport and storage apparatus may lead to some small beneficial outcome for certain diseases [10]. The biological relevance of these models, and of in vitro experiments (including our own) may be finally ascertained only after the fact.
Although unexpected, the results in this report and a critical review of other literature [9,11-18], suggest that an extracellular model for Cr(III) biochemistry with respect to insulin signaling may be plausible (see Supporting Information). Such a structural model would include the known aspects of INSR cycling and insulin degradation [34], and include the proposed interactions between Cr(III) and the INSR at the acidic site identified by our genomic search. This model is reminiscent of Mertz and co-workers' original proposal [12] of a ternary interaction between Cr(III), insulin and insulin receptor. It is substantially different from intracellular mechanisms for LMWCr action [16,18], redox mechanisms [33], and the iron homeostasis model [10]. In addition, the cycling of the insulin receptor and insulin degradation [34] may satisfy the problems of cellular distribution of Cr(III) and production of LMWCr via proteolysis. Experimentally observed insulin potentiating activity of Cr(III) may result from binding to the alpha subunit or bridging interaction between the two α subunits of an intact INSR molecule.
This model is a parsimonious alternative to current proposals of Cr action in the insulin signaling pathway. However, this model points directly back to significant kinetic and a thermodynamic questions about Cr(III) in biological systems. For instance, what is the physical form of Cr in the bloodstream? How is Cr(III) transported and exchanged between ligands in the serum? Is transport specific? What structure/activity relationship exists in Cr(III) complexes to allow their transport across biological membranes? Thermodynamically, a hydrolyzed, multinuclear Cr cluster should predominate at neutral pH, but transport by transferrin would presumably be in the mononuclear Fe binding sites. Alternatively, Cr(III) clusters may be transported non-specifically in serum by proteins, possibly including transferrin and serum albumin. At this point, there exist significant gaps in understanding the possible biochemistry of Cr(III) and what molecular processes it may affect.
The proposed extracellular model of Cr(III) action in this report is upstream of IRS1, a therapeutic target of White and others [7,8], and may lend itself to small-molecule therapeutic strategies for diabetes and other metabolic conditions [1]. We hope this model may pave the way for innovative experiments, better models of Cr(III) biochemistry and excretion, and further understanding of signaling events in complex biochemical pathways.
Conclusions
A bioinformatic search for an acidic decameric sequence matching reported stoichiometries of LMWCr amino acid composition indicates that the peptide may not be a contiguous sequence. An expanded search localized a pentameric sequence in the insulin receptor and suggests a possible identity for the Cr(III)-containing peptide fractions derived from liver. Disulfide linked penta- and hexameric peptides based on the identified sequence bind Cr(III) in a similar fashion to LMWCr fractions reported in the literature.
Methods
Search for LMWCr
The nr database was downloaded and Perl scripts used to search for exact matches corresponding to all unique permutations of EEEEGGCCDD. In a separate search, a set of 78 unique protein sequences from Swiss-Prot were obtained as a result of using the query "insulin human." This set was searched for exact matches corresponding to all unique permutations of EEGCD. In addition, the human proteome was also downloaded and searched for matches to the same set of pentameric peptides.
N-Acetylglutamylglutamylcysteinylglycylaspartyl carboxamide (AcEECGD-CONH2)
The peptide, AcEECGD-CONH2, was synthesized by continuous-flow automated solid-phase synthesis on a Perseptive Biosystems Pioneer Peptide Synthesis System. Peptide synthesis was performed by using standard Fmoc-protection strategies with TBTU/DIEA activation strategy. Typically, a 0.5 mmol scale, using Rink amide resin (loading ~0.65 mmol/g) and four times excess of the other reagents (TBTU, protected amino acid) were used. After the solid-phase synthesis was complete, acetylation at the amino terminus was carried out for 2 hours (50:50:1:: acetic anhydride:dimethylformamide:pyridine), and the resulting peptide cleaved from the resin. The peptide was cleaved from the solid phase resin by adding a slurry of 94% trifluoroacetic acid, 2.5% of water, 2.5% of ethanedithiol, and 1% of triisopropylsilane to the reaction vessel and shaking it for 5 h. The trifluoroacetic acid, containing the peptide, was filtered off by vacuum, and the resin was washed 2–3 times with trifluoroacetic acid. The filtrate was evaporated under nitrogen gas until the volume was reduced to 15 mL. 30 mL of ice-cold diethyl ether was added to the filtrate, causing the peptide to precipitate, and the mixture centrifuged to form a pellet of the peptide. The diethyl ether was decanted and the peptide pellet was washed with diethyl ether three times. Finally, the peptide was dissolved in 20 mL water containing 0.1% trifluoroacetic acid and extracted with diethyl ether three times. The aqueous portion was collected and freeze-dried to give the synthetic peptide. AcEECGD-CONH2, has a retention time of 5.3 minutes on a 250 mm × 4.5 mm C-18 reversed phase HPLC column using a gradient of 5 to 20% acetonitrile in 0.1% trifluoroacetic acid/water mobile phase running at 1 mL per minute and detected at a wavelength of 220 nm.
N-Acetylglutamylglutamylcystinylglycylaspartylcarboxamide dimer ((AcEECGD-CONH2)2)
The synthetic peptide, AcEECGD-CONH2, was dissolved to 10 mg/mL in a 0.1 M solution of ammonium carbonate at pH 7 and allowed to oxidize under ambient air for 48 hours. The product was isolated by lyophilization and analyzed by HPLC. The disulfide peptide dimer, (AcEECGD-CONH2)2, has a retention time of 6.0 minutes using the above conditions. The product, (AcEECGD-CONH2)2, has a calculated mass of 1183.35141 and exhibits a mass of 1183.3477 Daltons when analyzed by electrospray mass spectrometry.
Chromium N-acetylglutamylglutamylcystinylglycylaspartylcarboxamide (Cr3O(AcEECGD-CONH2)2)
Fifteen milligrams of (AcEECGD-CONH2)2 was weighed and dissolved in 30 mL of water. A portion of this solution (5 mL) was taken up in a 25 mL tube, and 4 equivalents of chromium(III) chloride were added as a solid or in an aqueous solution. The reaction of the two components took several minutes and was monitored by ultraviolet-visible spectrophotometry. The chromium peptide complex, Cr3O(AcEECGD-CONH2)2, exhibits characteristic spectral absorbance features at 432 nm and 615 nm. The features are consistent with chromium bound to oxygen atom donors and similar to the reported spectrum of LMWCr [24,32]. EPR spectra were collected using the following parameters: microwave frequency, 9.632 GHz; microwave power incident to the cavity, 2 mW; temperature, 10 K (LHe cryostat). Samples were prepared by incubating a solution of peptide with chromium chloride at a final concentration of 1 mM in metal with excess peptide. The complete characterization (EPR, MS, etc.) of this and analogous Cr-peptide complexes will be reported elsewhere.
Authors' contributions
DD constructed informatics scripts and searches and made sequence alignments. VM and KA constructed pathway maps and conducted informatics searches. SC assisted in peptide synthesis and conducted HPLC analysis. VM and BB collected ESR spectra. JDVH conceived of the study, synthesized peptides and Cr(III) complexes and drafted the manuscript, tables and graphics. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Supporting figures (3) include HPLC and MS data for AcEECGD-CONH2 and proposed Cr(III) model at INSR. Supporting tables (2) include FASTA sequences for insulin signaling map (20 proteins) and pentameric peptides found in genomic search (439 entries)
Click here for file
Acknowledgements
This work was supported by an award from the American Heart Association, and by funds from the University of Missouri (to J.D.V.H. and V.M.) and University of Missouri Research Board. EPR training and experiments at the National Biomedical EPR Center, Milwaukee, WI, were supported by NIH grant EB001980.
==== Refs
White MF Insulin signaling in health and disease. Science 2003 302 1710 1711 14657487
Amos AF McCarty DJ Zimmet P The rising global burden of diabetes and its complications: estimates and projections by 2010. Diabet Med 1997 14 S5 S85 9212322
King H Aubert RE Herman WH Global Burden of Diabetes, 1995-2025. Prevelence, numerical estimates and projections. Diabetes Care 1998 21 1414 1431 9727886
Kopelman PG Hitman GA Diabetes. Exploding Type II. Lancet 1998 352 SIV5 9872152
Moller DE New drug targets for type 2 diabetes and the metabolic syndrome Nature 2001 414 821 827 11742415
Patti ME Sun XJ Bruening JC Araki E Lipes MA White MF Kahn CR 4/PS/Insulin Receptor Substrate (IRS)-2 Is the Alternative Substrate of the Insulin Receptor in IRS-1-deficient Mice J Biol Chem 1995 270 24670 24673 7559579
Pender C Goldfine ID Manchem VP Evans JL Spevak WR Shi S Rao S Bajjalieh S Maddux BA Youngren JF Regulation of Insulin Receptor Function by a Small Molecule Insulin Receptor Activator J Biol Chem 2002 277 43565 43571 12213804
Lee YH White MF Insulin receptor substrate proteins and diabetes. Arch Pharm Res 2004 27 361 370 15180298
Levina A Codd R Dillon CT Lay PA Chromium in biology: Toxicology and Nutritional Aspects Prog Inorg Chem 2002 51 145 250
Stearns DM Is chromium a Trace essential metal? Biofactors 2000 11 149 162 10875302
Codd R Dillon CT Levina A Lay PA Studies on the Genotoxicity of Chromium: From the Test Tube to the Cell Coord Chem Rev 2001 216-217 537 582
Mertz W Toepfer EW Roginski EE Polansky MM Present knowledge of the role of chromium Fed Proc 1974 33 2275 2280 4372107
Mertz W Effects and Metabolism of Glucose Tolerance Factor Nut Rev 1975 33 129 135
Mertz W The Essential Trace Elements Science 1981 213 1332 1338 7022654
Anderson RA Chromium, Glucose Intolerance and Diabetes J Am Coll Nutr 1998 17 548 555 9853533
Davis CM Vincent JB Chromium Oligopeptide Activates Insulin Receptor Tyrosine Kinase Activity Biochemistry 1997 36 4382 4385 9109644
Davis CM Vincent JB Chromium in Carbohydrate and Lipid Metabolism J Biol Inorg Chem 1997 2 675 679
Vincent JB Elucidating a Biological Role for Chromium at a Molecular Level Acc Chem Res 2000 33 503 510 10913239
Wada O Wu GY Yamamoto A Manabe S Ono T Purification and Chromium-Excretory Function of Low-Molecular-Weight, Chromium-Binding Substances from Dog Liver Environ Res 1983 32 228 239 6617615
Yamamoto A Wada O Ono T Distribution and Chromium-Binding Capacity of a Low-Molecular-Weight, Chromium-Binding Substance in Mice J Inorg Biochem 1984 22 91 102 6502162
Yamamoto A Wada O Ono T Isolation of a Biologically Active Low-molecular-mass Chromium Compound from Rabbit Liver Eur J Biochem 1987 165 627 631 3595604
Yamamoto A Wada O Suzuki H Purification and Properties of Biologically Active Chromium Complex from Bovine Colostrum J Nutr 1988 118 39 45 3275760
Yamamoto A Wada O Manabe S Evidence that Chromium is an Essential Factor for the Biological Activity of Low-molecular-weight, Chromium-binding Substance Biochem Biophys Res Comm 1989 163 189 193 2775259
Davis CM Vincent JB Isolation and Characterization of a Biologically Active Chromium Oligopeptide from Bovine Liver Arch Biochem Biophys 1997 339 335 343 9056266
Gayatri R Rajaram R Rajaram A Govindaraju K Rao JR Nair BU Ramasami T Chromium induced structural changes in biomolecules Proc Indian Acad Sci (Chem Sci) 1997 109 307 317
Gayatri R Rajaram R Nair BU Chandrasekaran F Ramasami T Chromium-induced molecular assemblies and long range ordering in collagenous tissues: A conceptual insight into chromium tanning Proc Indian Acad Sci (Chem Sci) 1999 111 133 145
Kurohara K Matsuda Y Nagabukuro A Tsuji A Amagasa T Fujisawa-Sehara A Meltrin b (ADAM19) gene: cloning, mapping, and analysis of the regulatory region. Biochem Biophys Res Comm 2000 270 522 527 10753657
White M Insulin Signaling Pathway. Science's STKE (Connections Map, as seen October 2004)
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: A new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694
Garrett TP McKern NM Lou M Frenkel MJ Bentley JD Lovrecz GO Elleman TC Cosgrove LJ Ward CW Crystal structure of the first three domains of the type-1 insulin-like growth factor receptor Nature 1998 394 395 399 9690478
Ward CW Garrett TP McKern NM Lou M Cosgrove LJ Sparrow LG Frenkel MJ Hoyne PA Elleman TC Adams TE Lovrecz GO Lawrence LJ Tulloch PA The three dimensional structure of the type I insulin-like growth factor receptor Mol Pathol 2001 54 125 132 11376122
Jacquamet L Sun YJ Hatfield J Gu WW Cramer SP Crowder MW Lorigan GA Vincent JB Latour JM Characterization of chromodulin by X-ray absorption and electron paramagnetic resonance spectroscopies and magnetic susceptibility measurements J Am Chem Soc 2003 125 774 780 12526678
Mulyani I Levina A Lay PA Biomimetic Oxidation of Chromium(III): Does the Antidiabetic Activity of Chromium(III) Involve Carcinogenic Chromium(VI)? Angew Chem Int Ed Engl 2004 43 4504 4507 15340954
Di Guglielmo GM Drake PG Baass PC Authier F Posner BI Bergeron JJ Insulin receptor internalization and signalling. Mol Cell Biochem 1998 182 59 63 9609114
| 15603587 | PMC539358 | CC BY | 2021-01-04 16:30:50 | no | BMC Chem Biol. 2004 Dec 16; 4:2 | utf-8 | BMC Chem Biol | 2,004 | 10.1186/1472-6769-4-2 | oa_comm |
==== Front
BMC Complement Altern MedBMC Complementary and Alternative Medicine1472-6882BioMed Central London 1472-6882-4-191558833310.1186/1472-6882-4-19Research ArticleChinese herbal recipe versus diclofenac in symptomatic treatment of osteoarthritis of the knee: a randomized controlled trial [ISRCTN70292892] Teekachunhatean Supanimit [email protected] Puongtip [email protected] Noppamas [email protected] Kanit [email protected] Suwalee [email protected] Sorasak [email protected] Sumalee [email protected] Department of Pharmacology, Faculty of Medicine, Chiang Mai University, Thailand2 Department of Orthopedics, Faculty of Medicine, Chiang Mai University, Thailand3 Department of Radiology, Faculty of Medicine, Chiang Mai University, Thailand4 Division of Pharmaceutical Sceinces, Faculty of Pharmacy, Chiang Mai University, Thailand5 Department of Microbiology, Faculty of Medicine, Chiang Mai University, Thailand2004 13 12 2004 4 19 19 5 8 2004 13 12 2004 Copyright © 2004 Teekachunhatean et al; licensee BioMed Central Ltd.2004Teekachunhatean 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
Duhuo Jisheng Wan (DJW) is perhaps the best known and most widely used Chinese herbal recipe for arthralgia, but the clinical study to verify its efficacy is lacking. The purpose of this study was to compare the efficacy of DJW versus diclofenac in symptomatic treatment of osteoarthritis (OA) of the knee.
Methods
This study was a randomized, double-blind, double-dummy, controlled trial. The 200 patients suffering from OA of the knee, were randomized into the DJW and diclofenac group. The patients were evaluated after a run-in period of one week (week 0) and then weekly during 4 weeks of treatment. The clinical assessments included visual analog scale (VAS) score that assessed pain and stiffness, Lequesne's functional index, time for climbing up 10 steps, as well as physician's and patients' overall opinions on improvement.
Results
Ninety four patients in each group completed the study. In the first few weeks of treatment, the mean changes in some variables (VAS, which assessed walking pain, standing pain and stiffness, as well as Lequesne's functional index) of the DJW group were significantly lower than those of the diclofenac group. Afterwards, these mean changes became no different throughout the study. Most of the physician's and patients' overall opinions on improvement at each time point did not significantly differ between the two groups. Approximately 30% of patients in both groups experienced mild adverse events.
Conclusion
DJW demonstrates clinically comparable efficacy to diclofenac after 4 weeks of treatment. However, the slow onset of action as well as approximately equal rate of adverse events to diclofenac might limit its alternative role in treatment of OA of the knee.
==== Body
Background
Osteoarthritis (OA) is the most prevalent joint disorder characterized by articular cartilage degradation with an accompanying peri-articular bone response [1]. OA affects many joints, with diverse clinical patterns, but OA of the hip and knee is the major cause of disability [2]. A clinical manifestation of OA of the knee is pain in and around the joint that is typically worse with weight-bearing and at night. Other manifestations include morning stiffness, stiffness after rest, crepitation on motion, limited joint motion and/or joint deformity [3]. Although there are many treatment modalities, OA is still widely treated with nonsteroidal anti-inflammatory drugs (NSAIDs) [4]. Nonetheless, since the inflammatory component of OA is usually minimal, a need for the anti-inflammatory effect of NSAIDs used in this condition is still controversial [5-7]. Moreover, long-term use of NSAIDs is also directly related to many side effects, including gastrointestinal bleeding, hypertension, congestive heart failure, hyperkalemia, and renal insufficiency [8]. Although some of these disadvantages can be avoided by using paracetamol or selective cyclooxygenase II (COX-II) inhibitors, long-term use of paracetamol possibly leads to hepatotoxicity and chronic renal impairment [9,10]. In addition, the relatively high cost of selective COX-II inhibitors seems to be unsuitable for Thailand's present socio-economic status.
The use of Chinese and other foreign patent herbal medicines (pills and tablets) in arthralgia treatment is highly prevalent and increasing in Thailand, but importing these medicines from the People's Republic of China and other foreign countries is usually rather expensive. However, the cost of similar preparations can be minimized by using imported dried herbs available in Thailand as raw materials in the manufacturing process coupled with simple and inexpensive traditional drug manufacturing techniques. Thus, if clinical studies suggest that these herbal medicines are as effective and/or less toxic than conventional treatment, promotion of self-produced recipes in each community will lead to community-directed osteoarthritic treatment in Thailand.
The herbal recipe used in this study was "Duhuo Jisheng Wan (DJW)", which means pill of Pubescent angelica root and Mulberry mistletoe combination, and it was quoted from the book Bei Ji Qian Jing Yao Fang compiled by Sun Simiao in the Tang Dynasty (652 A.D.) [11,12]. Although this recipe is perhaps the best known and most widely used formula for arthralgia and also sold as a patent remedy [13], the clinical study to verify its efficacy (compared with conventional treatment) is lacking. Thus, the objectives of this study were to verify the efficacy of DJW and compare its efficacy versus diclofenac in symptomatic treatment of OA of the knee.
Methods
Research design
This randomized, double-blind, double-dummy, controlled trial was approved by the Medical Ethics Committee of the Faculty of Medicine, Chiang Mai University and was in compliance with the Helsinki Declaration.
Subjects
Two hundred out-patients of either sex were recruited. They were aged over 40 years, and had been suffering from unilateral or bilateral OA of the knee according to the criteria of the American College of Rheumatology [3] for more than 3 months. After the use of usual medications had ceased for 7 days, the visual analog scale (VAS) score that assessed pain during the most painful knee movement had to be more than 40, and Lequesne's functional index [14] had to be over 7 points. Participants had to be able to walk and give both verbal and written information regarding the study. Signed informed consent was obtained prior to entry. Exclusion criteria included an underlying inflammatory arthropathy, hyperuricemia, expectation of surgery in the near future, recent injury in the area affected by OA of the knee, intra-articular corticosteroid injections within the last 3 months, hypersensitivity to NSAIDs, abnormal liver or kidney function tests, major abnormal finding on complete blood count, history of coagulopathies, history of peptic ulceration and upper GI hemorrhage, uncontrolled hypertension, congestive heart failure, hyperkalemia, pregnancy, lactation and malignant tumors.
Treatment procedures
During a run-in period of 1 week (week 0), patients considered eligible for the study were informed to discontinue all analgesics, anti-inflammatory drugs, and other modalities for the treatment of arthralgia and arthritis. At the beginning of week 1, patients who still met the eligible criteria were randomized into 2 groups (DJW and diclofenac group) and treated for 4 weeks (Table 1). Other medications and treatment modalities for OA were prohibited during the study. In addition, a count of unused drugs and placebos was made weekly in order to check for the rates of compliance with medication.
Table 1 Treatment in the DJW and diclofenac group.
Treatment DJW group Diclofenac group
Capsule Placebo Diclofenac
Herbal capsule DJW Placebo
1. Diclofenac and its placebo
Twenty five mg film-coated tablets of commercially marketed diclofenac sodium (Voltaren®) were provided by Novartis (Thailand) Co., Ltd. In order to completely blind the patients, each diclofenac tablet was packed into a capsule with an appearance identical to its placebo. Either diclofenac or placebo was prescribed at 1 capsule, 3 times a day, immediately after meals.
2. DJW and its placebo
DJW and its placebo were prepared by the Department of Pharmaceutical Sciences, Faculty of Pharmacy, Chiang Mai University. It consisted of 7.75% each of Radix Angelicae Pubescentis (Duhuo), Radix Gentianae Macrophyllae (Qinjiao), Cortex Eucommiae (Duzhong), Radix Achyranthis Bidentatae (Niuxi), Radix Angelicae Sinensis (Danggui), Herba Taxilli (Sangjisheng), Radix Rehmanniae Preparata (Shudihuang), Rhizoma Chuanxiong (Chuanxiong), Cortex Cinnamomi (Rougui) and Radix Ledebouriellae (Fangfeng), 5% each of Radix Paeoniae Alba (Baishao), Radix Codonopsis (Dangshen), Radix Glycyrrhizae (Gancao) and Poria (Fuling), as well as 2.5% of Herba Asari (Xixin).
Xixin, Niuxi, Shudihuang and Rougui were imported from the Shantou Traditional Chinese Medicine Factory, the People's Republic of China (PRC). The remaining herbs were imported from the Qixin Co., Ltd. (Hebei Province), PRC. Each pulverized ingredient was mixed thoroughly together according to the formula mentioned above and prepared into honeyed pills, which were baked in a hot air oven until completely dry, and then pulverized. The pulverized powder was finally filled into capsules of 500 mg per capsule. The quality control and standardization of DJW (i.e., assessment of weight variation, disintegration time, screening for microorganisms and aflatoxin) were conducted by using guidelines recommended by the Food and Drug Administration of Thailand [15]. DJW and its placebo were prepared in 4 separate lots. Every lot had to pass for quality control and standardization before prescription and they were used within 8 weeks in order to ascertain the stability of active substances, and avoid microorganism and aflatoxin contamination during the study. DJW was prescribed at 6 capsules (3 g) each time, 3 times a day, immediately after meals. Its placebo, with identical appearance, was made from cane sugar and prescribed at the same dosage as the DJW.
Assessments
Clinical assessments were evaluated for base-line data at the end of a run-in period (week 0) and then weekly for 4 weeks. These assessments included 100-mm VAS that assessed pain (classified into walking pain, standing pain, pain during climbing up and down stairs, night pain, resting pain, total pain, pain during the most painful knee movement), 100-mm VAS that assessed stiffness (classified into morning stiffness, stiffness after rest and total stiffness), Lequesne's functional index that assessed the patient's daily activities (score ranging from 0–24) [14], and time for climbing up 10 steps. The participants self-rated the VAS and Lequesne's functional index, and they were allowed to view their own previously recorded scores.
At the end of week 1–4, 100-mm VAS that assessed the physician's and patients' overall opinions on improvement were also evaluated. The assessment forms were designed so that the patients and evaluator could view their own previously recorded scores, but they were not allowed to view each other's VAS. Clinical assessments were evaluated by the same physician who had been blinded to the treatment. Complete physical examination and non-directive questioning for adverse events were also performed weekly for 4 weeks in order to acquire a safety assessment.
Statistical analysis
In within the group analysis, the mean VAS and Lequesne's functional index between base-line and the following weeks were compared by a non-parametric Wilcoxon's signed-rank test, whereas, the average time for climbing up 10 steps was compared by the paired t-test.
In the analysis between the groups, a non-parametric Wilcoxon's rank-sum test was used to determine whether the two groups differed in the physician's and patients' overall opinions on improvement. In addition, the mean changes in VAS that assessed pain and stiffness, as well as Lequesne's functional index were compared by the same test. The student's t-test was used to compare the mean changes in the time for climbing up 10 steps.
Results
A total of 429 patients were recruited into this study, of whom 229 were excluded (Figure 1). The remaining 200 patients were randomized into the DJW and diclofenac group, 100 patients per group. In the DJW group, 4 patients withdrew from the study due to ineffectiveness (n = 3) and transportation problem (n = 1), 1 patient was lost to follow up and another one had a traffic accident during the study. In the diclofenac group, 3 patients were lost to follow up and 3 were withdrawn due to accidents. Thus, each group comprised 94 completers. The two treatment groups were not significantly different in demographic data e.g., sex, age, weight, height, duration of OA, location of OA (Table 2) and base-line data for the major outcome assessment (VAS, Lequesne's functional index and time for climbing up 10 steps). The radiographic findings at entry (Table 3) were not different between both groups. During the study, the rates of compliance with medication in the DJW group were 94%, whereas, those in the diclofenac group were 96%. Since few patients withdrew from the trial, the results were not substantially affected, whether the statistical method was performed by an intention to treat (ITT) analysis or an analysis on available completers. Thus, the following data showed the findings from the ITT analysis.
Figure 1 Flow chart of patients who participated in the clinical trial. 1DJW group received DJW plus placebo of diclofenac. 2Diclofenac group received diclofenac plus placebo of DJW.
Table 2 Demographic data and base-line data for the major outcome assessments of participants evaluated at the end of a run-in period (week 0).
Treatment groups
Characteristics DJW Diclofenac p value
n (M:F) 100 (22:78) 100 (19:81) NS
Age (y)* 62.66 (9.46) 62.38 (8.22) NS
Body weight (kg)* 60.47 (10.34) 60.13 (10.89) NS
Height (m)* 1.51 (0.07) 1.51 (0.07) NS
BMI (kg/m2)* 26.52 (4.38) 26.35 (3.85) NS
Duration of OA (y)* 5.46 (5.48) 4.79 (4.24) NS
Localization of OA NS
Right knee 17 17
Left knee 14 14
Both knees 69 69
VAS the assessed pain (mm)*
Walking pain 64.53 (24.92) 64.78 (25.14) NS
Standing pain 52.42 (25.87) 53.52 (24.69) NS
Pain during climbing up and down stairs 63.08 (20.87) 62.69 (23.21) NS
Night pain 50.15 (26.74) 48.45 (28.18) NS
Resting pain 38.48 (22.09) 37.12 (26.08) NS
Total paina 268.65 (88.87) 266.55 (89.33) NS
Pain during the most painful knee movement 82.25 (16.15) 81.17 (16.56) NS
VAS that assessed stiffness (mm)*
Morning stiffness 53.53 (27.38) 58.32 (26.40) NS
Stiffness after rest 68.52 (22.76) 70.45 (22.32) NS
Total stiffnessb 122.05 (41.98) 128.76 (42.34) NS
Lequesne's functional index* 14.20 (3.13) 14.80 (2.61) NS
Time for climbing up 10 steps* 13.44 (4.85) 13.32 (5.10) NS
*Data represent mean (SD). aSummation of VAS that assessed walking pain, standing pain, pain during climbing up and down stairs, night pain and resting pain. bSummation of VAS that assessed morning stiffness and stiffness after rest. NS: no statistical significance.
Table 3 The radiographic findings at entry into the study.
Treatment groups
Radiographic findings DJW (169 knees) Diclofenac (169 knees) p value
Kellgren and Lawrence X-ray grade [20] NS
Grade 2 31 23
Grade 3 71 80
Grade 4 67 66
Knee compartment with most severe changes NS
Medial tibiofemoral 131 135
Lateral tibiofemoral 16 8
Patellofemoral 22 26
The VAS that assessed pain and stiffness at the end of week 1–4 decreased significantly when compared to their own base-line values (within the group analysis), as did Lequesne's functional index and time for climbing up 10 steps (Table 4). At the end of week 4, the percentages of improvement in VAS that assessed pain and stiffness were higher than 65% in both groups, whereas, the percentages of improvement in Lequesne's functional index and time for climbing up 10 steps were approximately 40% and 20%, respectively.
Table 4 Mean VAS that assessed pain and stiffness, Lequesne's functional index and time for climbing up 10 steps in intent-to-treat patients (n = 100/group).
Variable Treatment Group Week 0 Week 1 Week 2 Week 3 Week 4 % improvementa
VAS that assessed pain (mm)
Walking pain DJW 64.53 (24.92) 47.58* (25.33) 37.72* (25.00) 28.00* (23.25) 18.06* (20.76) 72.01
Diclofenac 64.78 (25.14) 44.08* (23.43) 34.99* (22.07) 24.21* (21.00) 14.31* (16.10) 77.91
Standing pain DJW 52.42 (25.87) 39.81* (26.09) 31.61* (24.89) 24.29* (22.98) 16.89* (20.59) 67.78
Diclofenac 53.52 (24.69) 37.60* (24.06) 28.19* (22.40) 21.12* (21.16) 12.86* (16.69) 75.97
Pain during climbing up and down stairs DJW 63.08 (20.87) 46.31* (26.56) 36.40* (25.67) 28.16* (24.03) 18.41* (21.50) 70.81
Diclofenac 62.69 (23.21) 43.90* (22.29) 32.61* (22.42) 24.59* (21.79) 15.83* (19.65) 74.75
Night pain DJW 50.15 (26.74) 33.44* (27.27) 23.56* (22.79) 15.68* (18.14) 9.27* (15.04) 81.52
Diclofenac 48.45 (28.18) 28.93* (22.82) 20.87* (19.56) 15.02* (17.87) 8.65* (14.68) 82.15
Resting pain DJW 38.48 (22.09) 27.25* (21.99) 19.96* (19.98) 12.64* (15.56) 7.42* (13.09) 80.72
Diclofenac 37.12 (26.08) 22.84* (20.62) 16.26* (18.19) 11.30* (16.40) 6.58* (13.96) 82.27
Total painb DJW 268.65 (88.87) 194.38* (105.06) 149.24* (103.19) 108.76* (92.54) 70.04* (83.94) 73.93
Diclofenac 266.55 (89.33) 177.34* (85.49) 132.91* (84.50) 96.21* (81.94) 58.23* (70.43) 78.15
Pain during the most painful knee movement DJW 82.25 (16.15) 63.31* (26.35) 49.77* (28.70) 37.69* (28.45) 26.81* (27.70) 67.40
Diclofenac 81.17 (16.56) 56.79* (24.87) 43.64* (27.30) 33.10* (27.17) 22.84* (25.85) 71.86
VAS that assessed stiffness (mm)
Morning stiffness DJW 53.53 (27.38) 36.61* (25.56) 28.04* (23.86) 19.66* (20.53) 12.34* (17.69) 76.95
Diclofenac 58.32 (26.40) 38.73* (23.87) 28.52* (21.93) 20.19* (20.23) 12.90* (17.34) 77.88
Stiffness after rest DJW 68.52 (22.76) 51.69* (24.93) 39.40* (25.31) 29.05* (24.60) 19.62* (23.06) 71.37
Diclofenac 70.45 (22.32) 49.71* (24.68) 39.54* (24.97) 28.23* (24.17) 18.90* (20.60) 73.17
Total stiffnessc DJW 122.05 (41.98) 88.30* (45.93) 67.44* (46.25) 48.71* (42.82) 31.96* (38.84) 73.81
Diclofenac 128.76 (42.34) 88.44* (43.84) 68.06* (43.03) 48.42* (41.97) 31.80* (36.07) 75.30
Lequesne's functional index (score) DJW 14.20 (3.13) 11.60* (4.11) 11.05* (4.04) 9.93* (4.40) 8.92* (4.60) 37.18
Diclofenac 14.80 (2.61) 10.89* (3.38) 10.65* (3.55) 9.59* (3.52) 8.64* (3.83) 41.62
Time for climbing up 10 steps (s) DJW 13.44 (4.85) 11.65* (4.75) 11.42* (4.67) 10.94* (4.73) 10.50* (4.38) 21.88
Diclofenac 13.32 (5.10) 11.26* (5.12) 11.14* (5.72) 10.61* (5.51) 10.18* (4.46) 23.57
Data represent mean (SD). aCalculated by (meanweek0-meanweek4) × 100/meanweek0. bSummation of VAS that assessed walking pain, standing pain, pain during climbing up and down stairs, night pain and resting pain. cSummation of VAS that assessed morning stiffness and stiffness after rest. *p < 0.05 versus base-line value.
When the statistical analysis between groups was performed, the mean changes in VAS that assessed pain during climbing up and down the stairs, night pain, resting pain, total pain, and time for climbing up 10 steps did not differ significantly between both groups (Table 5). Nonetheless, the mean changes in VAS that assessed walking pain, standing pain, and stiffness were significantly different during week 0–1, whereas, differences in mean changes in Lequesne's functional index were found during week 0–1 and 0–2. Afterwards, the mean changes in these variables became no different throughout the study.
Table 5 Mean changes of VAS that assessed pain and stiffness, Lequesne's functional index and time for climbing up 10 steps in intent-to-treat patients (n = 100/group).
Variable Treatment Group Week 0–1 Week 0–2 Week 0–3 Week 0–4
VAS that assessed pain (mm)
Walking pain DJW -16.96 (1.68) -26.82 (1.97) -36.54 (2.31) -46.48 (2.41)
Diclofenac -20.70† (1.60) -29.80 (1.95) -40.58 (2.26) -50.47 (2.38)
Standing pain DJW -12.61 (1.80) -20.81 (2.23) -28.13 (2.28) -35.53 (2.34)
Diclofenac -15.93† (1.33) -25.33 (1.83) -32.41 (2.04) -40.66 (2.25)
Pain during climbing up and down stairs DJW -16.78 (1.95) -26.68 (2.30) -34.93 (2.26) -44.67 (2.22)
Diclofenac -18.79 (1.40) -30.08 (1.91) -38.11 (2.04) -46.86 (2.35)
Night pain DJW -16.71 (2.32) -26.60 (2.25) -34.47 (2.42) -40.88 (2.59)
Diclofenac -19.52 (1.98) -27.58 (2.30) -33.43 (2.51) -39.80 (2.81)
Resting pain DJW -11.23 (1.24) -18.52 (1.46) -25.84 (1.86) -31.06 (2.02)
Diclofenac -14.28 (1.34) -20.86 (1.91) -25.82 (2.07) -30.54 (2.38)
Total paina DJW -74.27 (6.53) -119.42 (7.42) -159.90 (7.85) -198.61 (8.51)
Diclofenac -89.21 (5.25) -133.64 (7.02) -170.34 (7.65) -208.33 (9.03)
Pain during the most painful knee movement DJW -18.94 (2.11) -32.48 (2.63) -44.56 (2.77) -55.44 (2.67)
Diclofenac -24.38 (2.10) -37.53 (2.51) -48.07 (2.57) -58.33 (2.59)
VAS that assessed stiffness (mm)
Morning stiffness DJW -16.93 (1.98) -25.50 (2.24) -33.87 (2.46) -41.19 (2.58)
Diclofenac -19.59† (1.69) -29.80 (2.10) -38.13 (2.47) -45.42 (2.63)
Stiffness after rest DJW -16.83 (1.97) -29.12 (2.41) -39.48 (2.50) -48.91 (2.54)
Diclofenac -20.74† (1.72) -30.91 (2.07) -42.22 (2.29) -51.55 (2.40)
Total stiffnessb DJW -33.76 (3.48) -54.62 (4.02) -73.35 (4.21) -90.10 (4.27)
Diclofenac -40.33† (3.05) -60.71 (3.72) -80.35 (4.21) -96.97 (4.47)
Lequesne's functional index (score) DJW -2.60 (0.34) -3.15 (0.32) -4.28 (0.37) -5.29 (0.38)
Diclofenac -3.92† (0.31) -4.16† (0.32) -5.22 (0.36) -6.16 (0.40)
Time for climbing up 10 steps (s) DJW -1.79 (0.33) -2.02 (0.31) -2.50 (0.32) -2.94 (0.32)
Diclofenac -2.05 (0.31) -2.18 (0.34) -2.71 (0.34) -3.13 (0.33)
Data represent mean (SD). aSummation of VAS that assessed walking pain, standing pain, pain during climbing up and down stairs, night pain and resting pain. bSummation of VAS that assessed morning stiffness and stiffness after rest. †p < 0.05 versus the DJW group at the same duration of treatment.
The physician's and patients' overall opinions on improvement, as measured on VAS, are shown in Table 6. The physician's overall opinion on improvement at each time point did not significantly differ between the two groups. However, differences between groups (DJW versus diclofenac group) were found in the patients' overall opinion at week 1 (32.58 ± 23.18 versus 37.48 ± 18.59), but no differences were demonstrated at the remaining time-points.
Table 6 VAS that assessed physician's and patients' overall opinions on improvementa during treatment (intent-to-treat data set).
Variable Treatment group n Week 1 Week 2 Week 3 Week 4
Physician's overall opinion DJW 98b 56.69 (11.32) 57.30 (11.32) 60.06 (12.47) 62.55 (11.67)
Diclofenac 97b 59.84 (7.53) 59.63 (7.74) 62.11 (7.57) 63.35 (7.90)
Patients' overall opinion DJW 98b 32.58 (23.18) 45.53 (24.74) 58.10 (26.84) 71.13 (24.68)
Diclofenac 97b 37.48* (18.59) 50.24 (18.79) 62.88 (19.75) 75.30 (17.95)
Data represent mean (SD). a0 = no improvement, 100 = best possible improvement. b2 patients in the DJW group and 3 patients in the diclofenac group could not be assessed due to loss to follow up or withdrawal during week 0. *p < 0.05 versus the DJW group at the corresponding week.
The majority of patients in both groups experienced no adverse events (72% vs. 73% for DJW and diclofenac groups, respectively). All adverse events reported were mild in intensity in both groups. The most common adverse events occurring in the DJW and diclofenac group were raised blood pressure (16% vs. 19%), central nervous system symptoms including dizziness, somnolence and drowsiness (16% vs. 11%), and gastrointestinal symptoms including nausea/vomiting, dyspepsia, diarrhea and constipation (12% vs. 5%). The least common adverse events were increased appetite, cramp, rash and flu. More than one adverse events might be occurred in some patients. However, the percentages of patients who experienced each adverse event in both groups were not significantly different.
In summary, the VAS that assessed pain and stiffness, Lequesne's functional index and time for climbing up 10 steps at each time point decreased significantly in the DJW and diclofenac group when compared to their own base-line values. The mean changes in all VAS that assessed pain, except those for walking and standing, did not differ significantly between both groups. The differences in mean changes in the VAS that assessed walking pain, standing pain and stiffness were found only during week 0–1, whereas, those in Lequesne's functional index were found during week 0–1 and 0–2.
Discussion
Since the preparations and dosages of DJW and diclofenac were different, this study was designed as a randomized, double dummy, controlled trial in order to completely blind both patients and physician (double-blind). Therefore, the placebo of DJW was also prescribed for the patients in the diclofenac group, and vice versa, the placebo of diclofenac was prescribed for the patients in the DJW group.
Among the 15 herbs used as raw materials in DJW, Xixin (Herba Asari) seemed to be the most toxic, due to its pungent taste and warm property [16]. Generally, a large dose of this herb is not recommended in a tropical country (such as Thailand) because of the potential aggravation of internal heat. Thus, the amount of Xixin in the DJW recipe used in this study was reduced from 7.75% to 2.5%.
In an ITT analysis (and analysis on completers), the mean changes in some variables between the two groups were significantly different after the first few weeks of treatment, and became no different afterwards. These differences suggest that the onset of DJW is significantly slower than diclofenac for at least 2 weeks (with respect to walking pain, standing pain, morning stiffness, stiffness after rest, total stiffness and patients' overall opinion) or 3 weeks (with respect to Lequesne's functional index). The reason why DJW needs a few weeks to exert its effect may be due to 3 possibilities. Firstly, from the pharmacokinetic point of view, the elimination half-life of the active ingredients in DJW might be too long, and therefore needs weeks to accumulate until a steady state concentration is reached (normally 4–5 half-lives) and its maximal therapeutic effect is evident. Secondly, from the pharmacodynamic point of view, DJW may exert its effects via several probable mechanisms (similar to many novel biologic treatments of arthropathy) involved modifications of cartilage metabolism, normalized viscosity and elasticity of synovial fluid, etc. These mechanisms of action might resemble many symptomatic slow acting drugs in osteoarthritis (SYSADOA) such as glucosamine sulfate, intra-artricular hyaluronan, and others. These interventions always need a period of time to exert their therapeutic action. Thirdly, the major effect of DJW might be the result from placebo effect and/or natural fluctuation of the OA symptoms. It could be simply that diclofenac worked quickly, but patients in both groups got better anyway by 2–3 weeks. Although the last possibility cannot be entirely ruled out, but it seems unlikely because even there is a tendency of OA symptoms to improve after placebo treatment, it has been reported that diclofenac was significantly superior to placebo in relieving pain, improving stiffness, and improving physical function after 4 weeks of treatment [17]. Furthermore, we also found that oral administration of the ethanol extract of DJW possessed both central and peripheral analgesic activities in animal model, even when the DJW extract was given in the equivalent dose used in human (mg/kg of human dose corrected by intra- and inter-specie variations) [to be published data]. In clinical practice, this slower onset of action and probable need for rescue analgesics (e.g., paracetamol as needed) during the first 2–3 weeks after initiation of DJW should be the important limitations of using DJW as an alternative treatment for OA of the knee. Moreover, the patient's compliance with such a high dosage of DJW (9 g/day or 18 capsules/day) is an important issue to be concerned.
Since this study demonstrated that approximately 30% of study subjects in each group experienced adverse events, this data suggest that the toxicity profiles of DJW are similar to diclofenac. Therefore, cautious use of DJW should be considered in the same manner as using diclofenac including other NSAIDs. However, the gastrointestinal adverse effects in the diclofenac group were quite low when compared to other short-term NSAIDs studies [18,19]. This might be due to the exclusion of patients with a high risk of adverse effects from NSAIDs during the screening visit. Since the relief of joint pain afforded by paracetamol is comparable with that achievable with NSAIDs, paracetamol merits a trial as initial therapy, based on its overall cost, efficacy, and toxicity profile [3,7]. In this circumstance, the rather high rate of adverse events from DJW should be another limitation of using DJW as an alternative, especially to paracetamol, in symptomatic treatment of OA of the knee.
Conclusion
DJW demonstrates clinically comparable efficacy to diclofenac after 4 weeks of treatment. However, the slow onset of action as well as approximately equal rate of adverse events to diclofenac might limit its alternative role in treatment of OA of the knee.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ST carried out the randomization, supervised data collection and analysis, and drafted the manuscript. PK participated in the design of the study and performed the statistical analysis. NR participated in the selection of patients eligible for the study. KS carried out the outcome assessments. SP participated in the report of radiographic findings of knee. SL participated in the preparation of DJW and its placebo. SP carried out the screening for microorganism contamination in DJW and its placebo. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by Faculty of Medicine, Chiang Mai University, Thailand. The authors would like to acknowledge Assoc. Prof. Dr. Pornngam Limtrakul for her valuable suggestion concerning the screening for aflatoxin contamination in DJW and its placebo by the ELISA technique.
==== Refs
Felson DT An update on the pathogenesis and epidemiology of osteoarthritis Radiol Clin North Am 2004 42 1 9 15049520
Drury PL Shipley M Kumar P, Clark M Rheumatology and bone disease In Clinical medicine 1998 4 London: WB Saunders 447 578
Hochberg MC Altman RD Brandt KD Clark BM Dieppe PA Griffin MR Moskowitz RW Schnitzer TJ Guidelines for the medical management of osteoarthritis. Part II. Osteoarthritis of the knee. American College of Rheumatology Arthritis Rheum 1995 38 1541 6 7488273
Hungin AP Kean WF Nonsteroidal anti-inflammatory drugs: overused or underused in osteoarthritis? Am J Med 2001 8 11 10.1016/S0002-9343(00)00628-8
Wegman A van der Windt D van Tulder M Stalman W de Vries T Nonsteroidal antiinflammatory drugs or acetaminophen for osteoarthritis of the hip or knee? A systematic review of evidence and guidelines J Rheumatol 2004 31 344 54 14760807
Brooks PM Potter SR Buchanan WW NSAIDs and osteoarthritis-help or hindrance? J Rheumatol 1982 9 3 5 7086778
Bradley JD Brandt KD Katz BP Kalasinski LA Ryan SI Comparison of an antiinflammatory dose of ibuprofen, an analgesic dose of ibuprofen, and acetaminophen in the treatment of patients with osteoarthritis of the knee N Engl J Med 1991 325 87 91 2052056
Buffum M Buffum JC Nonsteroidal anti-inflammatory drugs in the elderly Pain Manag Nurs 2000 1 40 50 11706458 10.1053/jpmn.2000.7779
Bolesta S Haber SL Hepatotoxicity associated with chronic acetaminophen administration in patients without risk factors Ann Pharmacother 2002 36 331 3 11847957 10.1345/aph.1A035
McLaughlin JK Lipworth L Chow WH Blot WJ Analgesic use and chronic renal failure: a critical review of the epidemiologic literature Kidney Int 1998 54 679 86 9734593 10.1046/j.1523-1755.1998.00043.x
Xu XC Yuan JR Tian JZ Chen SM Commonly used Chinese patent medicines 1994 Beijing: Higher Education Press
Geng JY Huang WQ Ren TC Ma XF Practical traditional Chinese medicine & pharmacology: Herbal Formulas 1997 Beijing: New World Press
Analysis of Chinese herb prescriptions for rheumatoid arthritis
Lequesne MG Mery C Samson M Gerard P Indexes of severity for osteoarthritis of the hip and knee validation-value in comparison with other assessment tests Scand J Rheumatology 1987 85 9
Thai Food and Drug Admistration, Ministry of Public Health Handbook of herbal products for community-directed economy 2000 Bangkok: Veteran Administration Publishing House
Asarum herb (Xixin)
Schnitzer TJ Beier J Geusens P Hasler P Patel SK Senftleber I Gitton X Moore A Sloan VS Poor G Efficacy and safety of four doses of lumiracoxib versus diclofenac in patients with knee or hip primary osteoarthritis: a phase II, four-week, multicenter, randomized, double-blind, placebo-controlled trial Arthritis Rheum 2004 51 549 57 15334426 10.1002/art.20525
Valat JP Accardo S Reginster JY Wouters M Hettich M Lieu PL A comparison of the efficacy and tolerability of meloxicam and diclofenac in the treatment of patients with osteoarthritis of the lumbar spine Inflamm Res 2001 30 4
Gagnier P Review of the safety of diclofenac/misoprostol Drug 1993 31 5
Kellgren JH Lawrence JS Radiological assessment of osteoarthritis Ann Rheum Dis 1957 16 494 502 13498604
| 15588333 | PMC539359 | CC BY | 2021-01-04 16:31:45 | no | BMC Complement Altern Med. 2004 Dec 13; 4:19 | utf-8 | BMC Complement Altern Med | 2,004 | 10.1186/1472-6882-4-19 | oa_comm |
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-4-221558833210.1186/1472-6947-4-22Research ArticleModification of the mean-square error principle to double the convergence speed of a special case of Hopfield neural network used to segment pathological liver color images Sammouda Rachid [email protected] Mohamed [email protected] Dept. of Computer Science, University of Sharjah, Sharjah, UAE2 Dept. of Computer & Information Science, Prince Sultan University, Riadh, Saudi Arabia2004 12 12 2004 4 22 22 14 3 2004 12 12 2004 Copyright © 2004 Sammouda and Sammouda; licensee BioMed Central Ltd.2004Sammouda and Sammouda; 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 analyzes the effect of the mean-square error principle on the optimization process using a Special Case of Hopfield Neural Network (SCHNN).
Methods
The segmentation of multidimensional medical and colour images can be formulated as an energy function composed of two terms: the sum of squared errors, and a noise term used to avoid the network to be stacked in early local minimum points of the energy landscape.
Results
Here, we show that the sum of weighted error, higher than simple squared error, leads the SCHNN classifier to reach faster a local minimum closer to the global minimum with the assurance of acceptable segmentation results.
Conclusions
The proposed segmentation method is used to segment 20 pathological liver colour images, and is shown to be efficient and very effective to be implemented for use in clinics.
==== Body
Background
Segmentation is an important step in most applications that use medical image data. For example, segmentation is a prerequisite for quantification of morphological disease manifestations and for radiation treatment planning [1,2], for construction of anatomical models [3], for definitions of flight paths in virtual endoscopies [4], for content-based retrieval by structure [5], and for volume visualization of individual objects [2].
A Number of algorithms based on approaches such as histogram analysis, regional growth, edge detection and pixel classification have been proposed in other articles of medical image segmentation. In recent years, Artificial Neural Networks (ANNs) have been proposed as an attractive alternative solution to a number of pattern recognition problems. In our previous works [6], we have explored the potential of a Special Case of Hopfield Neural Network (SCHNN) in segmenting cerebral images obtained using the Magnetic Resonance Imaging (MRI) technique.
Hopfield network for the optimization applications consists of many interconnected neuron elements. The network minimizes an energy function of the form:
where N is the number of neurons, Vk is the output of the kth neuron, Ik is the bias term, and Tkl is the interconnection weight between the kth and lth neurons. The energy function used in the segmentation problem is slightly different from the one defined by Hopfield and the arguments are given in [7].
The results that have been obtained in [6] were preferable to those obtained using Boltzmann Machine (BM) and the conventional ISODATA clustering technique. Also, in [8] we have shown that SCHNN is also able to make crisp segmentation of pathological liver colour images. However, during our study attempt to improve the segmentation process, we found that SCHNN segmentation results depend strongly on some parameters in the energy function formulating the classification problem. A summery of this study follows.
Methods
The segmentation problem of an image of N pixels is formulated in [8] as a partition of the N pixels among M classes, such that the assignment of the pixels minimizes a criterion function. The SCHNN classifier structure consists of a grid of N × M neurons with each row representing a pixel and each column representing a cluster. The network classifies the image of N pixels of P features among M classes, in a way that the assignment of the pixels minimizes the following criterion function:
where Rkl is the Mahalanobis distance measure between the kth pixel and the centroid of class l, Rkl is also equivalent to the error committed when a pixel k is assigned to a class l. The index n in is the power or weight of the considered error in the energy function of the segmentation problem, and Vkl is the output of the klth neuron. Nkl is a N × M vector of independent high frequency white noise source used to avoid the network being trapped in early local minimums. The term c(t) is a parameter controlling the magnitude of noise which is selected in a way to provide zero as the network reaches convergence. The minimization is achieved by using SCHNN and by solving the motion equations satisfying:
where Ukl is the input of the kth neuron, and μ(t) is a scalar positive function of time, used as heuristically motivated stopping criterion of SCHNN, and is defined as in [6] by:
β(t) = t(Ts - t) (4)
where t is the iteration step, and Ts is the pre-specified convergence time of the network which has been found to be 120 iterations [6]. The network classifies the feature space, without teacher, based on the compactness of each cluster calculated using Mahalanobis distance measure between the kth pixel and the centroid of class l given by:
where Xk is the P-dimensional feature vector of the kth pixel (here P = 3 with respect to the RGB color space components), is the P-dimensional centroid vector of class l, and Σl is the covariance matrix of class l. The segmentation algorithm is described as follows [8].
Step 1 Initialize the input of the neurons to random values.
Step 2 Apply the following input-output relation, establishing the assignment of each pixel to only and only one class.
Step 3 Compute the centroid and the covariance matrix Σl of each class l as follows:
where nl is the number of pixels in class l, and the covariance matrix is then normalized by dividing each of its elements by .
Step 4 Update the inputs of each neuron by solving the set of differential equations in (2) using Eulers approximation:
Step 5 if t <Ts, repeats from Step 2, else terminated.
For this study, a total of 20 liver tissue sections were provided by the pathological division of National Cancer Center in Tokyo. These sections were taken using needle biopsy, stained with hematoxylin and then magnified with an optical microscope. Figure 1 shows a true RGB color image of liver tissue of 768 × 512 pixels. We have used the above described SCHNN classifier with the image components in the R.G.B color space. The number of classes is fixed to five based on medical information. These classes are the contour of the image, the cell's nuclei, the cytoplasm, the fibrous tissues, and the class of both blood sinus and fat cells.
Figure 1 A sample of pathological liver colour image in true colour (Red, Green, and Blue). The cells nuclei are represented by a circular shape in dark purple colour, the cytoplasm regions are coloured purple, the circular objects in white represent the fat cells, and the remaining objects in wave shape and white colour represent the fibrous tissues and blood sinus. The contour of the image is black.
Figure 2 shows the curves of SCHNN energy function during the segmentation of the sample shown in Figure 1 with Ts values between 30 and 120 iterations. Similar curves were obtained for the rest of the images of the dataset. As it is illustrated in Figure 2. The curve corresponding to Ts = 120 iterations gives the optimal solution, the same as it is with MRI data [6].
Figure 2 SCHNN energy function curves during the segmentation of the sample shown in Figure 1 using different values of the pre-specified convergence time Ts.
In order to study the effect of the weight of the Mahalanobis distance Rkl in the cost function (2), we have provided a simple modification to the above algorithm as follows:
Step 1 Use the same random initialization N × M matrix, as input of the neurons, when minimizing the energy function (1) with different error's weight n.
This condition is added to the algorithm in order to make sure that the random field does not have any effect on the generated results.
Step 2 trough Step 5 remain the same.
Results
Figure 3 shows different curves of the optimization of the energy function of the segmentation of the sample shown in Figure 1 using SCHNN with the above modification (Step 1) with respect to different values of the variable n in equation (2). As aforementioned, the pre-specified convergence time of SCHNN is fixed to Ts = 120 iterations. However, we can clearly see from Figure 3 that with a higher value of n in Equation (2), the same convergence point or a close position is reached in half the time of the one reached with n = 2 and Ts = 120 iterations. So, this raises the following question: what is the type of relation between the variable n in (2) and the pre-specified convergence time Ts?
Figure 3 Shows different curves of the optimization of the energy function of the segmentation, of the sample in Figure 1, by considering different values of the variable n in equation (2) and with pre-specified convergence Time Ts = 120 iterations.
Before answering this question, it is essential to know at this level what is the best value of n that corresponds to the optimum solution with Ts = 120 iterations. From Figure 4, it can be seen that n = 6 gives the optimum solution with Ts = 120. Similar figures to Figure 3 and Figure 4 were obtained with the rest of the images in the dataset.
Figure 4 This curve is extracted from Figure 3, it connects the convergence values of the energy function of the segmentation problem, of the sample in Figure 1, by considering different values of the variable n in Equation (2), with the same random initialization matrix for all n values, and with a pre-specified convergence time Ts = 120 iterations.
Discussion
Analysis of the pre-specified convergence time effect
In order to study the effect of the pre-specified time, we repeated the above experiments with different Ts values. We realized that each value of Ts corresponds to a value of n, in Equation (2). When both (Ts and n) used together they give a local optima in the energy landscape of SCHNN. Figure 5 shows the curves linking the convergence values of SCHNN with respect to the value of n in Equation (2) that are obtained with Ts values 120, 60, and 30. We realized that the curves corresponding to Ts = 120 and Ts = 60 intersect in their optimum solutions obtained with n = 6, and the two curves are similar when n is in the range 5–10. However, the curve corresponding to Ts = 30, shows higher error at convergence of all values of n.
Figure 5 Curves of the energy function of SCHNN at convergence with respect to values of the variable n in Equation (2) and for different pre-specified convergence time Ts. The green and red curves correspond to Ts = 60 and Ts = 120, respectively, are almost identical when n is in the range 5–10.
Analysis of the SCHNN random initialization effect
In order to see the effect of the random initialization on the results of the algorithm described in section 3, we have executed the same algorithm with different initialization matrices and the curves of the convergence values of SCHNN corresponding to these initializations are shown in Figure 6. As it is clear from the curves, in Figure 6, the random initialization does not have effect on the variable n in (2) when it takes the value of six where SCHNN gives an optimum and acceptable results that agree with the pathological experts point of views. However, with other values of n, the random initialization may affect the solution of the problem, or in other words, may affect the error of the SCHNN at convergence as shown in Figure 6.
Figure 6 Curves of the energy function of SCHNN at convergence with respect to values of the variable n in Equation (2) and for different initialization matrices.
Conclusions
We analyzed the effect of considering the mean-square error in formulating the segmentation problem of multidimensional medical images. We have shown, empirically, that considering an integer power equal to six, of the error in the energy function of the problem, helped SCHNN to converge twice as fast as the same optimal solution obtained with the mean-square error algorithm. This result is promising to make our segmentation method useful for a Computer Aided Diagnosis (CAD) system for liver cancer and the like. In our future work, we will study deeply the effect of the random initialization and its effect on the segmentation result and on the SCHNN classifier.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Rachid carried out the theoretical study, the sequence alignment and drafted the manuscript. Mohammed participated in the design of the study and performed the analysis and helped to draft the manuscript. All authors read and approved the final manuscript.
Figure 7 Segmentation result of the sample in Figure 1, obtained using SCHNN in optimizing equation (1) with n = 2, and a pre-specified convergence time Ts = 120 iterations. The cells nuclei are represented by a circulate shape with white colour, surrounded by the red regions representing the cytoplasm of the cells, fat cells are coloured blue, and the fibrous tissues and blood sinus are coloured green.
Figure 8 Segmentation result of the sample liver pathological image in Figure 1, obtained using SCHNN in optimizing equation (2) with n = 6, and a pre-specified convergence time Ts = 120 iterations.
Figure 9 Segmentation result of the sample liver pathological image in Figure 1, obtained using SCHNN in optimizing equation (2) with n = 6, and a pre-specified convergence time Ts = 60 iterations.
Figure 10 Segmentation result of the sample liver pathological image in Figure 1, obtained using SCHNN in optimizing equation (2) with n = 12, and a pre-specified convergence time Ts = 30 iterations
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors want to thank the Research Center at the University of Sharjah for supporting this work. Also the authors thank Dr. Maher Moussa in the English Department of the University of Sharjah for editing this paper.
==== Refs
Chaney E Pizer S Defining anatomical structures from medical images Semin Radiat Oncol 1992 2 215 225 10717038
Tracton G Chaney E Rosenman J Pizer S Mask: Combining 2-D and 3-D segmentation methods to enhance functionality SPIE Conf Medical Imaging Bellingham, WA 1994 98 109
Brinkley JF A flexible, generic model for anatomic shape: Application to intyeractive two-dimensional medical image segmentation and matching Computer Biomed Res 1993 26 121 142 10.1006/cbmr.1993.1008
Lorensen W Ferenc A Kikinis R The exploration of cross-sectional data with a virtual endoscope Interactive Technology and the New Paradigm for Health Care, Japan 1995 IOS Press 221 230
Orphanoudakis SC Chronaki C Kostomanolakis S I/Sup 2/C: A system for the indexing, storage, and retrieval of medical images by content Med Inform 1994 19 109 122
Sammouda R Niki N Nishitani H A Comparison of Hopfield Neural Network and Boltzmann Machine in Segmenting MR Images of the Brain IEEE Transactions on Nuclear Science 1996 43 3361 3368 10.1109/23.552753
Amartur SC Piraino D Takefuji Y "Optimization neural networks for the segmentation of magnetic resonance images," IEEE Transactions on Medical Imaging 1992 11
Sammouda M Sammouda R Niki N Liver cancer detection system based on the analysis of digitized color images of tissue samples obtained using needle biopsy International Journal of Information Visualization 2002 1 Pagrave Press 130 138 10.1057/palgrave.ivs.9500012
| 15588332 | PMC539360 | CC BY | 2021-01-04 16:36:20 | no | BMC Med Inform Decis Mak. 2004 Dec 12; 4:22 | utf-8 | BMC Med Inform Decis Mak | 2,004 | 10.1186/1472-6947-4-22 | oa_comm |
==== Front
Cancer Cell IntCancer Cell International1475-2867BioMed Central London 1475-2867-4-71555506710.1186/1475-2867-4-7Primary Researchα1-antitrypsin and its C-terminal fragment attenuate effects of degranulated neutrophil-conditioned medium on lung cancer HCC cells, in vitro Zelvyte Inga [email protected] Tim [email protected] Ulla [email protected] Sabina [email protected] Lund University, Department of Medicine and Otholaryngology, University Hospital Malmo, 20502 Malmo, Sweden2004 21 11 2004 4 7 7 5 3 2004 21 11 2004 Copyright © 2004 Zelvyte et al; licensee BioMed Central Ltd.2004Zelvyte 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
Tumor microenvironment, which is largely affected by inflammatory cells, is a crucial participant in the neoplastic process through promotion of cell proliferation, survival and migration. We measured the effects of polymorphonuclear neutrophil (PMN) conditioned medium alone, and supplemented with serine proteinase inhibitor α-1 antitrypsin (AAT) or its C-terminal fragment (C-36 peptide), on cultured lung cancer cells.
Methods
Lung cancer HCC cells were grown in a regular medium or in a PMN-conditioned medium in the presence or absence of AAT (0.5 mg/ml) or its C-36 peptide (0.06 mg/ml) for 24 h. Cell proliferation, invasiveness and release of IL-8 and VEGF were analyzed by [3H]-thymidine incorporation, Matrigel invasion and ELISA methods, respectively.
Results
Cells exposed to PMN-conditioned medium show decreased proliferation and IL-8 release by 3.9-fold, p < 0.001 and 1.3-fold, p < 0.05, respectively, and increased invasiveness by 2-fold (p < 0.001) compared to non-treated controls. In the presence of AAT, PMN-conditioned medium loses its effects on cell proliferation, invasiveness and IL-8 release, whereas VEGF is up-regulated by 3.7-fold (p < 0.001) compared to controls. Similarly, C-36 peptide abolishes the effects of PMN-conditioned medium on cell invasiveness, but does not alter its effects on cell proliferation, IL-8 and VEGF release. Direct HCC cell exposure to AAT enhances VEGF, but inhibits IL-8 release by 1.7-fold (p < 0.001) and 1.4-fold (p < 0.01) respectively, and reduces proliferation 2.5-fold (p < 0.01). In contrast, C-36 peptide alone did not affect these parameters, but inhibited cell invasiveness by 51.4% (p < 0.001), when compared with non-treated controls.
Conclusions
Our data provide evidence that neutrophil derived factors decrease lung cancer HCC cell proliferation and IL-8 release, but increase cell invasiveness. These effects were found to be modulated by exogenously present serine proteinase inhibitor, AAT, and its C-terminal fragment, which points to a complexity of the relationships between tumor cell biological activities and local microenvironment.
==== Body
Background
The relationship between inflammation, innate immunity and cancer is widely accepted. Early in the neoplastic process, inflammatory cells and their released molecular species influence the growth, migration and differentiation of all cell types in the tumor microenvironment, whereas later in the tumorigenic process neoplastic cells also divert inflammatory mechanisms, such as metalloproteinase production, and chemokine/cytokine functions to favour tumor spread and metastasis [1-3]. Human polymorphonuclear neutrophils (PMN) comprise 50–70% of circulating leukocytes and induce inflammatory reactions that can be either cytotoxic for tumour cells or promote tumour growth and metastasis [4,5]. For example, animal studies show that circulating neutrophils isolated from tumor-bearing animals reduce the number of metastatic foci in the lung. Other studies reveal that neutrophils stimulate tumour cell attachment to endothelial monolayers [6,7] and neutrophil-released cytokines, growth and angiogenic factors enhance tumour growth and spread [8]. Most of the neutrophil-induced tumour-promoting effects are attributed to their abilities to release proteases. Neutrophil degranulation results in the release of serine proteases, such as elastase, cathepsin G and protease-3, which may contribute to the activation of matrix metalloproteinases (MMPs) that mediate tumor cell invasiveness [9-12]. Recently, Schwartz and co-workers found that neutrophil-secreted factors can activate pro-MMP-2, which is important in ECM degradation and tumor cell invasion [14]. It is now also apparent that tumor cells can be sources of proteolytic enzymes themselves. For example, some cells possess immunoreactive neutrophil elastase [15], which is suggested to play a role in increasing tumor cell invasiveness of surrounding tissue [16]. Levels of immunoreactive neutrophil elastase in tumour extracts are reported to be an independent prognostic factor for patients with breast cancer and non-small cell lung cancer [17-19].
A large number of studies support the notion that proteases play an important role in the progression of malignant tumors. Therefore, the expression of proteinase inhibitors is considered to be an antimalignant event [20]. Alpha-1 antitrypsin (AAT), a major inhibitor of human serine proteases in serum, is produced mainly by the liver, but also by extrahepatic cells, including neutrophils and certain cancer cells [21,22]. AAT is an acute phase protein and its concentration rises up to 3–4-fold above normal during acute inflammation [23]. Several types of cancer, including non-small cell lung adenocarcinoma, have been associated with increased serum levels of AAT [20,24]. Clinical studies have shown that high circulating levels of AAT directly correlate with tumour progression [24-26]. Recently we also found that plasma levels of AAT are significantly elevated in lung cancer patients and, particularly in cases with metastases [27]. The role of AAT is poorly understood in tumor diseases although it is suggested that it can inhibit tumor cell growth and invasion, in vitro [28]. Moreover, not only the native, inhibitory form of AAT, but also conformationally modified, non-inhibitory forms are suggested to play a role in modulating tumour growth and invasiveness. For example, Kataoka and co-workers have shown that the C-terminal fragment of AAT can enhance the growth and invasiveness of human pancreas adenocarcinoma cells, in vivo [29]. Our in vitro studies revealed that the C-terminal fragment of AAT, corresponding to amino acid sequence 358–396 (C-36 peptide), induces breast tumour cell proliferation and invasiveness [30]. Recently we found that C-36 peptide induces neutrophil chemotaxis, adhesion, degranulation and superoxide generation, in vitro [31], which we propose may indirectly affect tumor cell biological activities, in vivo. Together these findings prompted us to design an experimental model in vitro that would allow us to evaluate how neutrophil-derived products themselves and in combination with exogenously added native AAT or its C-36 peptide influence lung cancer HCC cell growth, invasion and metastatic properties.
Results
HCC cell proliferation
To examine the effect of the degranulated PMN-conditioned media alone or supplemented with exogenously added native AAT (0.5 mg/ml) or its C-36 peptide (0.06 mg/ml) on lung cancer HCC cell proliferation, we measured DNA biosynthesis using a [3H]-thymidine incorporation assay (Fig. 1). HCC cells cultured in PMN-conditioned medium for 24 h decrease proliferation by 3.9-fold, (p < 0.001), while under the same experimental conditions cells exposed to PMN-conditioned medium supplemented with native AAT show no changes in proliferation compared to control cells cultured in a regular media. HCC cells exposed directly to native AAT decrease [3H] thymidine incorporation by 2.5-fold, (p < 0.01) compared to control cells cultured in a regular medium. Thus, our data show that both PMN-conditioned medium and AAT loose their inhibitory effects on cancer cell proliferation when used in combination. Under the same experimental conditions, the C-36 peptide of AAT added either directly to HCC cells or as a supplement with PMN-conditioned medium did not exert any significant effects on HCC cell proliferation relative to controls (Fig. 1).
Figure 1 [3H]Thymidine incorporation assay. The HCC cells were exposed to AAT (0.5 mg/ml) or its C-36 peptide (0.06 mg/ml) in a regular medium or in a PMN-conditioned medium for 20 h following addition of [3H] thymidine for 4 h. Each bar represents mean ± standard deviation from three separate experiments with three repeats in each. *** p < 0.001, ** p < 0.01.
HCC cell invasiveness
To characterize lung cancer cell properties after their exposure to PMN-conditioned medium with and without added AAT or its C-36 peptide, we performed cell invasiveness assays. As illustrated in Fig. 2 and Table 1, cells cultured in a PMN-conditioned medium show markedly increased cell invasiveness (by 56.8%, p < 0.001) compared to controls. When PMN-conditioned medium is supplemented with native AAT or its C-36 peptide, its effects on cancer cell invasiveness are diminished by 41.5%, and 77%, (p < 0.001), respectively, compared to PMN medium alone. It must be pointed out that HCC cells cultured in a regular medium in the presence of C-36 peptide decreased cell invasion by 51.4%, (p < 0.001), while native AAT showed slight, but not significant up-regulation of cell invasiveness, compared to controls. Together these data show that native AAT as well as its C-36 peptide abolish the capacity of PMN-conditioned medium to stimulate HCC cell invasiveness.
Figure 2 HCC cell invasiveness after exposure to PMN-conditioned media, AAT or C-36 peptide separately and in combinations. HCC cells cultured in a regular medium (I) and in a PMN-conditioned medium (II) alone or supplemented with AAT or C-36 peptide (this figure represents one of three independent experiments).
Table 1 Quantitative evaluation of HCC cell invasiveness (migrated cells/per view, in ten randomly selected fields).
Stimulus HCC cells in a regular medium HCC cells in PMN-conditioned medium
mean♠) SEM mean SEM
Control (medium) 27.8 ± 0.85 43.6*** ± 1.6
AAT (0.5 mg/ml) 32.7 ± 2.1 25.5*** ± 1.99
C-36 (0.06 mg/ml) 13.5*** ± 1.5 10.0*** ± 0.86
♠ mean and standard error of 3 independent experiments; *** -p < 0.001
Release of Vascular Endothelial Growth Factor (VEGF) and interleukin-8 (IL-8)
Vascular endothelial growth factor (VEGF) and chemokine IL-8 are factors produced by tumor cells as well as by neutrophils [32-34], and are known to correlate with lung cancer angiogenesis in vivo [35]. We analysed for VEGF and IL-8 released by HCC cells alone or cultured in PMN-conditioned medium with and without added AAT or C-36 peptide for 24 h (Fig. 3 and 4). As shown in Fig 3, AAT increases VEGF release by 1.7-fold, p < 0.001, and inhibits IL-8 by 1.4-fold, p < 0.01, relative to control cells. Under the same experimental conditions, cancer cells cultured in a PMN-conditioned medium supplemented with AAT increase VEGF and IL-8 release by 3.7-fold, p < 0.001 and 1.6-fold, p < 0.01, respectively, compared to PMN-condition medium alone. In addition, HCC cells in a PMN-conditioned medium show a slight (~27%), but not significant, decrease in VEGF levels compared to cells grown in regular medium. It must be noted that C-36 peptide had no influence on VEGF and IL-8 levels under any experimental conditions used (data not shown).
Figure 3 VEGF release from HCC cells alone or exposed to PMN-conditioned medium with and without addition of AAT. Each bar represents the mean ± standard deviation of six repeats from two separate experiments with three repeats in each. *** p< 0.001
Figure 4 IL-8 release from HCC cells alone or exposed to PMN-conditioned medium with and without addition of AAT. Each bar represents the mean ± standard deviation of six repeats from two separate experiments with three repeats in each. ** p < 0.01, * p < 0.05.
Enzymatic activity in PMN-conditioned medium
Neutrophil degranulation results in a large release of proteases, and to verify this we monitored proteolytic activity in PMN-conditioned medium by zymography, an electrophoretic technique used for the qualitative evaluation of proteases. As illustrated in Fig. 5 and 6 line-1, PMN-conditioned medium reveals a specific pattern of gelatinolytic and caseinolytic activity at a position between 72 and 180 kDa, which most likely represents neutrophil collagenase (MMP-8) (75 kDa), gelatinase A (MMP-2) (72 kDa), gelatinase B (MMP-9) (92 kDa) and several MT-MMPs, such as MT-MMP-2, -5 (72–75 kDa). Lung cancer cells grown in regular medium (Fig. 5 and 6, line-7) or treated with AAT and C-36 peptide show no gelatinolytic (Fig. 5, lines 6, 5) and caseinolytic activity (Fig. 6, lines-5, 3). PMN-conditioned medium incubated with HCC cells for 24 h manifested a decrease in certain enzymatic activities (Fig. 5, line 4, Fig. 6, line 6) compared to PMN media alone (Fig 5 and 6, line-1). However, no further changes in enzymatic activity profiles were observed when HCC cells were grown in PMN-conditioned medium supplemented with native AAT (Fig. 5, line 3 and 6 line-5) or C-36 peptide (Fig 5 and 6, line-2).
Figure 5 Effects of PMN-conditioned medium alone and supplemented with AAT or C-36 peptide on HCC cell gelatinolytic activity. Lane 1-PMN-conditioned medium alone; lanes 2 and 3, PMN-conditioned medium supplemented with C-36 peptide and AAT, respectively; and incubated with cancer cells for 24 h, lane 4-PMN-conditioned medium incubated with HCC cells alone; lanes 5 and 6 – HCC cells exposed to C-36 peptide and AAT, respectively, in a regular medium; 7-HCC cells alone. This figure represents 1 of 3 separate experiments performed under the same experimental conditions.
Figure 6 Effects of PMN-conditioned medium alone and supplemented with AAT or C-36 peptide on HCC cell caseinolytic activity. Lane 1-PMN-conditioned medium alone; lanes 2-PMN-conditioned medium supplemented with C-36 peptide and incubated with HCC cells for 24 h, line 3-HCC cells exposed to C-36; lane 4-; PMN-conditioned medium supplemented with AAT and incubated with HCC cells for 24 h, lane 5-HCC cells exposed to AAT; lane 6-PMN-conditioned medium incubated with HCC cells; lane 7-HCC cells alone. This figure represents 1 of 3 separate experiments performed under the same experimental conditions.
Molecular profile of AAT in HCC cell culture supernatants
Cell culture supernatants collected from HCC cells cultured in a regular medium or in PMN-conditioned medium with and without addition of AAT were analysed for AAT by 7.5% SDS-PAGE followed by immunoblotting using polyclonal antibody against human AAT. As shown in Fig. 7, line-2, AAT added to HCC cells was unchanged in amount and form compared to native AAT alone (Fig. 7, line-5), whereas AAT added as a supplement into PMN-conditioned medium is altered in the distribution of its forms: in addition to monomeric AAT, a complexed form of AAT can be detected (Fig. 7, lane-4).
Figure 7 HCC cell culture medium alone and in the presence of AAT studied by Western blot analysis. Cell culture supernatants were applied to 10% SDS-PAGE and immunoblotted with polyclonal antibody against AAT. M, molecular size markers (myosin-205 000, β-galactosidase-123 000, bovine serum albumin-79 000, carbonic anhydrase-45 700), lane 1-HCC cells alone, lane 2-HCC cells stimulated with AAT, lanes 3 and 4-HCC cells cultured in PMN-conditioned medium alone and supplemented with AAT, respectively lane 5-ATT done. The arrow indicates complexed AAT. The AAT profile shown in this figure represents one of three similar experiments.
Discussion
Local and systemic inflammatory mediators derived by tumor-associated PMN and tumor cells are suggested to play a role in the development of lung tumors [36,37]. Different studies have reported an increase of proteolytic activity in cancer and have suggested a role of proteases in tumor progression and metastasis [38-40]. For example, Guner and co-workers have shown that the plasma concentration of elastase in patients with malignant lung carcinoma is 10-fold higher compared to benign lesions of the lungs [41]. An increased concentration of neutrophil elastase is found to be closely associated with progression of non-small cell lung cancer (NSCL). In parallel, an increase in levels of proteinase inhibitors, such as AAT, has also been reported in tumor cases, including lung tumors. These elevated levels of proteinase inhibitors are attributed to the inflammatory reactions accompanying the tumorigenesis [42,43].
PMN, the most abundant circulating blood leukocytes, are potent effectors of inflammation and release a wide profile of serine and metalloproteinases [44]. Infiltration of tumors with PMN is associated with a favourable prognosis in some studies in humans, however for individual patients there is no predictable relationship between PMN infiltration and cancer prognosis [45,46]. Thus, in the present study we aimed to investigate how degranulated neutrophil-conditioned medium alone and supplemented with the serine proteinase inhibitor, AAT, or its C-terminal peptide (C-36), affects functional activities of a non-small cell lung adenocarcinoma cell line (HCC), in vitro.
We found that PMN-conditioned medium expresses multiple effects on lung cancer HCC cell functional activities (decreases in cell proliferation and IL-8 release and stimulation of cell invasion through the ECM membrane), but has no significant effects on VEGF release. These dual neutrophil activities toward cancer cells have been described by other investigators. It has been suggested that through the release of cytokines, chlorinated oxidants and defensins, neutrophils may cause direct tumor killing [47,48] while in parallel, through release of serine proteinases and MMPs, breakdown of basement membranes and ECM and increase in tumor invasiveness are promoted [49]. Our findings that PMN-conditioned medium decreases HCC cancer cell proliferation and IL-8 release, but increases invasiveness, in part support the hypothesis that neutrophils can both, inhibit and promote tumor cell growth and invasiveness. In parallel experiments, we found that PMN-conditioned medium supplemented with AAT loses its effects on HCC cell invasion, proliferation and IL-8 release, but enhances VEGF. In contrast, C-36 peptide added to PMN-conditioned medium abolished the effects of medium on cell invasiveness, but had no significant influence on medium effects on cell proliferation, IL-8 and VEGF release.
The major function of AAT is to inhibit neutrophil derived serine proteases, particularly neutrophil elastase [50]. A local increase of proteinases and PMN elastase-AAT complexes has been demonstrated in bronchoalveolar lavage from patients with lung cancer [51]. AAT is also known to interact with other components of degranulated-PMN. For example, AAT binds to defensins and neutralizes their effects on cell cytotoxicity and migration [52]. In our in vitro model, AAT added to degranulated PMN-conditioned medium occurs in a complexed form. Thus, the modulating effects of AAT on PMN-medium activities toward HCC cancer cells can be attributed to AAT alone, but also to its interaction(s) with components of the medium. Verification of the effects of AAT on separate components released by degranulated PMN needs further investigations.
It is important to point out that AAT alone as well as in combination with PMN-conditioned medium induced VEGF release from HCC cells. VEGF is one of the most potent angiogenic molecules, regulating both angiogenesis and vascular permeability, and hence promotes tumor progression and development in NSCLC [53]. Many compounds, including anti-VEGF antibody and anti-VEGF receptor antibody have been developed as VEGF inhibitors and these compounds were reported to inhibit growth of a wide variety of tumor cell lines in vitro and in animal models, in vivo [54]. Therefore, our in vitro findings that AAT expresses potent effects on VEGF release allow classify AAT as a tumor promoter. Contrary data exist for a relationship between the levels of AAT and cancer advancement. Some investigators propose that AAT plays a protective role in tumorogenesis. For example, Finlay and co-workers showed that AAT might be directly involved in breast cancer progression by acting as a tumor suppressor [55]. Yavelow and co-workers have shown that AAT inhibits human breast cancer cells MCF-7 growth [56] and AAT was also shown to block the activation of pro MMP-2 and tumor cell invasion [11]. However, immunohistochemical studies revealed that patients with AAT-positive colon, gastric and lung adenocarcinomas had a worse prognosis than AAT-negative ones [57-59]. Together these findings suggest that AAT may play multiple roles in cancerogenesis in vivo, in addition to its role as proteinase inhibitor.
During tumor progression, inflammatory cells produce large amounts of IL-8, an autocrine growth factor [53,60]. Levels of IL-8 have also been shown to correlate with lung cancer angiogenesis. Current knowledge of the degranulation process of neutrophils in vitro suggests that the chemokine IL-8 promotes rapid release of all neutrophil granular stores after cell exposure to cytoskeleton-disrupting agents [61]. In the present study, we found that the amount of IL-8 released by HCC can be reduced by AAT or PMN-conditioned medium alone, whereas these inhibitory effects are abolished when AAT and PMN-conditioned media are added together. Based on our findings, we hypothesize that IL-8 and other angiogenic factors released from HCC cells are regulated by PMN components, which may potentially initiate tumor cell activities or induce anti-tumor responses dependent on the tumor cell microenvironment.
AAT is the main inhibitor of neutrophil elastase and proteinase 3 [62] and it has been reported that an imbalance between AAT and neutrophil elastase may predispose to lung cancer development [63]. Patients who carry the deficiency allele of AAT have a significantly higher risk of developing squamous cell or bronchoalveolar carcinoma of the lungs [64]. On the other hand it was reported that strong expression of AAT in lung adenocarcinoma correlates with poor prognosis [65], although it was not shown whether the inhibitory activity of AAT was normal and whether elastase levels were high among this group of patients. Moreover, AAT is a good substrate for MMPs: neutrophil collagenase, gelatinase-B, stromelysin-1 and -3, and matrilysin can effectively cleave AAT [66-69]. The cleavages by these MMPs occurs at peptide bonds within AAT active site loop, resulting in a generation of cleaved forms of AAT. Comparative proteome analysis performed to identify protein alterations in plasma of prostate, lung and breast-cancer patients showed significant elevation of AAT and its N-terminal fragment [70], which points to a role for different molecular forms of AAT in cancer progression. Recent studies provide evidence that the C-terminal fragment of AAT may enhance tumor growth and invasiveness in vitro and in vivo [29,30]. We found that the C-36 peptide displays striking concentration-dependent pro-inflammatory effects on human neutrophils, including induction of neutrophil chemotaxis, adhesion, degranulation and superoxide generation [31]. Interestingly, in our HCC cancer cell model, C-36 peptide reduced HCC cell invasiveness and also abolished PMN-conditioned medium-induced cell invasiveness, whereas there were no significant effects on other parameters measured, such as cell proliferation, VEGF and IL-8 release.
Conclusions
Our data show that AAT in various molecular forms expresses differential effects on lung tumour cell responses and provide an experimental evidence for complexity in the interactions of neutrophil-released molecular species and lung cancer cells.
Materials and methods
AAT and its C-terminal peptide
Native, purified human plasma AAT was purchased from the Clinical Chemistry Department (UMAS, Malmö). The quality of AAT preparations was confirmed by 7.5% SDS-PAGE and determination of anti-elastase activity as described by Gaillard et al. [71]. Synthetic C-terminal fragment of AAT (C-36, corresponding to residues 359–394) was obtained from Saveen, Biotech AB (Denmark) and was greater than 98% purity. The C-36 peptide was prepared in sterile, endotoxin-free Tris buffered saline (0.015 M Tris, pH 7.4 containing 0.15 M NaCl) just before use. The endotoxin content in AAT was tested by quantitative E-TOXATE Assay (Sigma, USA). According to international standards no more than 0.08 enzyme U/ml of endotoxin is allowed to be present in AAT solutions [72]. In our experimental model we used AAT preparation with endotoxin levels below 0.05 enzyme U/ml. If necessary, AAT samples were purified using Detoxi-Gel™ endotoxin removing gel (Pierce, Rockford, IL, USA) according to the manufacturers recommendations.
Neutrophil isolation
Neutrophils were isolated from the peripheral blood of healthy donors using Polymorphprep™ (Axis-Shield PoC AS, Oslo, Norway) according to the manufacturers recommendations. Neutrophils were harvested as the lower cellular band above the red cell pellet. Residual erythrocytes were removed by a hypotonic lysis using ice cold 0.2% NaCl (w/v) for 30 s, followed addition of an equal volume of 1.6% NaCl to restore isotonicity. The neutrophil purity was typically 90% as determined by an AC900EO AutoCounter (Swelab Instruments, Sweden) and cell viability exceeded 95% according to trypan blue staining.
Preparation of PMN-conditioned medium
PMN-conditioned medium was prepared by degranulating neutrophils (5 × 106 cells/ml) according to Videm and Strand [72]. Briefly, isolated neutrophils were incubated on a shaker for 1 h at 37°C and immediately after on ice for 3 min, and centrifuged at 240 × g for 10 min at 4°C. Supernatants were collected and analyzed for the cytokine release and enzyme activity.
Cell culture
Human non-small cell lung carcinoma (HCC) cells were established from the lung of 54-year old women with non-small cell lung the adenocarcinoma type (DSMC No. ACC 534). The cells were routinely grown at 37°C in a humid air containing 5% CO2, in 125-cm2 flasks in RPMI-1640 medium, supplemented with 10% FBS, 100 IU/ml penicillin, and 100 μg/ml streptomycin. Medium was changed every 2–3 days. Prior to experiments HCC cells were seeded into 12-wells plates at a density of 3 × 105 cells/ml, and cultured until confluent. After, cells were washed and a serum-free medium containing test substances was added for 24 h. Two sets of experiments were designed: the first, in which HCC cells were exposed to native AAT (0.5 mg/ml) or C-36 peptide (0.06 mg/ml) in a RPMI-1640 and the second, in which the cells were cultured in PMN-conditioned medium alone or supplemented with native AAT or C-36 peptide.
[3H] Thymidine incorporation assay
HCC cells were incubated with test substances for 20 h. [3H] Thymidine was then added (0.2 μCi/ml) for a further 4 h at 37°C. The medium was then aspirated, the cells were washed twice with 0.5 M NaCl and incubated for 5 min with 5% trichloroacetic acid. The cells were then washed with distilled water, dissolved in 1 ml 0.5 M NaOH, neutralised with 200 μl HCl and radioactivity determined in a β-counter (Packard 300CD liquid scintillation spectrometer; Packard Instruments). Protein concentration in cell lysates was determined by the Lowry method using HSA as a standard.
Matrigel invasion assay
Cell invasion assay was performed in an invasion chamber, a 24-well tissue culture plate with 12 inserts. The inserts contain an 8 μm pore size polycarbonate filters over which is placed a thin layer of ECMatrix™. HCC cells were suspended in a serum free medium containing 1 × 106 cells/ml alone or together with test substances. Each suspension was added to the upper chamber of an ECM Invasion system. Medium containing 10% of fetal bovine serum (FBS), serving as a chemoattractant, was added to the lower chamber. The chambers were incubated at 37°C in 5% CO2 for 24 h. Invasive cells were stained and counted using microscope (Olympus BX41, PC program Olympus MicroImage) at a 100 × magnification. The number of cells per field, from 10 randomly selected fields, are presented.
Enzyme activity assay
HCC cells were cultured alone or stimulated with test substances for 24 h. Cell free supernatants were then analysed by zymography using 10% and 12% Tris-Glycine gels containing 0.1% gelatin or β-casein, respectively (NOVEX, Invitrogen Life Technologies, UK). Briefly, supernatants were diluted in a sample buffer and separated by electrophoresis at 125 V for 90 min. The gels were then renatured and developed over night in a developing buffer (NOVEX, Invitrogen Life Technologies, UK) at 37°C. After washing, the gels were stained with Coomassie Blue R-250. Proteases that can utilize casein or gelatine as a substrate showed up as clear zones in the gels.
Endothelial growth factor (VEGF) and IL-8 analysis
HCC cell supernatants were analysed for the VEGF and IL-8 levels by using commercially available quantitative ELISA kits (R&D systems, Minneapolis, USA) according to manufacturers instructions. The lowest detectable concentration for VEGF and IL-8 was 15.6 pg/ml and 10 pg/ml, respectively.
Immunoblotting
HCC cells supernatants were analysed by 10% SDS-PAGE gels and when transferred to a polivinylidene fluoride (PVDF) membrane (Millipore, Millipore Corporation, Bedford, MA 01730) using a semi-dry immunoblot transfer system. The blot was visualised using polyclonal rabbit antibody to human AAT (1:500) (DAKO, A/S, Denmark), secondary horseradish peroxidase-conjugated swine anti-rabbit antibody (1:800) (DAKO, A/S, Denmark) and peroxidase substrate, DAB (3,3-diaminobenzidine tetrahydrochloride) (Sigma, USA).
Statistics
The differences in the means of experimental results were analysed for their statistical significance with the one-way ANOVA combined with a multiple-comparisons procedure (Scheffe multiple range test), with an overall significance level of α = 0.05. Statistical Package (SPSS for Windows, release 11.0) was used for the statistical calculations.
List of abbreviations
AAT – alpha1-antitrypsin; C-36 – C-terminal fragment of AAT; FBS – fetal bovine serum; NSCLC – non small-cell lung cancer; MMPs – matrix metalloproteinases; MT-MMP – membrane type matrix metalloproteinase; PBS – phosphate buffered saline; PMN – polymorphonuclear neutrophil; VEGF – vascular endothelial growth factor.
Authors contributions
Inga Zelvyte – AB, Tim Stevens – JY, Ulla Westin – JY, Sabina Janciauskiene ES, FG
Acknowledgements
This work was supported by grants from Tore Nilsson Foundation, Swedish Research Council, Astra Zeneca R&D Lund, and Medical Faculty Lund University.
==== Refs
Coussens LM Inflammation and cancer Nature 2002 420 860 867 12490959 10.1038/nature01322
Carney DN The biology of lung cancer Curr Opin Pulm Med 1995 1 271 277 9363064
Liotta LA Kohn EC The microenvironment of the tumour-host interface Nature 2001 411 375 379 11357145 10.1038/35077241
van Kessel KP van Strijp JA van Kats-Renaud HJ Miltenburg LA Fluit AC Verhoef J Uncoupling of oxidative and non-oxidative mechanisms in human granulocyte-mediumted cytotoxicity: use of cytoplasts and cells from chronic granulomatous disease patient J Leukoc Biol 1990 48 359 366 2168467
Simpson-Haidaris PJ Rybarczyk B Tumors and fibrinogen. The role of fibrinogen as an extracellular matrix protein Ann N Y Acad Sci 2001 936 406 425 11460495
Ishihara Y Iijima H Matsunaga K Contribution of cytokines on the suppression of lung metastasis Biotherap 1998 11 267 275 10.1023/A:1008070025561
Starkey JR Liggitt HD Jones W Hosick HL Influence of migratory blood cells on the attachment of tumor cells to vascular endothelium Int J Cancer 1984 34 535 543 6490205
McCawley LJ Matrisian LM Tumor progression: defining the soil round the tumor seed Curr Biol 2001 11 R25 27 11166192 10.1016/S0960-9822(00)00038-5
Okada Y Nakanishi I Activation of matrix metalloproteinase 3 (stromelysin) and matrix metalloproteinase 2 ('gelatinase') by human neutrophil elastase and cathepsin G FEBS Lett 1989 249 353 356 2544455 10.1016/0014-5793(89)80657-X
Schleiffenbaum B Moser R Patarroyo M Fehr J The cell surface glycoprotein Mac-1 (CD11b/CD18) mediates neutrophil adhesion and modulates degranulation independently of its quantitative cell surface expression J Immunol 1989 142 3537 3545 2565926
Shamamian P Pocock BJ Schwartz JD Monea S Chuang N Whiting D Marcus SG Galloway AC Mignatti P Neutrophil-derived serine proteinases enhance membrane type-1 matrix metalloproteinase-dependent tumor cell invasion Surgery 2000 127 42 47 10.1067/msy.2000.101155
Maeda H Wu J Okamoto T Maruo K Akaike T Kallikrein-kinin in infection and cancer Immunopharmacology 1999 43 115 128 10596842 10.1016/S0162-3109(99)00104-6
Van de Louw A Jean D Frisdal E Cerf C d'Ortho MP Baker AH Lafuma C Duvaldestin P Harf A Delclaux C Neutrophil proteinases in hydrochloric acid- and endotoxin-induced acute lung injury: evaluation of interstitial protease activity by in situ zymography Lab Invest 2002 82 133 145 11850527
Schwartz JD Shamamian P Monea S Whiting D Marcus SG Galloway AC Mignatti P Activation of tumor cell matrix metalloproteinase-2 by neutrophil proteinases requires expression of membrane-type 1 matrix metalloproteinase Surgery 1998 124 232 238 9706143 10.1067/msy.1998.90566
Yamashita JI Ogawa M Ikei S Omachi H Yamashita SI Saishoji T Nomura K Sato H Production of immunoreactive polymorphonuclear leucocyte elastase in human breast cancer cells: possible role of polymorphonuclear leucocyte elastase in the progression of human breast cancer Br J Cancer 1994 69 72 76 8286213
Inada M Yamashita J Ogawa M Neutrophil elastase inhibitor (ONO-5046-Na) inhibits the growth of human lung cancer cell lines transplanted into severe combined immunodeficiency (scid) mice Res Commun Mol Pathol Pharmacol 1997 97 229 232 9344234
Foekens JA Ries C Look MP Gippner-Steppert C Klijn JG Jochum M The prognostic value of polymorphonuclear leukocyte elastase in patients with primary breast cancer Cancer Res 2003 63 337 341 12543785
Yamashita J Tashiro K Yoneda S Kawahara K Shirakusa T Local increase in polymorphonuclear leukocyte elastase is associated with tumor invasiveness in non-small cell lung cancer Chest 1996 109 1328 1334 8625686
Yamashita J Ogawa M Abe M Hayashi N Kurusu Y Kawahara K Shirakusa T Tumor neutrophil elastase is closely associated with the direct extension of non-small cell lung cancer into the aorta Chest 1997 111 885 890 9106565
Kataoka H Itoh H Koono M Emerging multifunctional aspects of cellular serine proteinase inhibitors in tumor progression and tissue regeneration Pathol Int 2002 52 89 102 11940213 10.1046/j.1440-1827.2002.01320.x
Paakko P Kirby M du Bois RM Gillissen A Ferrans VJ Crystal RG Activated neutrophils secrete stored alpha 1-antitrypsin Am J Respir Crit Care Med 1996 154 1829 1833 8970377
du Bois RM Bernaudin JF Paakko P Hubbard R Takahashi H Ferrans V Crystal RG Human neutrophils express the alpha 1-antitrypsin gene and produce alpha 1-antitrypsin Blood 1991 77 2724 2730 2043769
Molmenti EP Ziambaras T Perlmutter DH Evidence for an acute phase response in human intestinal epithelial cells J Biol Chem 1993 268 14116 14124 7686149
Nash DR McLarty JW Fortson NG Pretreatment, prediagnosis immunoglobulin, and alpha 1-antitrypsin levels in patients with bronchial carcinoma J Natl Cancer Inst 1980 64 721 724 6966002
Bata J Colobert L Biron A Brune J Study of various serum proteins in lung cancer. Immunoglobulins A, G, M, haptoglobin, alpha-1-antitrypsin, alpha-2-macroglobulin Ann Biol Clin (Paris) 1977 35 297 303 74964
Micksche M Kokron O Serum levels of alpha1-antitrypsin and alpha2-macroglobulin in lung cancer Osterr Kneipp Mag 1977 3 116 119
Zelvyte I Wallmark A Piitulainen E Westin U Janciauskiene S Increased Plasma Levels of Serine Proteinase Inhibitors in Lung Cancer Patients Anticancer Res 2004
Yavelow J Tuccillo A Kadner SS Katz J Finlay TH Alpha 1-antitrypsin blocks the release of transforming growth factor-alpha from MCF-7 human breast cancer cells J Clin Endocrinol Metab 1997 82 745 752 9062476 10.1210/jc.82.3.745
Kataoka H Uchino H Iwamura T Seiki M Nabeshima K Koono M Enhanced tumor growth and invasiveness in vivo by a carboxyl-terminal fragment of alpha1-proteinase inhibitor generated by matrix metalloproteinases: a possible modulatory role in natural killer cytotoxicity Am J Pathol 1999 154 457 468 10027404
Zelvyte I Lindgren S Janciauskiene S Multiple effects of alpha1-antitrypsin on breast carcinoma MDA-MB 468 cell growth and invasiveness Eur J Cancer Prev 2003 12 117 124 12671535 10.1097/00008469-200304000-00005
Janciauskiene S Zelvyte I Jansson L Stevens T Divergent effects of alpha1-antitrypsin on neutrophil activation, in vitro Biochem Biophys Res Commun 2004 15 288 296 14766206 10.1016/j.bbrc.2004.01.055
Gupta MK Qin RY Mechanism and its regulation of tumor-induced angiogenesis World J Gastroenterol 2003 9 1144 1155 12800214
Werther K Bulow S Hesselfeldt P Jespersen NF Svendsen MN Nielsen HJ VEGF concentrations in tumour arteries and veins from patients with rectal cancer APMIS 2002 110 646 650 12529018 10.1034/j.1600-0463.2002.1100908.x
Werther K Christensen IJ Nielsen HJ Determination of vascular endothelial growth factor (VEGF) in circulating blood: significance of VEGF in various leucocytes and platelets Scand J Clin Lab Invest 2002 62 343 350 12387579 10.1080/00365510260296492
Masuya D Huang C Liu D Kameyama K Hayashi E Yamauchi A Kobayashi S Haba R Yokomise H The intratumoral expression of vascular endothelial growth factor and interleukin-8 associated with angiogenesis in nonsmall cell lung carcinoma patients Cancer 2001 92 2628 2638 11745198 10.1002/1097-0142(20011115)92:10<2628::AID-CNCR1616>3.0.CO;2-F
Di Carlo E Forni G Lollini PL Colombo MP Modesti A Musiani P The intriguing role of polymorphonuclear neutrophils in antitumor reactions Blood 2001 97 339 345 11154206 10.1182/blood.V97.2.339
Jakobisiak M Lasek W Golab J Natural mechanisms protecting against cancer Immunology Letters 2003 90 103 122 14687712 10.1016/j.imlet.2003.08.005
Choong PF Nadesapillai AP Urokinase plasminogen activator system: a multifunctional role in tumor progression and metastasis Clin Orthop 2003 Suppl 415 46 58
Freije JM Balbin M Pendas AM Sanchez LM Puente XS Lopez-Otin C Matrix metalloproteinases and tumor progression Adv Exp Med Biol 2003 532 91 107 12908552
Foekens JA Ries Ch Look MP Gippner-Steppert C Klijn JG Jochum M Elevated expression of polymorphonuclear leukocyte elastase in breast cancer tissue is associated with tamoxifen failure in patients with advanced disease Br J Cancer 2003 88 1084 1090 12671709 10.1038/sj.bjc.6600813
Guner G Kirkali G Baskin Y Karlikaya C Akkoclu A Elastase levels in small and non-small cell lung carcinoma Biochem Soc Trans 1993 Suppl 21 305
Ameshima S Ishizaki T Demura Y Imamura Y Miyamori I Mitsuhashi H Increased secretory leukoprotease inhibitor in patients with nonsmall cell lung carcinoma Cancer 2000 89 1448 1456 11013357 10.1002/1097-0142(20001001)89:7<1448::AID-CNCR6>3.0.CO;2-Q
Umeki S Niki Y Soejima R Elastase/antielastase systems in pulmonary diseases Am J Med Sci 1988 296 103 106 2456693
Tani K Ogushi F Shimizu T Sone S Protease-induced leukocyte chemotaxis and activation: roles in host defense and inflammation J Med Invest 2001 48 133 141 11694952
Caruso RA Bellocco R Pagano M Bertoli G Rigoli L Inferrera C Prognostic value of intratumoral neutrophils in advanced gastric carcinoma in a high-risk area in northern Italy Mod Pathol 2002 15 831 837 12181268 10.1097/01.MP.0000020391.98998.6B
Nagtegaal ID Marijnen CA Kranenbarg EK Mulder-Stapel A Hermans J van de Velde CJ van Krieken JH Local and distant recurrences in rectal cancer patients are predicted by the nonspecific immune response; specific immune response has only a systemic effect – a histopathological and immunohistochemical study BMC Cancer 2001 1 7 11481031 10.1186/1471-2407-1-7
Di Carlo E Modesti A Coletti A Colombo MP Giovarelli M Forni G Diodoro MG Musiani P Interaction between endothelial cells and the secreted cytokine drives the fate of an IL4- or an IL5-transduced tumour J Pathol 1998 186 390 397 10209488 10.1002/(SICI)1096-9896(199812)186:4<390::AID-PATH194>3.3.CO;2-Q
van Kessel KP Verhoef J A view to a kill: cytotoxic mechanisms of human polymorphonuclear leukocytes compared with monocytes and natural killer cells Pathobiology 1990 58 249 264 2076189
Nielsen BS Timshel S Kjeldsen L Sehested M Pyke C Borregaard N Dano K 92 kDa type IV collagenase (MMP-9) is expressed in neutrophils and macrophages but not in malignant epithelial cells in human colon cancer Int J Cancer 1996 65 57 62 8543396 10.1002/(SICI)1097-0215(19960103)65:1<57::AID-IJC10>3.0.CO;2-F
Stockley RA Proteases and antiproteases Novartis Found Symp 2001 234 189 199 11199096
Pajdak W Sieradzak-Fleituch M Dubin A Owsinski J Alpha-1-antitrypsin and alpha-1-antitrypsin-neutrophil elastase complex in bronchoalveolar lavage fluid of patients with pulmonary diseases (pilot study) Acta Med Hung 1991 48 103 109 1813851
Panyutich AV Hiemstra PS van Wetering S Ganz T Human neutrophil defensin and serpins form complexes and inactivate each other Am J Respir Cell Mol Biol 1995 12 351 357 7873202
Masuya D Huang C Liu D Kameyama K Hayashi E Yamauchi A Kobayashi S Haba R Yokomise H The intratumoral expression of vascular endothelial growth factor and interleukin-8 associated with angiogenesis in nonsmall cell lung carcinoma patients Cancer 2001 92 2628 2638 11745198 10.1002/1097-0142(20011115)92:10<2628::AID-CNCR1616>3.0.CO;2-F
Yihai C Angiogenesis inhibitors and their therapeutic potentials Adv Exp Med Biol 2003 532 109 120 12908553
Finlay TH Tamir S Kadner SS Cruz MR Yavelow J Levitz M alpha 1-Antitrypsin- and anchorage-independent growth of MCF-7 breast cancer cells Endocrinology 1993 133 996 1002 8365378 10.1210/en.133.3.996
Yavelow J Tuccillo A Kadner SS Katz J Finlay TH Yavelow J Tuccillo A Kadner SS Katz J Finlay TH Alpha 1-antitrypsin blocks the release of transforming growth factor-alpha from MCF-7 human breast cancer cells J Clin Endocrinol Metab 1997 82 745 752 9062476 10.1210/jc.82.3.745
Fukushima M Fukuda Y Kawamoto M Yamanaka N Elastosis in lung carcinoma: immunohistochemical, ultrastructural and clinical studies Pathol Int 2000 50 626 635 10972861 10.1046/j.1440-1827.2000.01103.x
Fucich LF Cheles MK Thung SN Gerber MA Marrogi AJ Primary vs metastatic hepatic carcinoma. An immunohistochemical study of 34 cases Arch Pathol Lab Med 1994 118 927 930 7521638
Tahara E Ito H Taniyama K Yokozaki H Hata J Alpha 1-antitrypsin, alpha 1-antichymotrypsin, and alpha 2-macroglobulin in human gastric carcinomas: a retrospective immunohistochemical study Hum Pathol 1984 15 957 964 6207098
Taniguchi K Yang P Jett J Bass E Meyer R Wang Y Deschamps C Liu W Polymorphisms in the Promoter Region of the Neutrophil Elastase Gene Are Associated with Lung Cancer Development Clinical Cancer Research 2002 8 1115 1120 11948122
Arenberg DA Polverini PJ Kunkel SL Shanafelt A Hesselgesser J Horuk R Strieter RM The role of CXC chemokines in the regulation of angiogenesis in non-small cell lung cancer J Leukoc Biol 1997 62 554 562 9365108
Rainger GE Rowley AF Nash GB Adhesion-Dependent Release of Elastase From Human Neutrophils in a Novel, Flow-Based Model: Specificity of Different Chemotactic Agents Blood 1998 92 4819 4827 9845549
Travis J Salvesen GS Human plasma proteinase inhibitors Annu Rev Biochem 1983 52 655 709 6193754 10.1146/annurev.bi.52.070183.003255
Olsen GN Gangemi JD Bronchoalveolar lavage and the immunology of primary lung cancer Chest 1985 87 677 683 3886317
Higashiyama M Doi O Kodama K Yokouchi H Tateishi R An evaluation of the prognostic significance of alpha-1-antitrypsin expression in adenocarcinomas of the lung: an immunohistochemical analysis Br J Cancer 1992 65 300 302 1739634
Liu Z Zhou X Shapiro SD Shipley JM Twining SS Diaz LA Senior RM Werb Z The serpin alpha1-proteinase inhibitor is a critical substrate for gelatinase B/MMP-9 in vivo Cell 2000 102 647 655 11007483 10.1016/S0092-8674(00)00087-8
Michaelis J Vissers MC Winterbourn CC Cleavage of alpha 1-antitrypsin by human neutrophil collagenase Matrix 1992 1 80 81 1480096
Zhang Z Winyard PG Chidwick K Murphy G Wardell M Carrell RW Blake DR Proteolysis of human native and oxidised alpha 1-proteinase inhibitor by matrilysin and stromelysin Biochim Biophys Acta 1994 1199 224 228 8123672
Pei D Majmudar G Weiss SJ Hydrolytic inactivation of a breast carcinoma cell-derived serpin by human stromelysin-3 J Biol Chem 1994 269 25849 25855 7523394
Vejda S Posovszky C Zelzer S Peter B Bayer E Gelbmann D Schulte-Hermann R Gerner C Plasma from cancer patientsfeaturing a characteristic protein composition mediumtes protectionagainst apoptosis Mol Cel Proteomics 2002 1 387 393 10.1074/mcp.M200004-MCP200
Gaillard MC Kilroe-Smith TA Nogueira C Dunn D Jenkins T Fine B Kallenbach J Alpha-1-protease inhibitor in bronchial asthma: phenotypes and biochemical characteristics Am Rev Respir Dis 1992 145 1311 1315 1595996
Guideline on validation of the lumulus amebocyte lysate test as an end-product endotoxin test for human and animal parenteral drugs, biological products and medical devices, US Department of Health and Services, FDA 1987
Videm V Strand E Changes in Neutrophil Surface-Receptor Expression After Stimulation with FMLP, Endotoxin, Interleukin-8 and Activated Complement Compared to Degranulation Scandinavian Journal of Immunology 2004 59 25 33 14723618 10.1111/j.0300-9475.2004.01351.x
| 15555067 | PMC539361 | CC BY | 2021-01-04 16:40:09 | no | Cancer Cell Int. 2004 Nov 21; 4:7 | utf-8 | Cancer Cell Int | 2,004 | 10.1186/1475-2867-4-7 | oa_comm |
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-2-701558830810.1186/1477-7525-2-70ResearchThe use of bootstrap methods for analysing health-related quality of life outcomes (particularly the SF-36) Walters Stephen J [email protected] Michael J [email protected] Medical Statistics Group, School of Health and Related Research, University of Sheffield, UK2004 9 12 2004 2 70 70 9 8 2004 9 12 2004 Copyright © 2004 Walters and Campbell; licensee BioMed Central Ltd.2004Walters and Campbell; 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.
Health-Related Quality of Life (HRQoL) measures are becoming increasingly used in clinical trials as primary outcome measures. Investigators are now asking statisticians for advice on how to analyse studies that have used HRQoL outcomes.
HRQoL outcomes, like the SF-36, are usually measured on an ordinal scale. However, most investigators assume that there exists an underlying continuous latent variable that measures HRQoL, and that the actual measured outcomes (the ordered categories), reflect contiguous intervals along this continuum.
The ordinal scaling of HRQoL measures means they tend to generate data that have discrete, bounded and skewed distributions. Thus, standard methods of analysis such as the t-test and linear regression that assume Normality and constant variance may not be appropriate. For this reason, conventional statistical advice would suggest that non-parametric methods be used to analyse HRQoL data. The bootstrap is one such computer intensive non-parametric method for analysing data.
We used the bootstrap for hypothesis testing and the estimation of standard errors and confidence intervals for parameters, in four datasets (which illustrate the different aspects of study design). We then compared and contrasted the bootstrap with standard methods of analysing HRQoL outcomes. The standard methods included t-tests, linear regression, summary measures and General Linear Models.
Overall, in the datasets we studied, using the SF-36 outcome, bootstrap methods produce results similar to conventional statistical methods. This is likely because the t-test and linear regression are robust to the violations of assumptions that HRQoL data are likely to cause (i.e. non-Normality). While particular to our datasets, these findings are likely to generalise to other HRQoL outcomes, which have discrete, bounded and skewed distributions. Future research with other HRQoL outcome measures, interventions and populations, is required to confirm this conclusion.
Health Related Quality of LifeSF-36Bootstrap SimulationStatistical Analysis.
==== Body
1. Introduction
Health Related Quality of Life (HRQoL) measures are now frequently used in clinical trials and health services research, both as primary and secondary endpoints [1]. Investigators are now asking statisticians for advice on how to plan and analyse studies that have used HRQoL measures.
HRQoL measures such as the Short Form (SF)-36, Nottingham Health Profile (NHP) and European Organisation for Research and Treatment of Cancer (EORTC) QLQ-C30 are described in Fayers and Machin [1] and are usually measured on an ordered categorical (ordinal) scale. This means that responses to individual questions are usually classified into a small number of ordered response categories, e.g. poor, moderate and good. The responses are often analysed by assigning equally spaced numerical scores to the ordinal categories (e.g. 0 = 'poor', 1 = 'moderate' and 2 = 'good') and the scores across similar questions are then summed to generate a HRQoL score. These 'summated scores' are usually treated as if they were from a continuous distribution and were Normally distributed. We will also assume that there exists an underlying continuous latent variable, Z, that measures HRQoL (although not necessarily Normally distributed), and that the actual measured outcomes, X, are ordered categories that reflect contiguous intervals along this continuum.
This ordinal scaling of HRQoL measures means they generate data with discrete, bounded and non-standard distributions, which may lead to several problems in determining sample size and analysing the data [2,3]. Since HRQoL outcome measures may not meet the distributional requirements (usually that the data have a Normal distribution) for parametric methods of sample size estimation and analysis, conventional statistical advice would suggest that non-parametric methods be used to analyse HRQoL data.
The bootstrap [4,5] is a data based simulation method for estimating sample size [6] and analysing data: including hypothesis testing (p-values), standard error (SE) and confidence interval (CI) estimation; which involves repeatedly drawing random samples from the original data, with replacement. So, in theory, computer intensive methods such as the bootstrap that make no distributional assumptions may be more appropriate for estimating sample size and analysing HRQoL data than conventional statistical methods.
Conventional methods of analysis of HRQoL outcomes are extensively described in Fayers and Machin [1] and Fairclough [7]. They did not use the bootstrap to analyse HRQoL outcomes. As a consequence of this omission, the aim of this paper is to compare bootstrap computer simulation methods with standard methods of analysis of HRQoL measures (particularly the SF-36). We used the bootstrap for hypothesis testing, estimation of SEs and CIs for parameters, in four datasets (which illustrate the different aspects of study design). We then compared the bootstrap with standard methods of analysing HRQoL outcomes. These standard methods included: t-tests; multiple regression/analysis of covariance (ANCOVA) models fitted via ordinary least squares (OLS); response feature analysis or summary measures such as the Area Under the Curve (AUC) [8] and Generalised Linear regression Models (GLMs) [Pages 21–44, [9]] fitted using Generalised Estimating Equations (GEE) [10].
The remainder of this paper is structured into the following sections. The SF-36 HRQoL outcome is briefly described in Section 2. Section 2 also describes how the bootstrap can be used for hypothesis testing and confidence interval estimation. Section 2 ends with a description of the four example datasets. The results of conventional methods of analysis and bootstrap methods are compared in Section 3. Other issues, such as withdrawals and study sizes are discussed in Section 4. The final section (5) ends with a summary and conclusions.
2. Methods
The bootstrap
The term bootstrap derives from the phrase "to pull oneself up by one's bootstraps". The phrase is thought to be based on one of the eighteenth century Adventures of Baron Munchausen by Rudolph Erich Raspe. The Baron had fallen to the bottom of a deep lake. Just when it looked like all was lost, he thought to pick himself up by his own bootstraps [Page 5, [4]]!
The basic idea of the bootstrap involves repeated random sampling with replacement from the original data, to produce random samples of the same size of the original sample, each of which is known as a bootstrap sample, and each provides an estimate of the parameter of interest, e.g. mean. The "with replacement" means that any observation can be sampled more than once in each bootstrap sample. It is important because sampling without replacement would simply give a random permutation of the original data, with many statistics such as the mean being exactly the same [Page 115, [11]]. Repeating the process a larger number of times provides the required information on the variability of the estimator, since the standard error is estimated from the standard deviation of the statistics derived from the bootstrap samples.
Bootstrap observed value of the test statistic
The bootstrap is mainly used as a method for assessing statistical accuracy i.e. SE, biases and CIs. Throughout this paper we shall use the observed value of the test statistic or parameter estimate as our best guess at the true value of the unknown parameter or statistic. For example, if we are interested in estimating the population mean (from a random sample) it may seem that the best estimator of the mean of the population is the mean of all the bootstrap estimates. This is turns out not to be the case as the mean of the all the bootstrap means is biased. The observed sample mean, from the original data, is always the best estimate of the population mean. The same result applies for other statistics such as the median and regression coefficients.
Confidence Interval estimation
Suppose we wish to calculate a 95% confidence interval for a mean HRQoL from a sample. We take a random sample, with replacement from this data, of the same size as the original sample, and calculate the mean HRQoL of the data, in this bootstrap random sample. We do this repeatedly, a large number of times, say 1000. So we now have 1000 bootstrap samples, and 1000 estimates of the sample mean, one from each bootstrap sample. If these 1000 bootstrap sample means are ordered in increasing value, a bootstrap 95% confidence interval for the mean would be from the 25th to the 975th largest values. This is known as the percentile method and although it is an obvious choice, it is not the best method for bootstrapping confidence intervals, because it can have a bias, which one can estimate and correct for. This leads to methods such as the preferred bias corrected and accelerated (BCa) method [4,5]. Using the bootstrap method, valid bootstrap confidence intervals can be constructed for all common estimators such as the sample mean, median, proportion, difference in means, and difference in proportions. We estimated BCa bootstrap confidence intervals using the bootstrap procedure in STATA v8 [12].
According to Efron and Tibshirani [Page 180, [4]] each interval , where and are the lower and upper bounds of the interval respectively, can be described by its length and shape,
'Shape' measures the symmetry of the interval about the point estimate . The standard Normal based intervals are symmetrical about , and hence have shape = 1.00. Shape is a measure of skewness of the CI about the point estimate. A shape > 1.00, implies the CI is positively skewed, with a long tail to the right, whereas shape < 1.00 implies the CI is negatively skewed.
Hypothesis testing with the bootstrap
Bootstrap methods can also be used for hypothesis testing. The two quantities that we must choose when carrying out a bootstrap hypothesis test are a test statistic and a null distribution for the data under the null hypothesis. Given these, we generate bootstrap values of the test statistic under the null distribution for the data and estimate the achieved significance level (ASL) by calculating the proportion of the bootstrap values of the test statistics, which are greater than or equal to the observed value of the test statistic from the original data.
Several bootstrap test statistics are available for comparing the distribution of sample data in two independent groups. In considering a bootstrap hypothesis for comparing the two means, there is no compelling reason to assume equal variances and so we do not make this assumption. We used a bootstrap test statistic for comparing two means that use only the assumption of a common mean, under the null hypothesis [Page 224, [4]].
Linear regression: Model (residual) and case resampling
Standard errors and CIs for regression coefficients can also be obtained using bootstrap methods. Two different approaches are possible, case and model (residual) resampling.
For example with the simple linear model, y = a + bx, where y is the outcome variable and x is a predictor or explanatory variable, a is the intercept and b is the slope or gradient of the line, with n (x, y) pairs of HRQoL observations. Then case-based resampling involves drawing a bootstrap sample of size n, with replacement from these n pairs. Ordinary least squares (OLS) are then used to estimate the regression coefficients for this bootstrap sample of paired cases. Again we do this repeatedly, say 1000 times, so we now have 1000 bootstrap samples and 1000 estimates of the regression coefficients, one from each bootstrap sample. The standard error of these estimated coefficients is simply the standard deviation of these 1000 estimates. As before we can calculate BCa confidence intervals for these estimated regression coefficients.
Case-based resampling may be entirely natural for situations where it is plausible that the (x, y) pairs have been drawn by random sampling from a population. However, case based resampling is less appealing if the x values were controlled for in some way, perhaps by the design of the study. In this situation the alternative model or residual based procedures could be used.
For model based resampling the conventional fitted values and residuals are first obtained from the observed data. A bootstrap sample of the residuals is then drawn. These residuals are then added to the original regression equation (and x values) to generate new bootstrap values for the outcome variable. Ordinary least squares are then used to estimate the new bootstrap regression coefficients, for this bootstrap sample. This process (resampling of the residuals, adding them to the fitted values and estimating the regression coefficients) is repeated lots of times to estimate standard errors and confidence intervals for the regression coefficients from the bootstrap samples.
Thus model based resampling is an example of the "parametric bootstrap" when the residuals from a parametric model are bootstrapped to give estimates of the standard error of the parameters. There is considerable debate about which form of resampling is more appropriate. Both forms of resampling can easily be implemented in STATA [12] and S-PLUS [13]. We now briefly describe the SF-36 outcome and the four example datasets.
SF-36 Health Survey
The SF-36 is one of the most commonly used HRQoL measures in the world today. It contains 36 questions measuring health across eight different dimensions – physical functioning (PF), role limitation because of physical health (RP), social functioning (SF), vitality (VT), bodily pain (BP), mental health (MH), role limitation because of emotional problems (RE) and general health (GH). Responses to each question within a dimension are combined to generate a score from 0 to 100, where 100 indicates "good health" [14]. Thus, the SF-36 generates a profile of HRQoL outcomes, on eight dimensions, with discrete, bounded and skewed distributions (see Figures 1 and 2) which makes statistical analysis and interpretation difficult [1].
Figure 1 Distribution SF-36 dimensions from CPSW data by group
Figure 2 Distribution SF-36 dimensions from CPSW data by group
The four datasets
There now follows a brief description of the four datasets which are used throughout the rest of this paper. These datasets illustrate the use of HRQoL outcomes across a variety of study designs. There are three types of study: observational (both cross-sectional and with baseline and a single follow-up assessment), two group randomised controlled trial (RCT) and longitudinal RCT (with several follow-ups).
CPSW Data: Costs & effectiveness of community postnatal support workers (CPSW): RCT [15]
This RCT aimed to establish the relative cost-effectiveness of postnatal support in the community compared to the usual care provided by community midwives. Six hundred and twenty-three postnatal women were allocated at random to Intervention (n = 311) or Control (n = 312) groups. The intervention consisted of up to 10 home visits in the first postnatal month of up to three hours duration by a community postnatal support worker (CPSW). The main outcomes were HRQoL as measured by the SF-36 at six weeks postnatally. This study is unusual since no baseline HRQoL assessment was made. It was felt that it was inappropriate to assess HRQoL just prior to or immediately after childbirth.
Our analysis is based on the 495 responders to the six-week postnatal questionnaire who completed all 36 items of the SF-36. This sample consisted of 241 women in the Control group and 254 women in the Intervention group. We will use this data to illustrate methods for simple two group cross-sectional comparisons of HRQoL scores using conventional (e.g. t-test and Mann-Whitney tests) and bootstrap hypothesis tests. We will also compare standard Normal theory (t-test) based CIs with their bootstrap BCa equivalent.
OA Knee Data [16]
The aim of this longitudinal observational study was to evaluate two condition specific and two generic health status questionnaires for measuring HRQoL in patients with Osteoarthritis (OA) of the Knee, and offer guidance to clinicians and researchers in choosing between them. Patients were recruited from two settings, knee surgery waiting listings and rheumatology clinics. Four self-completion questionnaires including the SF-36 were sent to the subjects on two occasions 6 months apart. Two hundred and thirty patients returned the questionnaire at initial assessment, consisting of 118 patients awaiting total knee replacement (TKR) Surgery and 112 patients attending Rheumatology outpatient clinics. At the six-month follow-up assessment, 211 patients returned the questionnaire (109 and 102 in the Surgery and Rheumatology groups respectively). The data used here are based on the 211 patients returning both assessments.
Since there was a difference in the baseline HRQoL and sociodemographic characteristics (age and gender) of the Clinic and Surgery groups, we use this dataset to illustrate multiple regression/ANCOVA methods with follow-up HRQoL as the outcome variable and baseline HRQoL, age, gender and group as covariates. We compare the conventional ordinary least squares (OLS) estimates of standard error (SE) and Confidence Interval (CI) for the group regression coefficient with their bootstrap counterparts.
Leg Ulcer RCT data [17]
The aim of this RCT, with one year of follow-up, was to establish the relative cost-effectiveness of community leg ulcer clinics that use four layer compression bandaging versus usual care provided by district nurses. Two hundred and thirty-three patients with venous leg ulcers were allocated at random to intervention (120) or control group (113). The intervention consisted of weekly treatment with four layer bandaging in leg ulcer clinic (Clinic group) or usual care at home by the district nursing service (Home group). The primary outcome was time to complete ulcer healing over the one-year follow-up. Secondary outcomes included HRQoL as measured by the SF-36 at baseline, three months and 12 months follow-up.
We use these data to illustrate the use of summary measures such as the AUC for analysing longitudinal data, using conventional and bootstrap hypothesis tests. We will also compare standard Normal theory (t-test) based CIs with their bootstrap BCa equivalent.
Early Rheumatoid Arthritis RCT data [18]
The Early Rheumatoid Arthritis or NAMEIT trial was a 48-week, randomised, double blind study to compare Neoral with methotrexate (Neoral) versus placebo plus methotrexate (Placebo) in patients with early severe rheumatoid arthritis (RA). The primary efficacy variable in this study was the attainment of American College of Rheumatology (ACR) criteria for improvement of rheumatoid arthritis. Secondary efficacy variables included patient assessment of health related quality of life (HRQoL).
In order to assess the impact of the treatments on patients' health related quality of life, the SF-36 was completed by subjects at seven time-points, Week 0 (baseline), Weeks 8, 16, 24, 32, 40, and Week 48 at the end of the study or at the time of premature withdrawal from the trial.
Three hundred and six subjects at 48 centres were actually entered into the study. One hundred and fifty-two subjects receiving methotrexate were randomised to the Neoral treatment group and 154 subjects receiving methotrexate were randomised to the Placebo group. Of the 306 subjects randomised, 227 completed the study. Seventy-nine randomised subjects discontinued from the study prior to completion.
We use these data to illustrate more complex statistical models for analysing longitudinal data e.g. a marginal GLM fitted with GEEs and compare bootstrap SEs and CIs for the parameters with their conventionally estimated counterparts.
3. Results
Dataset 1 CPSW Study: simple cross-sectional comparison of 6 week HRQoL for the Control vs. Intervention Groups
Figures 1 and 2 show the histograms of the SF-36 dimension scores at six weeks post-natally for Intervention and Control groups. The graphs clearly show the bounded, skewed and discrete nature of the data for the SF-36 from this study.
Table 1 shows the two sample t-test (with equal variances) and Mann-Whitney (MW) comparisons of the eight SF-36 dimension scores. If we assume a cut-off of p ≤ 0.05 for statistical significance, then the t-test suggests significant differences on two dimensions of the SF-36: RP and SF. On two other dimensions PF (p = 0.060) and BP (p = 0.065) the p-values are close to the arbitrary cut-off of 0.05, suggesting some differences although these may not be statistically reliable. The results of the MW tests suggest significant differences on four dimensions (PF, RP, BP and SF) of the SF-36. The only major contrast between the interpretation of the results of the MW and t-tests is on the BP and PF dimensions, where the former test suggests a difference and later not.
Table 1 CPSW Study Simple cross-sectional comparison of 6 week HRQoL for Control vs. Intervention Groups
SF-36 Dimension Group n mean sd Mean Diff t-test equal σ's P-value MW test P-value Bootstrap P-value
Physical Control 241 89.9 14.5 2.6 0.060 0.015 0.057
Function Intervention 254 87.3 15.8
Role Control 241 74.3 38.1 9.1 0.009 0.004 0.010
Physical Intervention 254 65.2 39.5
Bodily Control 241 75.6 23.7 4 0.065 0.040 0.062
Pain Intervention 254 71.6 23.8
General Control 241 77.7 17.7 2.4 0.139 0.147 0.131
Health Intervention 254 75.3 18.5
Vitality Control 241 51.1 20.7 1.3 0.498 0.596 0.494
Intervention 254 49.8 21.7
Social Control 241 81.6 22.7 4.7 0.025 0.015 0.024
Function Intervention 254 76.9 24.2
Role Control 241 77.9 36.4 1.1 0.734 0.503 0.737
Emotional Intervention 254 76.8 35.5
Mental Control 241 72.9 17.2 -0.2 0.902 0.972 0.904
Health Intervention 254 73.1 16.7
The bootstrap p-value is based on 5000 bootstrap replications.
The last column of Table 1 also shows the results of a bootstrap hypothesis test for comparing two means. It compares and contrasts the results of the p-values from a bootstrap hypothesis tests with the p-values from the standard two sample t-test with equal variances, and the MW test. Although they report quantitatively different p-values, the magnitudes are similar, and if we use a cut-off of p < 0.05 for statistical significance then the qualitative interpretation of the tests is the same. So in this example dataset there appears to be little advantage in using the bootstrap hypothesis tests compared to conventional hypothesis tests, such as the t-test, for testing equality of means.
A major limitation of non-parametric methods, such as the MW test, is that they do not allow for the estimation of confidence intervals for parameters or allow for the adjustment of confounding variables such as baseline covariates. One way to estimate non-parametric CIs is via the bootstrap method. Table 2 compares and contrasts the Normal/t-test (equal variances) based confidence intervals with the bootstrap BCa ones.
Table 2 Comparisons of parametric and bootstrap estimates of confidence intervals for the eight dimensions of the SF-36 from the CPSW Study for Control vs. Intervention Groups
SF-36 CIs Interval
Dimension Mean Difference Lower Upper Length Shape
Physical Normal (t-test) -2.6 -5.2 0.1 5.4 1.00
Function Bootstrap BCA -5.2 0.0 5.2 0.98
Role Normal (t-test) -9.1 -16.0 -2.3 13.7 1.00
Physical Bootstrap BCA -15.8 -2.3 13.5 1.02
Bodily Normal (t-test) -4.0 -8.2 0.2 8.4 1.00
Pain Bootstrap BCA -8.1 0.3 8.4 1.03
General Normal (t-test) -2.4 -5.6 0.8 6.4 1.00
Health Bootstrap BCA -5.6 0.8 6.4 0.99
Vitality Normal (t-test) -1.3 -5.0 2.5 7.5 1.00
Bootstrap BCA -5.1 2.4 7.5 0.98
Social Normal (t-test) -4.7 -8.9 -0.6 8.3 1.00
Function Bootstrap BCA -8.7 -0.6 8.1 1.03
Role Normal (t-test) -1.1 -7.5 5.3 12.7 1.00
Emotional Bootstrap BCA -7.1 5.6 12.7 1.11
Mental Normal (t-test) 0.2 -2.8 3.2 6.0 1.00
Health Bootstrap BCA -2.8 3.2 6.0 0.98
Mean difference = Intervention mean - Control mean
BCa confidence intervals based 5000 bootstrap replications.
The estimates and lengths of the CIs are almost identical. Table 2 also shows that the shape of the BCa CIs is almost symmetric about the point estimate of the mean difference except for the RE dimension, where there is some evidence of asymmetry. So again in this example dataset there appears little advantage in using the bootstrap BCa confidence intervals compared to conventional methods of confidence interval estimation.
The bootstrap (and Normal) confidence intervals are calculated for a characteristic of the distributions (for example mean difference). The groups may have differences in distributions but similar characteristics e.g. mean [15]. For example, the MW tests suggests a significant difference (in distributions) for the PF, RP, BP and SF dimensions, but the bootstrap and Normal confidence limits for two out of four of these dimensions (PF and BP) includes zero; suggesting no differences in the mean HRQoL between the groups.
When a hypothesis is tested using the bootstrap, the resampling is carried out assuming the null hypothesis H0 is true. Whereas when confidence intervals for mean differences between two groups are estimated the resampling is carried out separately for each group. A useful analogy is with the comparison of proportions in two independent groups. Here the standard error for the hypothesis test is different to the standard error of the difference between the observed proportions used for estimating a confidence interval [Page 45, [19]].
Dataset 2 OA Knee: comparison of OLS multiple regression, bootstrap case and model based resampling SE and CI estimates for the group (surgery vs. clinic) parameter
Table 3 shows the baseline socio-demographic and HRQoL characteristics of the two groups of OA patients those awaiting total knee replacement surgery (Surgical) and those having pharmacological treatment (Rheumatology). The group of patients awaiting surgery is significantly older and has significantly more men than the Rheumatology group. The Surgical group has significantly lower levels of PF prior to total knee replacement surgery than the Rheumatology group. Conversely the Surgical group has significantly higher levels of GH, V and MH compared to the Rheumatology clinic patients. For the other four dimensions of the SF-36 (RP, BP, SF and RE) there was no evidence of any difference in HRQoL between the two groups.
Table 3 Baseline characteristics of the TKR Surgery and Rheumatology Clinic patients from the OA Knee study.
Rheumatology Surgical 95% CI
N Mean SD N Mean SD Mean Diff Lower Upper P-value
Age (years) 102 64.2 (11.3) 109 71.1 (8.5) -6.9 -9.6 -4.2 0.001
SF-36 Dimensions
Physical Function 97 28.2 (22.4) 95 21.2 (18.2) 7.0 1.2 12.8 0.019
Role Physical 96 11.5 (22.0) 99 12.9 (26.3) -1.4 -8.3 5.4 0.684
Bodily Pain 100 32.0 (19.5) 104 36.3 (23.4) -4.3 -10.3 1.6 0.154
General Health 94 43.9 (22.9) 96 57.3 (23.8) -13.3 -20.0 -6.6 0.001
Vitality 98 36.9 (19.0) 99 42.3 (19.3) -5.4 -10.8 0.0 0.050
Social Function 100 53.1 (30.6) 101 53.6 (27.6) -0.5 -8.6 7.6 0.910
Role Emotional 95 41.1 (44.2) 99 44.1 (44.6) -3.1 -15.6 9.5 0.632
Mental Health 99 62.7 (20.9) 100 68.2 (18.8) -5.5 -11.0 0.1 0.054
Gender
Female 71 (69.6%) 59 (54.1%) (15.5%) (2.4%) (27.8%) 0.021†
Male 31 (30.4%) 50 (45.9%)
Total 102 (100%) 109 (100%)
P-values from two independent samples t-test except (†) χ2test.
We were interested in seeing whether or not there was a difference in HRQoL in OA patients after TKR surgery compared with pharmacologically treated patients. From previous studies using the SF-36 we know that HRQoL varies with age and gender [14,20]. Since there was a difference in the baseline HRQoL and socio-demographic characteristics (age and gender) of the Rheumatology clinic and TKR surgery groups, we use this dataset to illustrate multiple regression/ANCOVA methods with follow-up HRQoL as the outcome variable and baseline HRQoL, age, gender and group (TKR surgery or Rheumatology clinic) as covariates.
The analysis involved using OLS to fit the multiple regression model with six month follow-up HRQoL as the outcome variable and age in years at baseline; gender of the patient (coded 0 for males and 1 for females); baseline HRQoL and treatment group variable (coded 0 = Clinic, 1 = Surgery) as explanatory covariates. The group regression coefficient estimate represents the difference in six-month follow-up HRQoL between the Rheumatology Clinic and TKR Surgery groups after adjustment for the patient's age, gender and baseline HRQoL. A positive value for the regression coefficient indicates the Surgery group has a better mean HRQoL at six months follow-up than the Clinic group after adjustment for the other covariates.
Table 4 compares the OLS and bootstrap standard errors and confidence interval estimates for the group coefficient from the OA Knee data. All models include age, baseline HRQoL and gender as covariates in the regression. For the bootstrap methods the standard errors are the standard deviations of the coefficients from the 5000 bootstrap re-samples. For ease of interpretation and comparison only the estimates for the group coefficient are shown.
Table 4 Comparison of multiple regression, bootstrap case and model based resampling SE and CI estimates from the OA Knee data
Dependent GROUP coefficient 95% CI Interval
Variable Model N SE /SE p Lower Upper Length Shape
Physical Function OLS 165 13.3 3.07 4.31 0.001 7.19 19.32 12.14 1.00
Case 3.02 4.39 7.64 19.69 12.05 1.15
Model 3.05 4.35 7.49 19.49 12.00 1.08
Role Physical OLS 177 -0.5 4.89 -0.11 0.915 -10.16 9.12 19.29 1.00
Case 4.39 -0.12 -8.60 8.51 17.11 1.12
Model 4.86 -0.11 -10.11 8.93 19.04 0.99
Bodily Pain OLS 200 14.7 3.39 4.34 0.000 8.01 21.38 13.37 1.00
Case 3.41 4.30 7.81 21.41 13.60 0.98
Model 3.36 4.38 8.07 21.25 13.18 0.99
General Health OLS 173 4.7 2.01 2.32 0.021 0.71 8.65 7.95 1.00
Case 2.03 7.26 0.69 8.69 8.00 1.01
Model 1.98 2.37 0.70 8.48 7.78 0.95
Energy OLS 185 6.5 2.46 2.64 0.009 1.65 11.36 9.72 1.00
Case 2.50 2.60 1.75 11.63 9.88 1.08
Model 2.46 2.65 1.49 11.04 9.54 0.90
Social Function OLS 194 9.1 3.70 2.46 0.015 1.82 16.41 14.59 1.00
Case 3.52 2.59 2.14 16.06 13.92 1.00
Model 3.65 2.50 1.72 16.09 14.37 0.94
Role Emotional OLS 184 9.4 6.10 1.55 0.124 -2.60 21.48 24.08 1.00
Case 5.89 1.60 -1.90 20.85 22.75 1.01
Model 6.03 1.57 -2.37 20.90 23.27 0.97
Mental Health OLS 191 1.1 2.15 0.51 0.613 -3.15 5.33 8.48 1.00
Case 2.32 0.47 -3.54 5.42 8.96 0.94
Model 2.14 0.51 -3.17 5.12 8.28 0.95
Bootstrap BCa confidence intervals based 5000 bootstrap replications.
The regression analysis suggests that at six month follow-up TKR surgical patients have significantly better HRQoL than Rheumatology treated clinic patients on five dimensions of the SF-36 (PF, BP, GH, V and SF) after adjustment for age, gender and baseline HRQoL. As can be seen from Table 4 the standard error estimates are almost identical for the three methods. Similarly the length of the confidence intervals is virtually the same for all three methods. Although the bootstrap CIs tend to be asymmetric about the point-estimate of the regression coefficient.
Qualitatively all of the intervals from the three methods either include or exclude zero so the interpretation of the group regression coefficient is the same. Therefore, again in this example dataset, there appears to be little advantage in using bootstrap case or model based re-sampling to estimate standard errors and confidence intervals compared to conventional methods of confidence interval estimation from the OLS multiple regression model.
Dataset 3 Leg ulcer: simple cross-sectional comparison of AUC for Home vs. Clinic Groups
We are interested in comparing the HRQoL over the one-year follow-up between the Home and Clinic treated groups. The two groups were well matched at baseline for age, gender and HRQoL, except for the RE dimension of the SF-36, where there was some reliable statistical evidence of a difference (p = 0.052).
The overall HRQoL of the leg ulcer patients over the 12-month study period (and three HRQoL assessments) can be summarised by the AUC. If we set the time units for the AUC calculation as a fraction of a year, then an AUC value of 100 implies the leg ulcer patient has been in "good health" for the entire 12-month follow-up period. Conversely an AUC value of 0 implies the leg ulcer patient has been in "poor health" for the entire 12-month follow-up period.
Table 7 [See additional file 1] gives the results of simple comparisons of differences in mean AUC between the groups using the two independent samples t-test, the MW test and the bootstrap hypothesis test.
The p-values from the t-test and the ASL from the bootstrap hypothesis tests are very similar. None of the p-values for the eight SF-36 dimensions are less than 0.05. Therefore there is no reliable statistical evidence to suggest a difference in mean AUC between the Clinic and Home treated leg-ulcer patients. Only the results of the MW test on the RE dimension of the SF-36 provide (p = 0.071) any evidence of a difference in AUC distributions between the groups, although even this p-value is not statistically significant using the conventional cut-off of 0.05.
The table also contrasts the Normal theory based CI estimates from the t-test with the bootstrap BCa limits. The lengths of the intervals are very similar, although the bootstrap BCa intervals tend to have a non-symmetric shape. All the estimated CIs include zero, again suggesting no evidence of a difference in mean AUC (HRQoL) between the Clinic and Home group patients in the Leg Ulcer study.
Dataset 4 Early RA: Comparison of robust and bootstrap SE's and CI's for the time and group coefficients with a GEE marginal model and exchangeable autocorrelation
In the Early RA study, HRQoL assessment was carried out at 0, 8, 16, 24, 32, 40 and 48 weeks. With seven repeated HRQoL measurements, such as this, the best approach is to model the longitudinal data using GLMs.
The modelling of longitudinal data takes into account the fact that successive HRQoL assessments by a particular subject are likely to be correlated. We used a marginal model with the Early RA data and used GEEs to estimate the regression coefficients. Marginal models are appropriate when inferences about the population average are the focus. For example, in a clinical trial the average difference between control and treatment is most important, not the difference for any one individual. In a marginal model, the regression of the response on explanatory variables is modelled separately from the within-person correlation.
The marginal model is an extension of the linear regression model used with the OA Knee data. Longitudinal models require the specification of the auto- or serial correlation, which is the strength of the association between successive longitudinal measurements of a single HRQoL variable on the same patient.
Several underlying patterns of the auto-correlation matrix are used in the modelling of HRQoL data. The error structure is independent (sometimes termed random) if the off diagonal terms of the auto-correlation matrix are zero. The repeated HRQoL observations on the same subject are then independent of each other, and can be regarded as though they were observations from different individuals. On the other hand, if all the correlations are approximately equal or uniform then the matrix of correlation coefficients is termed exchangeable, or compound symmetric. This means that we can re-order (exchange) the successive observations in any way we choose in our data file without affecting the pattern in the correlation matrix. As the time or lag between successive observations increases, the auto-correlation between the observations decreases. A correlation matrix of this form is said to have an autoregressive structure (sometimes called multiplicative or time series).
Table 5 summarises the resulting 21 auto-correlation pairs for the assessments until week 48. The pattern of the observed auto-correlation matrix, gives a guide to the so-called error structure associated with the successive HRQoL measurements. Table 5 shows that the autocorrelation coefficients range between 0.19 and 0.85. For three dimensions of the SF-36, PF, GH and MH, the autocorrelation coefficients are moderately large (between 0.5 and 0.85). The pattern of values suggests that the assumption of compound symmetry is not unreasonable.
Table 5 Auto-correlation matrices for the eight dimensions of the SF-36 from RA patients in the Early RA study assessed at seven time points
a) Physical Function (n = 218) e) Vitality (n = 216)
Week 0 8 16 24 32 40 48 Week 0 8 16 24 32 40 48
0 1.00 0 1.00
8 0.61 1.00 8 0.55 1.00
16 0.63 0.74 1.00 16 0.48 0.58 1.00
24 0.57 0.69 0.75 1.00 24 0.47 0.54 0.71 1.00
32 0.56 0.68 0.80 0.79 1.00 32 0.50 0.59 0.68 0.71 1.00
40 0.55 0.67 0.77 0.81 0.86 1.00 40 0.42 0.49 0.67 0.68 0.77 1.00
48 0.53 0.64 0.74 0.81 0.81 0.85 1.00 48 0.47 0.53 0.66 0.72 0.72 0.76 1.00
b) Role Physical (n = 212) f) Social Function (n = 219)
Week 0 8 16 24 32 40 48 Week 0 8 16 24 32 40 48
0 1.00 0 1.00
8 0.40 1.00 8 0.44 1.00
16 0.35 0.53 1.00 16 0.43 0.53 1.00
24 0.29 0.39 0.57 1.00 24 0.39 0.55 0.63 1.00
32 0.19 0.30 0.56 0.67 1.00 32 0.36 0.46 0.63 0.70 1.00
40 0.34 0.42 0.52 0.60 0.61 1.00 40 0.38 0.51 0.58 0.64 0.71 1.00
48 0.27 0.40 0.59 0.67 0.64 0.71 1.00 48 0.34 0.45 0.58 0.64 0.71 0.71 1.00
c) Bodily Pain (n = 219) g) Role Emotional (n = 206)
Week 0 8 16 24 32 40 48 Week 0 8 16 24 32 40 48
0 1.00 0 1.00
8 0.43 1.00 8 0.46 1.00
16 0.45 0.55 1.00 16 0.35 0.47 1.00
24 0.44 0.47 0.61 1.00 24 0.34 0.40 0.59 1.00
32 0.37 0.46 0.51 0.68 1.00 32 0.31 0.32 0.56 0.62 1.00
40 0.40 0.42 0.57 0.60 0.69 1.00 40 0.34 0.46 0.53 0.56 0.54 1.00
48 0.42 0.46 0.59 0.63 0.68 0.76 1.00 48 0.31 0.37 0.49 0.58 0.54 0.69 1.00
d) General Health (n = 209) h) Mental Health (n = 218)
Week 0 8 16 24 32 40 48 Week 0 8 16 24 32 40 48
0 1.00 0 1.00
8 0.55 1.00 8 0.57 1.00
16 0.56 0.68 1.00 16 0.57 0.62 1.00
24 0.60 0.67 0.80 1.00 24 0.55 0.59 0.72 1.00
32 0.58 0.67 0.77 0.83 1.00 32 0.52 0.55 0.65 0.69 1.00
40 0.59 0.65 0.72 0.79 0.84 1.00 40 0.50 0.54 0.70 0.70 0.74 1.00
48 0.58 0.65 0.75 0.84 0.82 0.85 1.00 48 0.56 0.55 0.68 0.72 0.73 0.77 1.00
Correlations are Pearson's product moment coefficient.
The process of fitting marginal models using GEE begins by assuming the simple independence form for the autocorrelation matrix, and fitting the model as if each assessment were from a different patient. Once this model is obtained the corresponding residuals are calculated and these are then used to estimate the autocorrelation matrix assuming it is of the exchangeable (or autoregressive) type. This matrix is then used to fit the model again, the residuals are once more calculated, and the autocorrelation matrix obtained. The iteration process is repeated until the corresponding regression coefficients that are obtained in the successive models converge or differ little on successive occasions [1].
Fayers and Machin [Pages 183–202, [1]] and Diggle et al [10] emphasise the importance of graphical presentation of longitudinal data prior to modelling. Figure 3 shows the mean levels of HRQoL in patients with RA, before and during treatment, for the eight dimensions of the SF-36. The curves for some dimensions of the SF-36 overlap (e.g. PF, GH, RE, and MH dimensions) suggesting that it may be unrealistic to assume that the mean difference in HRQoL values on these dimensions remains constant over time. For other dimensions such as BP, V and SF there is some evidence to suggest that for later HRQoL measurements the curves are parallel and that the mean difference between treatments is now fairly constant.
Figure 3 Profile of mean SF-36 scores over time by treatment group EARLY RA data (Patients who completed all seven HRQoL assessments)
The overlapping lines on some of the graphs in Figure 3 imply there may be a 'Treatment × Time' interaction. It is therefore important to test for any such interaction in any regression model. Fortunately, with the marginal model approach this is relatively easy to do and simply involves the addition of an extra regression coefficient to the model. If treatment is coded as a 0/1 variable (i.e. 0 = Placebo and 1 = Neoral) and assessment time as a continuous variable, then the additional interaction term is simply the product of these two variables (which will be 0 for all the Placebo group patients and equal to the HRQoL assessment time in the Neoral Group patients).
Early RA marginal model analysis
The marginal model we used for the Early RA data for analysing the seven HRQoL assessments over time was,
Yij = β1 + βBasexBase_i + βAgexAge_i + βSexxSex_i + βTimetij + βGroupxGroup_i + εij, (2)
where Yij is the HRQoL at time tij post-baseline; tij is the time of the QoL assessment, in weeks post baseline, of patient i at visit j; xBase_i is the baseline HRQoL assessment for subject i; xAge_i is the age (in years) of subject i at time 0 (baseline); xSex_i is the gender of subject i; xGroup_i is the treatment group (0 = Placebo, 1 = Neoral) for subject i; β1 is a constant and εij is the residual error.
The marginal regression models were fitted in STATA [12] using the xtgee command with an identity link function (link (iden)) and the robust standard errors option. The observed correlation matrices in Table 5 clearly show the off-diagonal terms are non-zero and that the assumption of an independent auto-correlation matrix for the marginal model is unrealistic. We will not consider models with an independent auto-correlation structure and will concentrate on reporting the results of models with an exchangeable correlation.
None of the interaction term coefficients for the eight SF-36 dimensions were statistically significant (from zero). Thus there was no reliable evidence of a 'Treatment × Time' interaction on any dimension of the SF-36 (p > 0.05), irrespective of the autocorrelation structure. Therefore we will only report the results of the simpler model (2), without the interaction term.
The beauty of the marginal model and the GEE methodology is that it is very flexible and can in principle deal with all the observed data from a HRQoL study. The subjects are not required to have exactly the same numbers of assessments, and even the assessments can be made at variable times. The latter allows the modelling to proceed even if a subject misses a HRQoL assessment. So it seems unrealistic and unreasonable to use bootstrap resampling methods for marginal models that can only utilise a balanced data set, with equally spaced QoL assessments. Since we are interested in fitting a marginal model and we are likely to have an unbalanced dataset with unequal observations per subject we used simple bootstrap case-resampling.
Figure 4 shows the estimated within subject correlation matrices for the eight dimensions of the SF-36 if we fit the longitudinal model and assume a compound symmetric structure. The lower diagonal gives the observed matrix before the model fitting. The fitted autocorrelations ranged from 0.43 for the RE dimension to 0.63 for the PF and GH dimensions. On the whole, the model correlation estimates tend to be lower than the actual observed autocorrelations, for HRQoL assessments that are close together. Conversely the model correlation estimates tend to be larger than the observed correlations for HRQoL observations further apart in time. It will usually be the case that after model fitting the autocorrelations will appear to have been reduced [1]. The observed deviations between the fitted model and observed autocorrelations are not too great, suggesting that the assumption of compound symmetry is not unreasonable (Figure 4).
Figure 4 Observed and estimated within-patient auto-correlation matrices (exchangeable model) from RA patients in the EARLY RA study. The lower diagonal gives the observed matrix before model fitting whilst the upper gives the exchangeable form after model-fittinga
Table 6 shows the estimated regression coefficients for the group and time variables. There is some evidence that HRQoL increases over time for three dimensions of the SF-36, PF, BP and V. However, we are interested in the effect of treatment and comparing HRQoL over time across the Placebo and Neoral treated groups. Since there is no reliable evidence of a 'Group × Time' interaction the interpretation of the treatment group coefficient is relatively straightforward. The p-values for the treatment group regression coefficients in Table 6 suggest significant differences in HRQoL between the Neoral and Placebo groups on three dimensions of the SF-36 (RP, GH and BP).
Table 6 Comparison of robust and bootstrap SE's and CI's from the EARLY RA data with a Marginal Model and exchangeable autocorrelation
Dependent Coefficients 95% CI Interval
Variable SE /SE p Lower Upper Length Shape
Physical Function (n = 222) time 0.11 0.03 3.63 0.001 0.05 0.18 0.12 1.00
0.03 3.72 0.05 0.18 0.12 0.97
group 2.82 2.25 1.25 0.211 -1.60 7.24 8.84 1.00
1.72 1.64 -0.51 6.13 6.64 0.99
Role Physical (n = 221) time -0.06 0.07 -0.90 0.366 -0.19 0.07 0.26 1.00
0.07 -0.94 -0.19 0.06 0.25 0.91
group 9.49 3.93 2.42 0.016 1.79 17.19 15.40 1.00
3.22 2.95 3.63 16.62 12.99 1.22
Bodily Pain (n = 222) time 0.16 0.03 4.69 0.001 0.10 0.23 0.14 1.00
0.03 4.88 0.10 0.23 0.13 1.05
group 4.23 1.97 2.14 0.032 0.36 8.10 7.74 1.00
1.50 2.82 1.44 7.25 5.81 1.08
General Health (n = 221) time 0.04 0.03 1.67 0.095 -0.01 0.09 0.10 1.00
0.03 1.68 -0.01 0.09 0.10 0.95
group -4.61 1.96 -2.35 0.019 -8.46 -0.76 7.69 1.00
1.51 -3.04 -7.36 -1.28 6.07 1.21
Vitality (n = 220) time 0.09 0.03 3.09 0.002 0.03 0.14 0.11 1.00
0.03 3.05 0.04 0.15 0.11 1.24
group 2.67 1.80 1.48 0.14 -0.87 6.20 7.07 1.00
1.41 1.89 -0.19 5.42 5.61 0.96
Social Function (n = 222) time 0.03 0.03 0.77 0.442 -0.04 0.09 0.13 1.00
0.03 0.79 -0.04 0.09 0.13 0.97
group 2.40 2.11 1.14 0.255 -1.73 6.54 8.27 1.00
1.65 1.46 -0.72 5.88 6.61 1.11
Role Emotional (n = 221) time -0.02 0.07 -0.32 0.752 -0.17 0.12 0.29 1.00
0.08 -0.31 -0.18 0.12 0.30 0.93
group 4.14 3.91 1.06 0.29 -3.52 11.81 15.33 1.00
2.93 1.41 -1.54 10.11 11.64 1.05
Mental Health (n = 221) time 0.02 0.03 0.69 0.489 -0.03 0.07 0.10 1.00
0.03 0.68 -0.03 0.07 0.10 1.15
group 1.53 1.67 0.92 0.359 -1.74 4.79 6.53 1.00
1.34 1.14 -0.93 4.42 5.35 1.18
Note: The bootstrap estimates of SE and BCa Confidence Intervals are shown in italics below the model based estimates and are based on 1000 resamples.
The bootstrap and robust standard errors for the time and group coefficients are different, although the bootstrap SE estimate tends to be the same size or somewhat smaller than its robust counterpart. However both bootstrap and robust SE estimates are of a similar order of magnitude. More importantly, the ratios of the estimated coefficient to its standard error are of similar size.
A crude test of statistical significance is to examine this ratio, if it is bigger than 2.0 then the estimated regression coefficient is likely to be significantly different from zero. Table 6 shows that for all the models where the original (group or time) regression estimates are significant (i.e. ratios of estimate/SE > 2) then so too is the ratio of the estimate to its bootstrap standard error.
When we compare the bootstrap BCa confidence intervals with the model- based estimates in Table 6 then the length of the bootstrap intervals tend to be the same size or slightly narrower than its robust counterpart. As before the bootstrap estimates are not constrained to be symmetric about the point-estimate of the regression coefficient. Qualitatively both the bootstrap and model based intervals include zero when the estimated regression coefficient is non-significant and exclude zero when the estimated coefficient is significant. Therefore, the actual practical interpretation of the confidence interval estimates is the same. That is for the RP, BP, and GH dimensions there is some evidence that the Neoral group has a better HRQoL than the Placebo group patients over time, after allowing for baseline HRQoL, age and gender.
The use of the bootstrap to estimate SEs and CIs for marginal longitudinal models appears to offer little advantage (in the Early RA data) compared to the conventional robust estimates.
4. Discussion
In the datasets and outcomes studied, and for the specific conventional analyses we used, we have shown that use of the bootstrap does not lead to different p-values, SE and CI estimates compared to conventional methods. On this basis, we cannot conclude the use of the bootstrap is more appropriate than conventional methods. The explanation for this conclusion and the extent of its generalisability deserve discussion.
Ordinality of HRQoL outcomes
One of the fundamental assumptions we have made, is that there exists an underlying continuous latent variable that measures HRQoL, and that the actual measured outcomes are ordered categories that reflect contiguous intervals along this continuum. If the goal of the analysis is to assess the magnitude of the treatment effect on this ordered outcome, then an appealing approach is to assign numeric scores to the ordered categories and then to compare means between groups using conventional linear regression methods. If interest lies elsewhere, for example in comparing the relative frequencies of cumulative probabilities in the ordered categories between treatments, then other techniques such as the proportional odds model would be more appropriate [9,2,21]. Heeren and D'Agostino [26] have demonstrated the robustness of the two independent samples t-test when applied to three-, four- and five point ordinal scaled data using assigned scores, in sample sizes as small as 20 subjects per group. Sullivan and D'Agostino [27] have expanded this work to account for a covariate when the outcome is ordinal in nature. They again assign numeric scores to the distinct response categories and compare means between treatment groups adjusting for a covariate reflecting a baseline assessment measured on the same scale. Their simulation study shows that in the presence of three-, four- and five point ordinal data and small sample sizes (as low as 20 per group) that both ANCOVA and the two independent sample t-test on difference scores are robust and produce actual significance levels close to the nominal significance levels.
Generalisability
The generalisability of the results could be called into question as they only apply to the limited number of datasets studied (four) and the SF-36 outcome. The SF-36 outcome is the most widely used generic HRQoL measure in the world today, so that is one obvious reason to use it [22]. Secondly, we had easy access to a variety of datasets that had previously used the SF-36 outcome. The four studies (CPSW, OA Knee, Leg Ulcer and Early RA), and datasets were well known to us. They illustrate the use of HRQoL outcomes across a variety of studies including cross-sectional surveys, RCTs, non-randomised before and after studies and longitudinal designs. So on practical and pragmatic grounds, we felt it was appropriate to use such datasets because of their familiar nature and the analysis was easy to understand and interpret.
The SF-36 is a multi-dimensional outcome with eight dimensions. As described in the Introduction the eight dimensions have a variety of distributions. We believe these distributions are not atypical of other generic HRQoL measures such as the NHP and EORTC QLQ-C30. The distributions we considered were chosen based on our experiences with HRQoL data in a variety of settings. So we believe that our results about the bootstrap may have generalisability to other HRQoL outcomes (besides the SF-36) used in other studies and populations, although strictly speaking our results only apply to the SF-36 outcome and the observed datasets. Hence, we cannot make sweeping generalisations about the impact of the bootstrap on other HRQoL outcomes, used in other studies. Therefore, these results need to be replicated with other HRQoL measures in other datasets and populations.
Missing values
It should be noted that in the all four example datasets there is missing data. We assumed that any missing HRQoL values in these datasets were Missing Completely at Random (MCAR). This means that the probability of the HRQoL response being missing is independent of the scores on the previous observed questionnaires and independent of the current and future scores had they been observed. We have assumed that the reduced dataset represents a randomly drawn sub-sample of the full dataset and the inferences drawn can be considered reasonable. This is a strong assumption and unlikely to hold for missing HRQoL data [1,7,23-25].
Sample sizes of the example datasets
The various datasets used in this study all had a sample size in excess of 100 patients. Some caution should be used in applying the results to smaller sample sizes. However the robustness of the conventional two-sample t-test and ANCOVA, for three-, four- and five point ordinal scale data using assigned scores has been demonstrated for sample sizes as small as 20 [26,27]. Simple bootstrapping may not be very successful in small samples anyway (say < 9 observations), since the observations themselves are less likely to be representative of the study population. As Campbell [Page 118, [11]] states, "In very small samples even a badly fitting parametric analysis may outperform a non-parametric analysis, by providing less variable results at the expense of a tolerable amount of bias."
The bootstrap
Bootstrap case resampling vs. model based resampling
The results with the OA Knee data show that there is little to choose from between the case and model based resampling for the multiple linear regression model for estimating SEs and CIs. Since there was very little difference in the SE and CI estimates from the datasets used, for simplicity one would tend to favour a case based resampling approach. Indeed this was the resampling method for the longitudinal marginal model for the Early RA data.
Bootstrap model based resampling for marginal model
In the longitudinal Early RA for simplicity we used only a simple case based resampling for the marginal model and effectively carried out a stratified random resampling with replacement. That is we sampled with replacement blocks or clusters of each patients' repeated HRQoL responses. In theory, one should be able to use model or residual based resampling for the marginal model. The resampling procedure would be rather complex particularly for autoregressive autocorrelation structures and for unbalanced datasets, with HRQoL assessments at unequally spaced time points. One would have to take into account that the residuals were not independent and uncorrelated, and for the autoregressive correlation structure, that the correlation between residuals within a patient declined over time. This is a very interesting avenue and requires further exploration with other longitudinal datasets.
Are the results surprising or unexpected?
Finally, are the results all that surprising or unexpected? We have shown that the use of bootstrap methods for analysis (calculation of p-values, SE and CIs) appears to offer little advantage compared to standard methods in the four datasets studied.
If we assume that there exists an underlying continuous latent variable that quantifies the HRQoL response of interest and that the goal of the analysis is to assess the magnitude of a treatment effect on the HRQoL outcome, by comparing means between groups. Then statistical theory says that if the distribution of the HRQoL data is Normal, so will be the distribution of the sample mean. Much more importantly, even if the distribution of HRQoL data is not Normal, as is frequently the case, that of the sample mean will become closer to a Normal distribution as the sample size gets larger. This is a consequence of the Central Limit Theorem (CLT) [Pages 304-7, [28]]. The Normal distribution is strictly only the limiting form of the sampling distribution as the sample size increases to infinity, but it provides a remarkable good approximation to the sampling distribution even when the sample size is small and the distribution of the data is far from Normal [Page 94, [29]]. This implies, for example, that the distribution of the sample means for the SF-36 HRQoL data shown in Figures 1 and 2 will be approximately Normal.
Thus, if the investigator is planning a large study and the sample mean is an appropriate summary measure of the HRQoL outcome, then pragmatically there is no need to worry about the distribution of the HRQoL outcome and we can use standard methods to estimate sample sizes and analyse the data. Since dramatic effects are unlikely in HRQoL studies using the SF-36 as an outcome, large samples sizes are likely to be required [2,3,6]. So perhaps unsurprisingly, the results reflect the robustness of conventional methods with large sample sizes and the application of the CLT to sample means even for HRQoL data with such bounded, discrete and skewed distributions as shown in Figures 1 and 2.
So our research using the SF-36 HRQoL outcome and the four datasets has shown that bootstrap methods appear to produce p-values, SEs and CIs similar to conventional methods. When the standard and the bootstrap methods agree, we can be more confident about the inference we are making and this is an important use of the bootstrap [Page 118, [11]]. When they disagree more caution is needed, but the relatively simple assumptions required by the bootstrap method for validity mean that in general it is to be preferred. Thus, there appears to be little advantage in using the bootstrap for the analysis of SF-36 data, particularly if one is interested in comparing mean HRQoL between treatment groups.
5. Conclusions
In the datasets we studied, using the SF-36 as an outcome measure, bootstrap methods produce results similar to conventional statistical methods. This is likely because the t-test and OLS multiple regression are robust to the violations of assumptions that HRQoL data are likely to cause (i.e. non-Normality). While particular to our datasets, these findings are likely to generalise to other HRQoL outcomes, which have discrete, bounded and skewed distributions. They may not generalise to HRQoL studies with smaller sample sizes of less than 100 subjects. Future research with other HRQoL outcome measures, interventions and populations, is required to confirm this conclusion.
Supplementary Material
Additional File 1
Table 7 – Leg Ulcer study simple cross-sectional comparison of AUC for Home vs. Clinic Groups
Click here for file
==== Refs
Fayers PM Machin D Quality of Life Assessment, Analysis and Interpretation 2000 Chichester: Wiley
Walters SJ Campbell MJ Lall R Design and Analysis of Trials with Quality of Life as an Outcome: a practical guide Journal of Biopharmaceutical Statistics 2001 11 155 176 11725929 10.1081/BIP-100107655
Walters SJ Campbell MJ Paisley S Methods for determining sample sizes for studies involving health-related quality of life measures: a tutorial Health Services & Outcomes Research Methodology 2001 2 83 99 10.1023/A:1020102612073
Efron B Tibshirani RJ An Introduction to the Bootstrap 1993 New York: Chapman & Hall
Davison AC Hinkley DV Bootstrap Methods and their Applications 1997 Cambridge: Cambridge University Press
Walters SJ Sample size and power estimation for studies with health related quality of life outcomes: A Comparison of Four Methods Using the SF-36 Health Qual Life Outcomes 2004 2 26 15161494 10.1186/1477-7525-2-26
Fairclough DL Design and Analysis of Quality of Life Studies in Clinical Trials 2002 New York: Chapman & Hall
Matthews JNS Altman DG Campbell MJ Royston P Analysis of serial measurements in medical research British Medical Journal 1990 300 230 235 2106931
McCullagh P Nelder JA Generalized Linear Models 1989 2 London: Chapman & Hall/CRC
Diggle PJ Heagerty P Liang K-Y Zeger SL Analysis of Longitudinal Data 2002 2 Oxford: Oxford University Press
Campbell MJ Statistics at Square Two: Understanding Modern Statistical Applications in Medicine 2001 London: British Medical Journal
StataCorp Stata Statistical Software: Release 80 College Station, Texas: Stata Corporation 2003
MathSoft S-PLUS 2000 Guide to Statistics, Seattle 1999 2 Washington: Data Analysis Products Division, MathSoft
Brazier JE Harper R Jones NMB O'Cathain A Thomas KJ Usherwood T Westlake L Validating the SF-36 health survey questionnaire: new outcome measure for primary care British Medical Journal 1992 305 160 164 1285753
Morrell CJ Spiby H Stewart P Walters S Morgan A Costs and effectiveness of community postnatal support workers: randomised controlled trial British Medical Journal 2000 321 593 598 10977833
Brazier JE Harper R Munro JF Walters SJ Snaith ML Generic and condition-specific outcome measures for people with osteoarthritis of the knee Rhemautology 1999 38 870 877 10.1093/rheumatology/38.9.870
Morrell CJ Walters SJ Dixon S Collins KA Brereton LML Peters J Brooker CGD Cost-effectiveness of community leg ulcer clinics: randomised controlled trial British Medical Journal 1998 316 1487 1491 9582132
Allard S and the NAME IT Study Group Phase IIIB Clinical Study Report of a Double-Blind, Randomised, Controlled Study to Compare Methotrexate plus Neoral® versus Methotrexate plus Placebo in Subjects with Early Severe Rheumatoid Arthritis 2000 Basel, Switzerland: Novartis Pharma AG
Altman DG Machin D Bryant TN Gardner MJ Statistics with Confidence: Confidence intervals and statistical guidelines 2000 2 London: British Medical Journal
Walters SJ Munro JF Brazier JE Using the SF-36 with older adults: cross-sectional community based survey Age & Ageing 2001 30 337 343 11509313 10.1093/ageing/30.4.337
Lall R Campbell MJ Walters SJ Morgan K and MRC CFAS A review of ordinal regression models applied on Health related Quality of Life Assessments Statistical Methods in Medical Research 2002 11 49 67 11923993 10.1191/0962280202sm271ra
Kaplan RM Staquet MJ, Hays RD, Fayers PM Profile versus utility based measures of outcome for clinical trials Quality of Life Assessment in Clinical Trials: Methods and Practice 1998 Oxford: Oxford University Press 69 90
Curran D Molenberghs G Fayers PM Machin D Incomplete quality of life data in randomised trials: missing forms Statistics in Medicine 1998 17 697 709 9549817 10.1002/(SICI)1097-0258(19980315/15)17:5/7<697::AID-SIM815>3.3.CO;2-P
Fayers PM Curran D Machin D Incomplete quality of life data in randomised trials: missing items Statistics in Medicine 1998 17 679 696 9549816 10.1002/(SICI)1097-0258(19980315/15)17:5/7<679::AID-SIM814>3.3.CO;2-O
Troxel AB Fairclough DL Curran D Hahn EA Statistical analysis of quality of life with missing data in cancer clinical trials Statistics in Medicine 1998 17 653 666 9549814 10.1002/(SICI)1097-0258(19980315/15)17:5/7<653::AID-SIM812>3.3.CO;2-D
Heeren T D'Agostino R Robustness of the two independent samples t-test when applied to ordinal scaled data Statistics in Medicine 1987 6 79 90 3576020
Sullivan lM D'Agostino RB Robustness and power of analysis of covariance applied to ordinal scaled data as arising in randomized controlled trials Statistics in Medicine 2003 22 1317 1334 12687657 10.1002/sim.1433
Hogg RV Tanis EA Probability and Statistical Inference 1998 3 New York: McMillan
Armitage P Berry G Matthews JNS Statistical Methods in Medical Research 2002 4 Oxford: Blackwell Science
| 15588308 | PMC543443 | CC BY | 2021-01-04 16:38:11 | no | Health Qual Life Outcomes. 2004 Dec 9; 2:70 | utf-8 | Health Qual Life Outcomes | 2,004 | 10.1186/1477-7525-2-70 | oa_comm |
==== Front
Ann Clin Microbiol AntimicrobAnnals of Clinical Microbiology and Antimicrobials1476-0711BioMed Central London 1476-0711-3-251554648510.1186/1476-0711-3-25ResearchLaboratory diagnosis and susceptibility profile of Helicobacter pylori infection in the Philippines Destura Raul V [email protected] Eternity D [email protected] Leah J [email protected] Cirle S [email protected] Venancio I [email protected] Ma Lourdes O [email protected] Richard L [email protected] Center for Global Health, Division of Infectious Disease and International Health, University of Virginia, Charlottesville, Virginia, USA2 Section of Infectious Diseases, Philippine General Hospital, Manila, Philippines3 Section of Gastroenterology, Department of Medicine, Philippine General Hospital, Manila, Philippines4 National Institute of Health-University of the Philippines, Manila, Philippines2004 16 11 2004 3 25 25 7 9 2004 16 11 2004 Copyright © 2004 Destura et al; licensee BioMed Central Ltd.2004Destura 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
Helicobacter pylori diagnosis and susceptibility profile directs the applicability of recommended treatment regimens in our setting. To our knowledge, there is no published data on the culture and local susceptibility pattern of Helicobacter pylori in the Philippines.
Methods
52 dyspeptic adult patients undergoing endoscopy from the Outpatient Gastroenterology clinic of the University of the Philippines-Philippine General Hospital underwent multiple gastric biopsy and specimens were submitted for gram stain, culture, antimicrobial sensitivity testing, rapid urease test and histology. Antimicrobial susceptibility testing was done by Epsilometer testing (Etest) method against metronidazole, clarithromycin, amoxicillin, and tetracycline.
Results
Sixty percent (60%) of the study population was positive for H. pylori infection (mean age of 44 years ± 13), 70% were males. H. pylori culture showed a sensitivity of 45% (95% CI [29.5–62.1]), specificity of 98% (95%CI [81.5–100%]), positive likelihood ratio of 19.93 (95% CI [1.254–317.04]) and a negative likelihood ratio of 0.56 (95% CI [0.406–0.772]). All H. pylori strains isolated were sensitive to metronidazole, clarithromycin, amoxicillin and tetracycline.
Conclusion
Knowledge of the antibiotic susceptibility patterns in our setting allows us to be more cautious in the choice of first-line agents. Information on antibiotic susceptibility profile plays an important role in empiric antibiotic treatment and management of refractive cases.
==== Body
Background
Helicobacter pylori is a gram-negative bacterium that colonizes the gastric mucosa of more than half of the world's population [1,2] Since its isolation in 1982, the association between H. pylori infection and the subsequent development of chronic active gastritis, peptic ulcer disease, gastric cell carcinoma and B cell MALT lymphoma has been well established[3]. The principal reservoir of infection is the human stomach and transmission has been epidemiologically linked to person to person contact [4]. The prevalence of infection is greater in developing countries and is influenced by socioeconomic conditions, ethnic background and age[5,6] In the Philippines, there is scarcity of published data regarding the epidemiology of this bacterium. Locally unpublished reports revealed a prevalence of 5.6% seropositivity rate in children and 60% among 136 adult Filipino patients with dyspepsia using the Clotest® (Cabahug et. al. 2003 and Caballero et al., 1997, unpublished data). A lower prevalence rate of 42% was reported by Daez et. al. in 2002 (unpublished data) among 375 patients undergoing endoscopy at the Philippine General Hospital utilizing the rapid urease test and histopathology.
Microbiological isolation of the organism is the theoretical gold standard for the detection of H. pylori infections. However, isolation of the organism by culture has been highly variable. Success rates depend on the technical expertise of the microbiology laboratory, ranging from 30% to 73%[7,8]. Failure to detect the organism may be due to sampling error, inappropriate transport or culture media and insufficient incubation period.
In clinical practice, gastric biopsy with culture is not routinely performed due to the availability of more rapid diagnostic tests in the detection of H. pylori such as urease broth tests, urea breath tests, serologic methods and stool antigen detection. However, the increasing prevalence of resistant strains makes culture and antibiotic sensitivity testing valuable to determine alternative treatment regimens after failure of initial eradication regimen.
In the Philippines, due to methodological difficulties in isolating the organism, detection of the organism by culture methods has not been popular. Realizing the increasing prevalence of antimicrobial resistance in other countries and its potential negative impact on the efficacy of many treatment eradication regimens, it is important in clinical practice to determine the prevailing local antibiotic susceptibility patterns when choosing appropriate eradication regimens for H. pylori infections in the empiric setting.
This study aims to evaluate the use of culture in the diagnosis of H. pylori infection among patients with dyspepsia, to determine the sensitivity and specificity of culture technique in the detection of H. pylori infection, and to determine the antibiotic susceptibility patterns of H. pylori organisms isolated by culture among Filipino patients.
Methods
This is a prospective, cross-sectional study involving adult patients with dyspepsia, who had independently been determined to have clinical indications for an endoscopy at the out-patient gastrointestinal clinic of the Philippine General Hospital, a tertiary training university hospital in Manila. Eligible patients were enrolled in the study after informed consent to undergo the required diagnostic testing of endoscopy samples.
Patients were excluded if they were less than 18 years old, had a history of proton pump inhibitors (PPI) intake within 2 weeks, H2 antagonists within 1 week and antibiotic intake within 1 month prior to inclusion in the study.
The nature and purpose of the study were discussed with the patient until fully understood. All patients with dyspepsia undergoing endoscopy who fulfilled the inclusion criteria had a complete history and physical examination. Data were obtained using a data collection form. Participants underwent upper gut endoscopy as clinically indicated. Pre-procedure preparations for Esophagogastroduodenoscopy were performed according to standard methods. Biopsy of gastric tissue were collected from the antrum and body of the stomach and specimens were sent for (1) histopathologic study, (2) gram staining, (3) culture and sensitivity and (4) rapid urease broth test. Those interpreting results of the above diagnostic tests were blinded.
Case Definition
A patient with Helicobacter pylori infection was defined as those patients independently assessed by their attending physician based on clinical symptoms and a positive test for any of the two diagnostic tests (histology and rapid urease test). In the evaluation of the diagnostic performance of H. pylori culture, the above clinico-laboratory case definition were used as the comparator reference standard.
Description of the Diagnostic Tests
Gram stain and Culture
Two pieces of gastric tissue were obtained and placed in 0.2 mL sterile saline and transported to the microbiology laboratory for processing. The biopsy specimen was placed in a sterile petri dish and minced with 2 sterile scalpel blades. Specimens were inoculated in both 7% Horse Blood + Brain Heart Infusion Agar (HBAP) and Brucella Blood Agar with Skirrow's supplement (5% defibrinated Sheep's Blood with Trimethoprim (5 mg/L), Vancomycin 10 mg/L and Polymixin B (2500 units/L))[9] Plates were incubated at 37°C for 7–10 days in a microaerophilic incubation environment and examined every other day (Pack-Microaero, Mitsubishi gas co., Japan). H. pylori colonies are typically small, flat and translucent to grey. On days 4 to 5, all plates with no characteristic colonies, were subcultured to a fresh HBAP to promote growth of slow or fastidious strains and incubated for an additional 3 to 5 days. Plates were examined until the tenth day before reporting a negative growth. Suspected H. pylori colonies were tested for urease, oxidase and catalase production. A modified gram stain was performed on a methanol-fixed smear using crystal violet for 1 minute followed by a water wash and then a safranin counter stain for 30 minutes prior to a final washing with tap water The smear was air-dried and examined under oil immersion.
The presence of H. pylori is confirmed by the presence of a gram negative curved bacilli and a positive test for urease, oxidase and catalase production. Five H. pylori isolates that survived the shipping and handling were verified at University of Virginia Center for Studies of Diseases Due to H. pylori.
Sensitivity Study
Epsilometer test (Etest, AB Biodisk, USA) was used to determine the minimal inhibitory concentrations (MIC). MIC values were read as the intercept of the elliptical zone of inhibition with the graded strip for the Etest. Strains were considered resistant when the MIC was >8 g/ml for metronidazole, >1 g/ml for clarithromycin and >0.5 g/ml for amoxicillin. These breakpoints were used based on the recommendations from the National Committee for Clinical Laboratory Standards (NCCLS) and a large clinical trial [10,11]. For tetracycline, resistance was determined at an MIC of >2 μg/ml based on a previous publications[12,13]. Sensitivity results were compared with a standard susceptible strain of H. pylori (NCTC # 12822) and the University of Virginia Center for Studies of Diseases Due to H. pylori metronidazole resistant culture strain #Cp 2124 and clarithromycin resistant strain #Cp 5535. In the absence of a resistant control for amoxicillin and tetracycline, susceptibility breakpoints set by NCCLS and large clinical trials were used[10,11,13].
Rapid Urease Test
The Rapid Urease Test (RUT) was performed by placing 0.5 ml of 8% (weight/vol) unbuffered urea in distilled water (pH 6.8) in a clear 0.7 ml Eppendorf tube, to which one drop of 1% phenol red (free acid) suspension was added. The urea solutions was stored at 4°C and prepared on the day of use to ensure color stability. Two gastric biopsy specimens from the antrum and body were placed in the tube. A positive test was indicated by a rapid color change of the media surrounding the biopsy from yellow to magenta followed by a rapid generalized color change throughout the media. A negative result was indicated when there was no change in color appreciated after 2 hours of observation.
Histopathologic Examination
Specimens were sent to the pathology lab and gastric tissues were fixed and stained with Giemsa and Hematoxylin-Eosin dye. A specimen was read as positive if curve bacilli organisms were seen on microscopy. Pathologists were blinded to the results of the other diagnostic tests.
Statistical Analysis
Demographic data was described using rates and percentages for categorical variables. For continuous variables, means and standard deviations were used. Measures of accuracy for H. pylori culture were expressed as sensitivity and specificity rates, positive and negative predictive values and likelihood ratios with a 95% confidence interval.
Results
Patient Characteristics
Among 52 patients with dyspepsia, 31 (60%) were positive for H. pylori infection based on the pre-defined case definition. The mean age for H. pylori infected individuals was 44 years ± 13. Seventy percent were males with a male:female ratio of 2:1. The majority of infected patient were married (80%), and had reached only up to the secondary level of education (70%). Fifty-five percent were unemployed. Seventy-five percent of infected patients had access to piped water. No significant differences among the demographic characteristics of H. pylori positive and negative cases were observed.
Laboratory Diagnosis of H. pylori
All included patients underwent the 3 diagnostic tests for the diagnosis of H. pylori infection (histopathology, rapid urease test (RUT) and culture) (figure 1). Fourteen of the 52 patients grew H. pylori on culture. Ten of the 14 positive culture samples were also positive for both histology and RUT, 14 and 10 were positive for RUT and histology alone, respectively. To validate the accuracy rate of H. pylori culture, results of the culture studies were compared with clinically defined cases of H. pylori infection, in this case patients who presents with abdominal symptoms and positive for at least one of the two diagnostic tests, histopathology and RUT. H. pylori culture showed a sensitivity of 45% (95% CI [29.5–62.1%]), specificity of 98% (95%CI [81.5–99.8%]), positive likelihood ratio of 19.93 (95% CI [1.254–317.04]) and a negative likelihood ratio of 0.56 (95% CI [0.406–0.772]). The positive predictive value was 97% (95% CI [74.7–99.7%])) and the negative predictive value was 55% (95% CI [39.8–69.7%]).
A total of 14 H. pylori organisms were isolated from 52 clinical specimens. The mean number of incubation time was 3.8 days ± 1 day. All isolates grew on primary plates. All isolates were highly sensitive to amoxicillin (mean MIC of 0.016 ug/ml by Etest)), tetracycline (mean MIC of 0.164 ug/ml SD ± 0.16 SD by Etest), metronidazole (mean MIC of 0.061 ug/ml SD ± 0.04 by Etest) and clarithromycin (mean MIC of 0.016 SD ± 0 by Etest) (table 1).
Discussion
This pilot study reported an H. pylori culture sensitivity rate of 45% and a specificity rate of 98–100% which are comparable to those reported in other countries[8,14-17]. High positive predictive values coupled with an intermediate to high likelihood ratio demonstrates that gastric tissue culture is highly specific, making it a useful confirmatory test in the diagnosis of H. pylori infection. Its low sensitivity is acceptable since this method is not recommended as a screening test. This study also supported previous studies that the rapid urease test whether in a gel or liquid preparation is a highly sensitive tool which qualifies as a good screening test among suspected H. pylori infected individuals[13,18,19].
Antibiotic resistance has increasingly been recognized as a major cause of treatment failure for H. pylori infection. Primary antimicrobial resistance against clarithromycin and metronidazole is now commonplace in several countries[2,20-26]. Regional variations in susceptibility and resistance patterns may be ascribed to differences in local antibiotic prescription practices, antibiotic usage in the community and mass eradication programs for H. pylori infection as part of gastric cancer prevention strategies. These factors may well be expected to influence success of eradication therapy [27-29].
All 14 strains isolated showed sensitivity to all the first line antibiotics namely metronidazole, amoxicillin, clarithromycin and tetracycline. No resistant strains were isolated based on the Etest method. Susceptibility patterns in Europe and the United States revealed that the highest resistance is to metronidazole ranging from 33.1% to 36.9%. Clarithromycin resistance was observed to be 10% in both areas. In contrast, Japanese data showed that clarithromycin resistance was 29% closely followed by metronidazole at 24%. Amoxicillin resistance remained low at 0–1.4% in all three geographic locations[3,21,23]. While susceptibility studies were done on large numbers of isolates in foreign data, this local study comprises one of the pioneering attempts, to determine the antibiotic susceptibility pattern of H. pylori infection in the Philippines.
Potential reasons for the absence of resistant H. pylori strains in our pilot study point to the type of population enrolled. Based on the selection criteria, these patients had no exposure to previous antibiotic nor had previous H. pylori eradication treatment, a strong risk factor for the development of acquired resistance. Another possible explanation for the low resistance of H. pylori isolates as compared to other Asian countries, is the difficulty procuring antibiotics due to their restrictive cost. In this study, 70% of H. pylori positive patients were unemployed with average incomes below the poverty level, defined as income below the annual per capita poverty threshold of PHP 18,000.00 for the National Capital Region of the Philippines (2001 Philippine Health Situation, Department of Health, Manila, Philippines, ). A study of the World Health Organization's Programme for appropriate Health Care Technology (ATH) has shown a correlation between the occurrence of multi-resistant bacteria and antibiotic consumption patterns. The Philippines has the highest percentage in 1983 of antibiotic utilization among countries (including USA, Japan, United Kingdom) surveyed (>25%). However, majority of the people whether rich or poor allot minimum expenses for medical care at 2.7% and 1.2%, respectively[30]. Such paradox in resistance patterns may well be explained by the capacity of these patients to actually afford the prescribed duration of antibiotic therapy. While many Filipinos may be (mis-) guided on the appropriate choice of antibiotic therapy by media or product representatives, the cost of these drugs still limits the access to these largely economically disadvantaged group. Although the presence of primary resistance of H. pylori has been well documented in other studies, the absence of primary resistance in our results may also be an underestimate of the true prevalence of H. pylori resistance because of the smaller sample size compared with published literature[2,20,25,26]. Only with continued surveillance of susceptibility patterns and a larger sample size of isolates will provide a more substantial answer to the issue of resistance of H. pylori in the Philippines.
Knowledge of the antibiotic susceptibility patterns in our setting allows us to be more cautious in the choice of first-line agents. The use of culture technique in the diagnosis of H. pylori infection approximates that in published literature abroad. In the absence of standard disk diffusion zone sizes for regimens used in H. pylori eradication regimen except for metronidazole, further establishment of the susceptibility pattern of locally occurring isolates by comparing zone size breakpoints with Etest, agar dilution method and as well as molecular genotyping of resistant strains will be the future direction of this pilot study.
Conclusions
While the use of culture is not an ideal test for the rapid diagnosis of H. pylori infection, information on antibiotic susceptibility profile plays an important role in empiric antibiotic treatment and management of refractive cases
Authors' contributions
RVD, LJB, CSA contributed in the microbiologic isolation of H. pylori, manuscript writing and editing
EDL, MOD, VD contributed in the specimen processing of biopsy samples, manuscript writing and editing
RLG contributed in the research design planning, manuscript content and final editing
Table 1 Sensitivity study Helicobacter pylori isolates*
ISOLATE NO. AMOXICILLIN TETRACYCLINE METRONIDAZOLE CLARITHROMYCIN
*Etest Resistance Breakpoints
> 0.5 ug/ml > 2 ug/ml > 8 ug/ml > 1 ug/ml
1 0.016 0.032 0.016 0.016
2 0.016 0.19 0.125 0.016
3 0.016 0.016 0.016 0.016
4 0.016 0.38 0.064 0.016
5 0.016 0.38 0.064 0.016
6 0.016 0.19 0.125 0.016
7 0.016 0.032 0.016 0.016
8 0.016 0.38 0.016 0.016
9 0.016 0.032 0.064 0.016
10 0.016 0.19 0.016 0.016
11 0.016 0.38 0.016 0.016
12 0.016 0.032 0.125 0.016
13 0.016 0.032 0.064 0.016
14 0.016 0.032 0.125 0.016
Strain # CP2124§ 0.015 0.015 33 0.015
Strain # CP5535§ 0.015 0.125 1.5 257
* all Etest studies done on the Philippine isolates were in the sensitive ranges
§UVa (University of Virginia) Metronidazole resistant isolate
¥UVa Clarithromycin resistant isolate
Figure 1 Comparative positivity of diagnostic tests: RUT (rapid urease test) and histology to H. pylori culture. All patients underwent biopsy, RUT and culture studies. Among the 31 clinico-laboratory defined cases, 14 were culture positive. Ten of the H. pylori culture positive cases were also positive for both histology and RUT. All 14 culture positive isolates tested positive for RUT while only 10 were positive for histology.
Acknowledgements
We would like to extend our gratitude to Dr. Myrna T. Mendoza and Mrs. Concepcion Ang of the Section of Infectious Disease Microbiology Research Laboratory, and Section of Gastroenterology Research laboratory and Endoscopy Unit of the University of the Philippines-Philippine General Hospital, for the valuable assistance with specimen collection and processing.
Dr. Destura is supported by the NIH/FIC ITRIED Research Fellowship award (ITREID: Grant number: 5-043-TW000909-05) of the Center for Global Health, University of Virginia.
==== Refs
Go MF Treatment and management of Helicobacter pylori infection Curr Gastroenterol Rep 2002 4 471 477 12441037
Poon SK Chang CS Su J Lai CH Yang CC Chen GH Wang WC Primary resistance to antibiotics and its clinical impact on the efficacy of Helicobacter pylori lansoprazole-based triple therapies Aliment Pharmacol Ther 2002 16 291 296 11860412 10.1046/j.1365-2036.2002.01184.x
Glupczynski Y Megraud F Lopez-Brea M Andersen LP European multicentre survey of in vitro antimicrobial resistance in Helicobacter pylori Eur J Clin Microbiol Infect Dis 2001 20 820 823 11783701 10.1007/s100960100611
Go MF Review article: natural history and epidemiology of Helicobacter pylori infection Aliment Pharmacol Ther 2002 16 Suppl 1 3 15 11849122 10.1046/j.1365-2036.2002.0160s1003.x
Moayyedi P Axon AT Feltbower R Duffett S Crocombe W Braunholtz D Richards ID Dowell AC Forman D Relation of adult lifestyle and socioeconomic factors to the prevalence of Helicobacter pylori infection Int J Epidemiol 2002 31 624 631 12055165 10.1093/ije/31.3.624
Adeyemi EO Danial MF Helal T Benedict S Abdulle AM The outcome of a 2-week treatment of Helicobacter pylori-positive duodenal ulcer with omeprazole-based antibiotic regimen in a region with high metronidazole resistance rate Eur J Gastroenterol Hepatol 1999 11 1259 1263 10563537
Grove DI Koutsouridis G Cummins AG Comparison of culture, histopathology and urease testing for the diagnosis of Helicobacter pylori gastritis and susceptibility to amoxycillin, clarithromycin, metronidazole and tetracycline Pathology 1998 30 183 187 9643503
E B and S F Bailey and Scott's Diagnostic Microbiology 1990 8 The C.V. Mosby Company 440 440
National Committee for Clinical Laboratory Standards: Performance standards for antimicrobial susceptibility testing. Approved standard M7-A5. Informational supplement M100-S10. NCCLS, Wayne, PA 2000
Megraud F Lehn N Lind T Bayerdorffer E O'Morain C Spiller R Unge P van Zanten SV Wrangstadh M Burman CF Antimicrobial susceptibility testing of Helicobacter pylori in a large multicenter trial: the MACH 2 study Antimicrob Agents Chemother 1999 43 2747 2752 10543758
Tolia V Brown W El Baba M Lin CH Helicobacter pylori culture and antimicrobial susceptibility from pediatric patients in Michigan Pediatr Infect Dis J 2000 19 1167 1171 11144378
Kwon DH Kim JJ Lee M Yamaoka Y Kato M Osato MS El Zaatari FA Graham DY Isolation and characterization of tetracycline-resistant clinical isolates of Helicobacter pylori Antimicrob Agents Chemother 2000 44 3203 3205 11036054 10.1128/AAC.44.11.3203-3205.2000
Ozcay F Kocak N Temizel IN Demir H Ozen H Yuce A Gurakan F Helicobacter pylori infection in Turkish children: comparison of diagnostic tests, evaluation of eradication rate, and changes in symptoms after eradication Helicobacter 2004 9 242 248 15165260 10.1111/j.1083-4389.2004.00230.x
Fabre R Sobhani I Laurent-Puig P Hedef N Yazigi N Vissuzaine C Rodde I Potet F Mignon M Etienne JP . Polymerase chain reaction assay for the detection of Helicobacter pylori in gastric biopsy specimens: comparison with culture, rapid urease test, and histopathological tests Gut 1994 35 905 908 8063217
Weiss J Mecca J da Silva E Gassner D Comparison of PCR and other diagnostic techniques for detection of Helicobacter pylori infection in dyspeptic patients J Clin Microbiol 1994 32 1663 1668 7929755
Lage AP Godfroid E Fauconnier A Burette A Butzler JP Bollen A Glupczynski Y Diagnosis of Helicobacter pylori infection by PCR: comparison with other invasive techniques and detection of cagA gene in gastric biopsy specimens J Clin Microbiol 1995 33 2752 2756 8567918
Montes H Salmen S Dolfo W Sotolongo A Petrosino P Donis J Berrueta L Evaluation of a liquid urease test (LUT) for detection of Helicobacter pylori Acta Gastroenterol Latinoam 2003 33 73 76 14708499
Kuo CH Wu DC Lu CY Su YC Yu FJ Lee YC Wu IC Lin SR Liu CS Jan CM Wang WM The media of rapid urease test influence the diagnosis of Helicobacter pylori Hepatogastroenterology 2002 49 1191 1194 12239903
Ikezawa K Kashimura H Kojima M Aikawa T Nakahara A Mutoh H Tanaka N Pretreatment antimicrobial susceptibilities of paired gastric Helicobacter pylori isolates: antrum versus corpus Helicobacter 1999 4 218 221 10597389 10.1046/j.1523-5378.1999.99051.x
Kato S Fujimura S Udagawa H Shimizu T Maisawa S Ozawa K Iinuma K Antibiotic resistance of Helicobacter pylori strains in Japanese children J Clin Microbiol 2002 40 649 653 11825987 10.1128/JCM.40.2.649-653.2002
Kim JJ Reddy R Lee M Kim JG El Zaatari FA Osato MS Graham DY Kwon DH Analysis of metronidazole, clarithromycin and tetracycline resistance of Helicobacter pylori isolates from Korea J Antimicrob Chemother 2001 47 459 461 11266421 10.1093/jac/47.4.459
Meyer JM Silliman NP Wang W Siepman NY Sugg JE Morris D Zhang J Bhattacharyya H King EC Hopkins RJ Risk factors for Helicobacter pylori resistance in the United States: the surveillance of H. pylori antimicrobial resistance partnership (SHARP) study, 1993-1999 Ann Intern Med 2002 136 13 24 11777360
Miyaji H Azuma T Ito S Suto H Ito Y Yamazaki Y Sato F Hirai M Kuriyama M Kato T Kohli Y Susceptibility of Helicobacter pylori isolates to metronidazole, clarithromycin and amoxycillin in vitro and in clinical treatment in Japan Aliment Pharmacol Ther 1997 11 1131 1136 9663841 10.1046/j.1365-2036.1997.00258.x
Pilotto A Rassu M Leandro G Franceschi M Di Mario F Prevalence of Helicobacter pylori resistance to antibiotics in Northeast Italy: a multicentre study. GISU. Interdisciplinary Group for the Study of Ulcer Dig Liver Dis 2000 32 763 768 11215555 10.1016/S1590-8658(00)80352-7
Wolle K Leodolter A Malfertheiner P Konig W Antibiotic susceptibility of Helicobacter pylori in Germany: stable primary resistance from 1995 to 2000 J Med Microbiol 2002 51 705 709 12171304
Vicente R Sicilia B Gallego S Revillo MJ Ducons J Gomollon F [Helicobacter pylori eradication in patients with peptic ulcer after two treatment failures: a prospective culture-guided study] Gastroenterol Hepatol 2002 25 438 442 12139836
Wong WM Wong BC Lu H Gu Q Yin Y Wang WH Fung FM Lai KC Xia HH Xiao SD Lam SK One-week omeprazole, furazolidone and amoxicillin rescue therapy after failure of Helicobacter pylori eradication with standard triple therapies Aliment Pharmacol Ther 2002 16 793 798 11929398 10.1046/j.1365-2036.2002.01223.x
Lui SY Yeoh KG Ho B Metronidazole-resistant Helicobacter pylori is more prevalent in patients with nonulcer dyspepsia than in peptic ulcer patients in a multiethnic Asian population J Clin Microbiol 2003 41 5011 5014 14605132 10.1128/JCM.41.11.5011-5014.2003
A. B The antibiotic crisis. In problem drugs Amsterdam Health Action International 1993 51 68
ER D Inappropriate Antibiotic Use in the Philippines Phil J Microbiol Infect 1997 26 77 87
| 15546485 | PMC543444 | CC BY | 2021-01-04 16:38:17 | no | Ann Clin Microbiol Antimicrob. 2004 Nov 16; 3:25 | utf-8 | Ann Clin Microbiol Antimicrob | 2,004 | 10.1186/1476-0711-3-25 | oa_comm |
Subsets and Splits